International Journal of Smart Electrical Engineering-8

187
International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

pp.187:193
ISSN: 2251-9246
EISSN: 2345-6221
Unit Commitment in Presence of Wind Power Plants and

Energy Storage
Reza Khanzadeh1, Mahmoud Reza Haghifam2
1
Electrical Engineering Department, South Tehran Branch, Islamic Azad University, Tehran, Iran. Email: khanzadehreza@yahoo.com
2
Electrical Engineering Department, South Tehran Branch, Islamic Azad University, Tehran, Iran. Email: haghifam@ieee.com
Abstract
As renewable energy increasingly penetrates into power grid systems, new challenges arise for system operators to keep the
systems reliable under uncertain circumstances, while ensuring high utilization of renewable energy. This paper presents unit
commitment (UC) which takes into account the volatility of wind power generation. The UC problem is solved with the
forecasted intermittent wind power generation and possible scenarios are simulated for representing the wind power
volatility. The iterative process between the commitment problem and the economic dispatch(ED) problem will continue
until we find the optimum mode of committing the units. Furthermore we have considered a hydro pump storage (HPS) unit
to be a part of operating system in order to mitigating wind power forecasting errors and peak shaving. Numerical
simulations indicate the effectiveness of the proposed UC for managing the security of power system operation by taking into
account the intermittency and volatility of wind power generation.
Keywords: Unit Commitment, Economic Dispatch, Wind Power, Hydro Pump Storage Unit, Mont Carlo Simulation.
2013 IAUCTB-IJSEE Science. All rights reserved
Nomenclatures
Index
NG
NW
NH
NS
i
s
t
T
Binary
Variables
Ii,t
Yi,t
Si,t
IGHt
IPHt
IIHt
Number of thermal units

Number of wind units
Number of pump storage units
Number of scenarios
Power generation unit index
Scenario index
Time index
Time horizon
Commitment status of unit i at time t

If unit i starts up at time t , is equal to 1
If unit i shuts down at time t , is equal to 1
HPS generating mode decision variable

HPS pumping mode decision variable
HPS inactive mode decision variable
Parameters
SUi,t
SDi,t
Pi,min
Pi,max
RUi
RDi
Zbegin
Zlast
Vh,max
Vh,min
A1
A2
Lh,tin
Uh,tin
Lh,tout
Uh,tout
SGCh,t
SPCh,t
Startup cost of unit i at time t

Shutdown cost of unit i at time t
Lower limit of real power generation of unit i
Upper limit of real power generation of unit i
Ramp up rate of unit i

Ramp down rate of unit i
Initial water reserve inventory of HPS
Target water reserve inventory of HPS
Upper limit of reservoir volume in HPS
Lower limit of reservoir volume in HPS
Efficiency of pumping cycle of HPS
Efficiency of generating cycle of HPS
Lower limit of consumed power by HPS
Upper limit of consumed power by HPS
Lower limit of generated power by HPS
Upper limit of generated power by HPS
Srarting to generate , cost of HPS
Starting to pump , cost of HPS
188
Tion
Tioff
Variables
Pi,t
Psw,i,t
Pinh,i,t
Pouth,i,t
Dts
Rs,i,t
Rs,t
Zt
Vh,t
Xoni,t
Xoffi,t
Function
FCi,t

Minimum up time of unit i
Minimum down time of unit i
Generation of unit i at time t

Generation of wind unit i at time t in scenario s
Absorbed power by HPS i at time t

Generated power by HPS i at time t
Demand at time t in scenario s
Spining reserve prepeared by unit i at time t
System spinning reserve requirement at time t
Water reserve inventory at time t

Water volume in HPS at time t
Operating duration of unit i at time t
Shutdown duration at time t
Generating cost of thermal unit i at time t
1. Introduction
Wind energy has become increasingly popular
across the globe. It is reported by the Global Wind
Energy Council (GWEC) that global wind energy
installations rose by 11 531 MW in 2005, which
represent an annual increase of 40.5% [1]. Such
figures demonstrate the prosperous future of wind
power development. However, the intermittent and
volatile nature of wind power generation may impact
power system characteristics such as voltages,
frequency and generation adequacy which can
potentially increase the vulnerability of power
systems. Intermittency refers to the unavailability of
wind for an extended period and volatility refers to
the smaller and hourly fluctuations of wind within its
intermittent characteristics. The cumulative wind
power (representing several wind farms) in a power
system might not be intermittent. However, the
power output of a single wind farm could be
intermittent within a 24-h period. The intermittency
of individual wind farms is considered in the
proposed UC in order to ensure that prevailing
constraints are satisfied.
There are several techniques for predicting the
quantity of intermittent wind power [2], [3].Wind
forecasting is conducted by simulation, statistical
method, or a combination of the two. The simulation
method is based on a large number of wind scenarios
and starts by a numerical weather prediction (NWP)
followed by local wind pattern predictions using
analytical methods. The statistical method also starts
from NWP followed by statistical, artificial neural
network, or fuzzy logic methods instead of analytical
methods for calculating the hourly quantity of
intermittent wind power, in which large data sets are
needed and spikes in wind data are hard to predict
[4]. Although wind power is predictable to a limited
extent, it cannot be forecasted with 100% accuracy
for dispatching purposes. Hence, it is possible that
ISSN: 2251-9246
EISSN: 2345-6221
the actual wind power would be different from its

forecasted value. The uncertainty is characterized in
this paper by considering the volatility in multiple
scenarios.
The wind power forecasting and associated
forecasting accuracy issues are important in
analyzing the impact of wind power on power system
operation. However, a complete discussion on wind
speed, wind forecast, and wind power data analyses
is beyond the scope of this paper, and deserves
another full paper. Furthermore, the modeling of load
forecasting error (load profile) is also performed in
this work . Other uncertainties such as the modeling
of generation and transmission outages are also
important subjects for power system operation which
are beyond the scope of this paper.
Wind farms could be managed by utility
companies in which the real-time on/off status of
non-wind units would be decided based on the hourly
load behavior and the availability of intermittent
wind power. However, in certain parts of the United
States, the intermittency of wind could amount to
several hundred megawatts in a matter of hours.
Likewise, the volatility of wind power could have a
tremendous impact on power system operations
which poses new challenges for the electricity market
management. Control room operators and ISOs in
competitive electricity markets apply optimization
methods for managing the security of the system
while utilizing the merits of wind power generation
[5][7].
In [8] the impact of intermittent wind generation
on the operations of the Tennessee Valley Authority
(TVA) power system is investigated and the
operations of the TVA power system are outlined. In
reference [9] authors have presented a new method
for solving efficiently a large scale optimal unit
commitment problem that was included three types
of units (i) usual thermal units (ii) fuel constrained
thermal units and (iii) pumped storage hydro units .
The solution method in this paper uses a lagrangian
relaxation. A new simulation method that could fully
assess the impacts of large-scale wind power on
system operations from cost, reliability, and
environmental perspectives was introduced in [10].
For coordinating the wind and thermal generation
scheduling problem a hybrid approach of combining
branch and bound algorithm with a dynamic
programming algorithm was developed in [11]. In
[12] a security constrained unit commitment (SCUC)
algorithm which takes into account the volatility of
wind power generation is proposed. A stochastic cost
model and a solution technique for optimal
scheduling of the generators in a wind integrated
power system considering the demand and wind
generation uncertainties is presented in [13]. In [14]
the effects of stochastic wind and load on the unit
commitment and dispatch of power systems with
189
high levels of wind power is examined and showed

that stochastic optimization results in less production
cost and better performing schedules than
deterministic
optimization.
A computational
framework for integrating the state-of-the-art
numerical weather prediction (NWP) model in
stochastic UC/ED formulations is proposed in [15]
that accounts for wind power uncertainty. In [16]
authors have presented an efficient formulation of the
stochastic unit commitment problem that is designed
for use in scheduling simulations of single-bus power
systems. A robust optimization approach for
accommodating wind output uncertainty is proposed
in [17] that aims to providing a robust unit
commitment schedule for the thermal generators in
the day ahead market that minimizes the total cost
under the worst wind power scenario. In [18] a
stochastic dynamic programming approach to unit
commitment and dispatch has proposed that
minimizes the operating cost by making optimal unit
commitment, dispatch and storage decisions in the
face of uncertain wind generation. A novel approach
to the security constrained unit commitment with
uncertain wind power generation is presented in [19]
that its goal is to solve the problem considering
multiple stochastic wind power scenarios but while
significantly reducing the computational burden
associated with the calculation of the reserve
deployment for each scenario. In [20] a robust
optimization approach is developed to derive an
optimal unit commitment decision for the reliability
unit commitment runs by ISOs/RTOs with the
objective of maximizing total social welfare under
the joint worst-case wind power output and demand
response scenario.
The rest of this paper is organized as follows.
Section 2 presents the uncertainty modeling
technique. Section 3 proposes the formulation of the
problem and the solution methodology. One case
study is studied in section 4. Section 5 concludes the
discussion.
2. Uncertainty Modeling Technique
In order to taking into account wind power and
demand forecasting uncertainty, we use an
uncertainty modeling technique that is based on
scenario generation for uncertain parameter. In this
approach we use monte carlo simulation technique to
generate a large number of scenarios subject to a
normal distribution of forecasting errors that have
engendered in the past predictions. Since the number
of scenarios is very large, using all of those scenarios
in solving progress increases the computational
burden of our problem. Therefore, reducing the
number of scenarios is one of the necessities. So,
scenario generation and reduction methods are as
follows:
ISSN: 2251-9246
EISSN: 2345-6221
2.1. Scenario Generation

For scenario generation, first we have to
calculate the forecasting errors that have occurred in
the past wind power and demand predictions and
assume that they are subject to a normal or other
statistical distribution with an expected value () and
a percentage of as its volatility (). Then, using
monte carlo sampling technique, monte carlo paths
will create by sampling from this distribution and
juxtapose them. Now for constructing possible
scenarios we must add the obtained samples to the
predictions for next 24 hour, each scenario is
assigned an occurrence probability.
2.2. Scenario Reduction
The scenario reduction technique is employed
to decrease the number of obtained scenarios.
Scenario reduction will remove scenarios that have
low occurrence probability and conjunct those
scenarios that are the same as each other, in one
scenario [21], [22]. By reducing the number of
scenarios consequently the computational burden and
time will decrease remarkably.
3. Problem Formulation And Solution
Methodology
3.1. Stochastic Programming
In many situations there is a need to make, an
optimal, decision under conditions of uncertainty.
There is a disagreement, however, with how to deal
with such situations. Uncertainty can come in many
different forms, and hence there are various ways
how it can be modeled. In a mathematical approach
one formulates an objective function f: Rn R
which should be minimized subject to specified
constraints. That is, one formulates a mathematical
programming problem:
Min f (x),
(1)
xX
Where the feasible set X Rn is typically

defined by a (finite or even infinite) number of
constraints, say X := {x Rn : gi(x) 0, i I} (the
notation := means equal by definition).
Inevitably the objective and constraint functions
depend on parameters, which we denote by vector
Rd. That is, f (x, ) and gi(x, ), i I, can be
viewed as functions of the decision vector x Rn and
parameter vector Rd.
3.2. Problem Formulation
We formulate the UC problem in presence of
wind power plants and hydro pump storage unit in
(2) (17) as a stochastic optimization problem. The
objective function (2) consists of generator operating
190
cost, start up and shutdown costs of thermal power

plants and the cost of staring to generate or absorb
power by pump storage unit. The constraints in our
UC problem including system constraints, thermal
power plant constraints, wind power plant constraints
and pump storage unit constraints are as follows.
System power balance constraint (3), system
spinning reserve requirements (4), unit generation
limits (5), unit minimum up time (6), unit minimum
down time (7), unit ramping up limits (8), unit
ramping down limits (9), unit initial state (10), hydro
water inventory constraints (11), constraints (12) and
(13) describe the upper and lower bounds of
electricity absorbed and generated by the pumpedstorage unit , constraints (14) and (15) give the initial
and target water inventory level for the pumpedstorage unit , water reservoir volume limit in pump
storage unit constraint (16) and finally constraint (17)
that ensures that the pumped-storage unit cannot
absorb and generate electricity simultaneously within
any specific time period .
NG T
i 1 t 1
NG
P
i 1
NG
i ,t
R
i 1
h 1 t 1
NW
i 1
As a case study we have considered a single

bus system that includes 6 thermal units , 2 wind
units and one hydro pump storage unit that have
shown in fig.1 . The characteristics of thermal units
and pump storage unit are presented in table.1 and
table.2 respectively.
Fig.1. Studied system

Table.1
Generator data
(2)
(3)
* I i ,t RS ,t
(4)
(5)
X ion,(t 1) Ti on * Ii ,(t 1) Ii ,t 0
(6)
off
X ioff
* Ii ,t Ii ,(t 1) 0
,( t 1) Ti
(7)
Pi ,t Pi ,(t 1) (2 Ii ,(t 1) Ii ,t ) Pi ,min (1 Ii ,(t 1) Ii ,t ) RUi
(8)
Pi ,(t 1) Pi ,t (2 Ii ,(t 1) Ii ,t ) Pi ,min (1 Ii ,(t 1) Ii ,t ) RDi
Yi ,t Si ,t I i ,t I i ,(t 1)
Zt Z (t 1) A1 Phin,t
L P U
in
h ,t
Unit
Pmin
(MW)
Pmax
(MW)
Min
ON
(H)
Min
Off
(H)
Ramp
Up
(MW/H)
Ramp
Down
(MW/H)
IniT
(H)
G1
100
580
10
10
250
250
10
G2
100
450
210
210
G3
100
380
175
175
G4
100
330
150
150
G5
100
300
150
150
G6
25
100
50
50
h 1
Pi ,min Ii ,t Pi ,t Pi ,max * Ii ,t
in
h ,t
4.1. Applying on a Sample System
NH
in
s
* I i ,t Pws,i ,t Phout
,i ,t Ph ,i ,t Dt
S ,i ,t
4. Case Study
NH T
min Fc ,i ( Pi ,t )* I i ,t SU i ,t SDi ,t SGCh,t SPCh,t
A2
(a)
Unit
St
Mbtu
Fuel
Price
($/Mbtu)
af
(Mbtu/MW2h)
bf
(Mbtu/MWh)
cf
(Mbtu/h)
G1
300
0.0109
8.6
70
G2
250
0.01059
8.3391
64.16
G3
100
0.003
10.76
32.96
G4
440
0.01088
12.8875
6.78
G5
100
0.01088
12.8875
6.78
G6
50
0.0128
17.82
10.15
(9)
(10)
out
h ,t
ISSN: 2251-9246
EISSN: 2345-6221
(11)
in
h ,t
(12)
out
out
Lout
h ,t Ph ,t U h ,t
(13)
Z0 Zbegin
(14)
ZT Zlast
(15)
Vh,min Vh,t Vh,max
(16)
IGH h,t IPH h,t IIH h,t 1
(17)
(b)
Each of the wind farms and required demand

also has a predicted value and some scenarios for
modelling uncertainty of the forecasted quantities.
Figures 2, 3 and 4 show characteristics of wind unit
1, wind unit 2 and demand, respectively.
191

Table.2
Pump storage unit data
Unit
Pump
Cycle
Eff
Gen
Cycle
Eff
Max
Gen
Lim
(MW)
Min
Gen
Lim
(MW)
Max
Abs
Lim
(MW)
Min
Abs
Lim
(MW)
Min
ON
(H)
Min
OFF
(H)
0.8
0.8
40
40
Uphill Reservoir
Unit
Ini
Vol
(Hm3)
Tgt
Vol
(Hm3)
Up
Lim
Vol
(Hm3)
180
60
250
Downhill Reservoir
Low
Lim
Vol
(Hm3)
Ini
Vol
(Hm3)
Up
Lim
Vol
(Hm3)
Low
Lim
Vol
(Hm3)
Gen
And
pump
St
Cost($)
50
380
600
200
75
ISSN: 2251-9246
EISSN: 2345-6221
GAMS24.1.3 and its CPLEX solver to solve the

Mixed Integer Program (MIP) that proposed in
section 3.2.
After solving the problem we must choose the
most optimum schedule of committing units in order
to minimize the production cost. The output results
of the program showed that the cost of the power
production in the considered day is 353834.056$ and
allocated power to each unit can be showed like
table.3.
Table.3
Commitment and dispatch of thermal and pump storage units
Fig.2. Forecasted power and scenarios for wind1
Hour
G1
(MW)
G2
(MW)
G3
(MW)
G4
(MW)
G5
(MW)
G6
(MW)
H
OUT
(MW)
H
IN
(MW)
460.4
25
246.6
25
113.6
25
100
8.4
118.5
172.4
422.4
165.8
562.524
100
175
21.376
580
139
350
10
580
181.6
380
100
11
580
233.5
380
100
12
580
271.7
380
100
25
13
580
255.5
380
100
25
40
14
580
323.3
380
100
25
15
580
242.9
380
100
16
580
293.9
380
100
17
580
276.7
380
100
18
580
293
380
100
19
580
340.2
380
100
100
20
580
450
350
100
106.8
21
580
450
175
100
136.5
22
500
450
100
133.9
23
250
405.49
100
40
24
364.5
100
Fig.3. Forecasted power and scenarios for wind2
Fig.4. Forecasted demand and scenarios
By means of the introduced system we solve the

UC and ED problem. In this paper we have used
In order to ensure system reliability and

security we have allocated spinning reserve for each
hour. The spinning reserve is provided by thermal
units by means of committing those thermal units in
each hour that the sum of their maximum production
capacity is greater than or equal by 1.15*Demand
in that hour.
In table.3 we can see that the pump storage
unit , in the period of times that demand is low and
wind production is high , treats like a load and
absorbs power to pump water from downhill
reservoir to uphill reservoir that . This work not only
causes an increase in energy storage but also
decreases the need to wind curtailment. Moreover
this unit has used for peak shaving in times that a
192
transition peak load has occurred and thus prevents

from more start up in thermal units.
It is obvious that if the capacity of the pump
storage unit be different from the value that we have
used, its commitment will be different too. Figures 5
and 6 show the change in uphill and downhill water
volume that have engendered because of the
generating and absorbing power during the day. It is
clear that the water volume had not transgressed
from its limits both in downhill and uphill reservoirs
and the volume in uphill reservoir have reached to
the predefined value at the end of the day.
ISSN: 2251-9246
EISSN: 2345-6221
It is sensible in fig.7 that the production cost at

first decreases gradually as the maximum capacity of
pump storage unit increases, but after a specific
capacity the change in cost is more remarkable. It is
why by increasing the capability of pump storage
unit to take part in supplying the load and saving
energy, the efficiency of it, is also increased.
Besides, Fig.8 shows that if the volume of the
water that exists in the uphill reservoir at the
beginning of the day be more than we considered, the
production cost will also less than we obtained from
the implemented volume as the initial water volume
for the uphill reservoir of the hydro pump storage
unit.
Fig.5. Uphill reserve inventory changes

Fig.7. Cost change by varying maximum capacity of the pump
storage unit
Fig.8. Cost change by varying initial water level in uphill reservoir

Fig.6. Downhill reserve inventory changes
4.2. Sensitivity Analysis

In order to evaluate the effect of some
parameters on the problem we have done sensitivity
analysis. By this goal we solved the problem without
wind power plants and pump storage unit. Results
show that this change increases the production cost
to 394171.138$ that is equal to 11.4% increase and
moreover the aggregate operating hour of units was
increased too.
We have also examined the effect of changes in
pump storage unit capacity and initial water volume
in uphill reservoir on the production cost. Figures 7
and 8 show the change in production cost by varying
the pump storage maximum capacity and initial
water volume in the uphill reservoir, respectively.
5. Conclusion
In this paper, we proposed an approach that
includes applying optimization concepts and
incorporating pumped-storage units to hedge wind
power output uncertainty and peak shaving. We
provided scenarios that can capture the wind power
unpredicted changes and our proposed approach can
provide an optimal solution that minimizes the total
cost under the wind power fluctuations that can occur
in the system, while ensuring the higher penetration
of wind power. Meanwhile, this solution is feasible
with a high probability under wind power output
uncertainty. In addition, by incorporating pumped
storage hydro units in the real time, our optimization
model contains discrete decision variables in
problems. Finally, our computational results verify
the effectiveness of the presence of wind units and
193
pump storage unit for the system and power

production cost.
References
[1] Global Wind Energy Council. [Online].
Available:http://www.gwec.net/.
[2] G. Sideratos and N. D. Hatziargyriou, An Advanced
Statistical Method for Wind Power Forecasting, IEEE Trans.
Power Syst., Vol.22, No.1, pp.258265, Feb. 2007.
[3] G. Kariniotakis, I. H.-P. Waldl, I. Marti, G. Giebel, T. S.
Nielsen, J. Tambke, J. Usaola, F. Dierich, A. Bocquet, and S.
Virlot, Next Generation Forecasting Tools for the Optimal
Management of Wind Generation, in Proc. Int. Conf.
Probabilistic Methods Applied to Power Systems (PMAPS
2006), Jun. 1115, 2006.
[4] J. Soens, Impact of Wind Energy in a Future Power Grid,
Ph.D. dissertation, Katholieke Univ. Leuven, Leuven,
Belgium, Dec. 2005.
[5] True Wind Solutions, LLC and AWS Scientific, Inc.,
Overview of Wind Energy Generation Forecasting. [Online].
Available:
www.uwig.org/forecst_overview_report_dec_2003.pdf.
[6] W. El-Khattam, Y. Hegazy, and M. Salama, Uncertain Power
Flow Analysis of Electrical Distributed Generation Systems,
in Proc. IEEE Power Eng. Soc. General Meeting, pp.1141
1144, 2003.
ISSN: 2251-9246
EISSN: 2345-6221
[13] Venkata Swaroop Pappala, Istvan Erlich, Kurt Rohrig, Jan

Dobschinski, A Stochastic Model for the Optimal Operation
of a Wind-Thermal Power System, IEEE Trans on Power
Systems, Vol.24, No.2, May 2009.
[14] Aidan Tuohy, Peter Meibom, Eleanor Denny, Mark OMalley
Unit Commitment for Systems With Significant Wind
Penetration, IEEE Trans on Power Systems, Vol.24, No.2,
May 2009.
[15] Emil M. Constantinescu, Victor M. Zavala, Matthew Rocklin,
Sangmin Lee, and Mihai Anitescu, A Computational
Framework for Uncertainty Quantification and Stochastic
Optimization in Unit Commitment With Wind Power
Generation, IEEE Trans on Power Systems, Vol.26, No.1,
February 2011.
[16] Alexander Sturt and Goran Strbac, Efficient Stochastic
Scheduling for Simulation of Wind-Integrated Power
Systems, IEEE Trans on Power Systems, Vol.27, No.1,
February 2012.
[17] Ruiwei Jiang, Jianhui Wang, Yongpei Guan, Robust Unit
Commitment With Wind Power and Pumped Storage Hydro
IEEE Trans on Power Systems, Vol.27, No.2, May 2012.
[18] Jeremy J. Hargreaves and Benjamin F. Hobbs Commitment
and Dispatch With Uncertain Wind Generation by Dynamic
Programming, IEEE Trans on Sustainable Energy, Vol.3,
No.4, October 2012.
[7] J. Liu, M. Salama, and R. Mansour, Identify the Impact of

Distributed Resources on Congestion Management, IEEE
Trans. Power Del., Vol.20, No.3, pp.19982005, Jul. 2005.
[19] Amir Kalantari, Jose F. Restrepo, FranciscoD. Galiana,

Security-Constrained Unit Commitment With Uncertain
Wind Generation: The Loadability Set Approach, IEEE
Trans on Power Systems, Vol.28, No. 2, May 2013.
[8] N. D. Sadanandan, D. W. Hilson, Impact Assessment of

Wind Generation on the Operations of a Power System IEEE
Transactions on Power Apparatus and Systems, Vol. -PAS102, No. 9, September 1983.
[20] Chaoyue Zhao, Jianhui Wang, Jean-Paul Watson, Yongpei

Guan, Multi-Stage Robust Unit Commitment Considering
Wind and Demand Response Uncertainties IEEE Trans on
Power Systems, Vol.28, No.3, August 2013.
[9] K. Aoki, T. Satoh, M. Itoh, Unit Commitment in a LargeScale Power System Including Fuel Constrained Thermal and
Pumped-Storage Hydro, IEEE Transactions on Power
Systems, Vol. PWRS-2, No.4, November 1987.
[21] J. Dupacov, N. Grwe-Kuska, and W. Rmisch, Scenario

reduction in stochastic programming: An approach using
probability metrics, Math. Program. Series A, vol. 3, pp.
493511, 2003.
[10] Bart C. Ummels , Madeleine Gibescu, Engbert Pelgrum, Wil

L. Kling, Arno J. Brand, Impacts of Wind Power on Thermal
Generation Unit Commitment and Dispatch, IEEE Trans on
Energy Conversion, Vol.22, No.1, March 2007
[22] N. Grwe-Kuska, H. Heitsch, and W. Rmisch, Scenario

reduction and scenario tree construction for power
management problems, in Proc. IEEE Power Tech Conf.,
Bologna, Italy, Vol.3, pp.2326, Jun. 2003.
[11] Chun-Lung Chen, Optimal WindThermal Generating Unit

Commitment, IEEE Trans on Energy Conversion, Vol.23,
No.1, March 2008.
[12] Jianhui Wang, Mohammad Shahidehpour, Zuyi Li, SecurityConstrained Unit Commitment With Volatile Wind Power
Generation, IEEE Trans on Power Systems, Vol.23, No.3,
August 2008.
195

pp.195:200
ISSN: 2251-9246
EISSN: 2345-6221
Maximum Power Point Tracking of Wind Energy Conversion

System using Fuzzy- Cuckoo Optimization Algorithm
Strategy
Mohammad Sarvi1, Mohammad Parpaei2
1
2
Electrical Engineering Department, Imam Khomeini International University, Qazvin, Iran. Email: sarvi@eng.ikiu.ac.ir
Electrical Engineering Department, Imam Khomeini International University, Qazvin, Iran. Email: parpaei@edu.ikiu.ac.ir
Abstract
Nowadays the position of the renewable energy is so important because of the environment pollution and the limitation of
fossil fuels in the world. Energy can be generated more and more by the renewable sources, but the fossil fuels are nonrenewable. One of the most important renewable sources is the wind energy. The wind energy is an appropriate alternative
source of fossil fuel. The replacement rate of renewable energy to fossil fuels is rising, although the production cost is higher
than fossil fuels. To further reduce cost of wind production, many methods have been proposed. One of the suitable
approaches is the maximum power point tracking strategy. In this paper, a new intelligent maximum power point tracker
called Fuzzy- Cuckoo strategy for smallscale wind energy conversion systems is proposed. The maximum power point
tracker proposed uses measured wind speed to detect the maximum output power and its respective optimal rotational speed.
The main contribution of the proposed approach is to exactly track the maximum power point, so the output power
fluctuations captured by wind turbine are less than conventional approaches. The simulations are performed in
MATAL/SIMULINK software. The superiority of the proposed approach is validated in two situations, low and rapid
changes in wind speed. The maximum power point of wind energy conversion systems can be tracked by the proposed
approach in any situation. The higher accuracy of the Fuzzy- cuckoo strategy than the conventional trackers is another
advantage of the proposed approach.
Keywords: Intelligent controller, Metaheuristic optimization approach, Wind energy conversion systems.
1. Introduction
Fossil fuel reserves reduction causes that the
whole countries, especially the countries that have
not enough fossil fuel sources, pay special attention
to the renewable energy as a second energy source.
China and USA are two countries that concentrate on
the wind energy conversion systems (WECS) than
other countries. WECSs are used to change the wind
energy to electrical energy by electrical machines
such as the permanent magnet synchronies
generators (PMSGs). The small- Scale WECSs are
suitable alternative sources for urban regions or
remote places that connection to power grid is
impossible [1]. The main disadvantage of the

renewable energy is that the electricity production
costs from various renewable energy sources are
higher than fossil fuel. To improve this problem,
maximum power point tracking (MPPT) is a matter
that is expressed. The maximum power point trackers
control the WECSs at the optimal output power.
There are many approaches to track the maximum
power point, but all approaches are based on three
main classifications. The first strategy is the methods
based on iteratively search, the second strategy is the
methods based on the static parameters of the wind
turbine and wind speed, and the third strategy is the
methods based on hill- climb searching (HCS)
196
control. The many approaches have been proposed to

implement the maximum power point trackers in the
WECSs. Reference [2] presents a relationship
between DC electrical variables and mechanical
variables of turbine. Reference [3] proposes an
alternative approach using mechanical model of the
permanent magnet synchronous generator (PMSG) in
order to estimate both position and speed of the wind
turbine. Reference [4] develops a new approach
using perturb and observe (P&O) and optimum
relationship based (ORB). Reference [5] presents a
modified MPPT controller using Fuzzy logic
controller (FLC). Reference [6] proposes a new
MPPT approach using neural networks (NNs) for
modeling of wind turbine systems and NNs describe
relationship between input and output. Reference [7]
introduces a novel approach of radial basis function
neural network (RBFNN) and particle swarm
optimization (PSO). Reference [8] presents a two
stages MPPT approach. In the first stage, the large
iterative steps and in the second stage, the
conventional P&O have been used to exactly track
the maximum power point. Reference [9] describes a
direct control using output observation and directly
adjusting duty cycle of boost converter. Reference
[10] develops a control strategy for the generator side
converter with output maximization of a PMSG.
Reference [11] present a new maximum power point
tracking based on adaptive neuro- fuzzy interface
system (ANFIS). Reference [12] presents a new
MPPT strategy based on Wilcoxon radial basis
function network (WRBFN) with hill-climb
searching (HCS). Reference [13] proposes a new
sensorless control strategy based on a model
reference adaptive system (MRAS) observer for
estimating the rotational speed. Reference [14]
proposes a new adaptive intelligent optimization
algorithm that uses a new advance P&O to detect the
maximum power point. Reference [15] uses a new
approach to estimate position and speed of the wind
turbine in order to track the maximum power point.
In reference [16], the rotor position is estimated
based on the flux linkage.
In this paper, a novel strategy based on an
intelligent optimization algorithm, namely Cuckoo
optimization algorithm (COA), and the fuzzy logic
controller is proposed. The proposed approach is
based on the optimum relationship- based (ORB).
The main contribution of the proposed approach is to
exactly track the maximum power point, so the
output power fluctuations captured by wind turbine
are less than the conventional approaches such as
PSO and fuzzy logic trackers. The higher accuracy of
the Fuzzy- Cuckoo strategy than the conventional
trackers is another advantage of the proposed
approach.
This paper has been organized as follows:
section 2 describes the wind energy conversion
ISSN: 2251-9246
EISSN: 2345-6221
system. Section 3 introduces the Cuckoo

optimization algorithm. Section 4 presents the
proposed maximum power point tracker. The
simulation results in several case studies are given in
Section 5, and finally section 6 states the conclusion.
2. Description of Wind Energy Conversion
System
The proposed approach will be applied to the
following WECS as Figure 1. As shown in Fig.1,
PMSG is coupled to a wind turbine. The mechanical
output power is transferred to the electrical power by
PMSG. The AC electrical power produced by PMSG
will be rectified by a three-phase diode bridge
rectifier, and the rectified power is fed to the boost
converter.
L
C dc
Co
PMSG
Fig.1. Wind energy conversion system
2.1. Wind Energy Conversion System Characteristics

The output power derived by the wind turbine
blades is expressed as [16]:
P=
1
A Cp V3
2
(18)
Where P is the output power (in watt), is the

air density (in kg/m3 ), Cp is the power coefficient,
and V is the wind speed (in m/s). A is area swept
by the blades (in m2 ) determined as:
A = R2
(19)
Where R is radius of blades (in m).

Power coefficient Cp is a nonlinear function of
the pitch angle and tip speed ratio as follows:
1
= 0.5176 (116 ( ) 0.4 5)
(20)
1
exp (21 ( )) + 0.0068
1
1
0.035
=
3
i + 0.08 + 1
=
Rt
V
(21)
(22)
Where is the tip speed ratio (TSR), is the pitch

angle (in degree), and t is the rotational speed of
turbine (in rps).
2
1.5
1
The locus of
maximum output
power
0.5
0
10
15
Rotational speed (rad/s)

Fig.2. Output power versus rotational speed
3. Cuckoo Optimization Algorithm

Cuckoo optimization algorithm (COA) is a new
intelligent evolutionary algorithm that is proper for
continuous non- linear problems. It is inspired by the
special kind of bird called Cuckoo. The superiority of
COA has been proven than particle swarm
optimization (PSO) and genetic algorithm (GA) by
[17]. The main advantage of the COA is that this
optimization algorithm is robust to dynamic changes.
The special lifestyle of Cuckoo has been formulated
as an optimization algorithm in order to fine the
maximum/minimum value of objective functions.
The following section meticulously describes the
COA.
To start the optimization process, it is necessary
to initialize the starting points as an array. Each of
the evolutionary optimization algorithms specifies a
special name for this array. For instant, in PSO, DE,
GA this array is called Particle Position,
Stochastic Population, and Chromosome,
respectively. In cuckoo optimization algorithm, this
array is also called Habitat.
In order to solve a N- dimensional optimization
problem, we need to form an array of 1N as
follows:
)
= (1
, 2
, ,
(23)
Where d is dimension of the objective function,

i is ith habitat, and t is t th generation. Therefore, a
candidate habitat matrix of size NP Nd (NP is the
number of habitats) is randomly generated.
In nature, each cuckoo lays 5 to 20 eggs, so the
number of eggs is randomly determined from 5 to 20
for each cuckoo at each generation of the
optimization process. Each cuckoo lays in the special
Where K is an integer number, xu , xl are upper

and lower of variable limits, respectively.
Cuckoos lay their eggs in other host birds nests
according to their ELR. Some of the eggs in host
birds nests will be recognized and thrown out by
host bird, so p% of all eggs will be killed. When the
cuckoos eggs hatch, they throw the host birds eggs
out from nests. Cuckoos chicks grow in the host
birds nests. When they become mature, they
immigrate to new area with more food and more
similarity of eggs to host birds eggs. To immigrate,
the best value experienced will be determined as a
new area that other cuckoos immigrate to there.
When they want to immigrate to a new area, they do
not fly all way to the destination habitat. They only
fly a part of whole way with a deviation like Figure
3.
New area
th
Output power (W)
2.5

( ) (24)

ng
x 10
Cuckoo's location
le
distance from its habitat called egg laying radius

(ELR) as follows:
le
Fig.2 shows the output power captured by wind

turbine versus rotational speed. According to this
figure, the output power function has a one global
optimum point, and it is a non-linear function of the
rotational speed.
ISSN: 2251-9246
EISSN: 2345-6221
for next generation
W
ho
197
Ap
art
of
le
ho
len
gth
Current
Cuckoo's location
Fig.3. Immigration of a cuckoo to a new location for next
generation
4. Fuzzy logic controller

The five main steps of fuzzy controller are input
and output variables, fuzzy rule, fuzzification,
inference, and defuzzification. The maximum power
point Pm,optimal and its respective rotational speed
optimal are calculated by COA. These values are
compared with the actual values of the output power
Pm,actual and of the rotational speed actual ,
respectively. As seen in Figure 4, we have two input
variables for fuzzy logic controller; the first input is
the difference between the maximum power point
and the actual power point as Equation (8). The
second input is the difference between the optimal
rotational speed that belongs to the maximum power
point and the actual rotational speed as Equation (9).
198
Table.2.
Parameter values of the simulated system
Air density ()
1.2929 (/3 )
Pm,actual
Fuzzy Rule
Optimization
Gain
Pm,optimal
Cuckoo
Radius of blade ()
7.4 (m)
Pitch angle ()
Fuzzy Interface
Fuzzifier
optimal
ISSN: 2251-9246
EISSN: 2345-6221
Defuzzifier
Gain
Parameter values of PMDG
Gain
Pm,actual
actual
Stator phase resistance
0.98 ()
Stator phase resistance
2.83 (mH)
Fig.4. Configuration of the proposed MPP tracker
Pole pairs
Inertia
30 (kg.m2)
= , ,
(25)
Parameter values of BOSST converter
= optimal actual
(26)
Capacitance
Inductance
400e-6 (H)
800e-6 (F)
Load
A mamdani inference is used as fuzzy inference

system. Fuzzy controller based on these inputs and
fuzzy rules change the duty cycle of boost converter.
The input and output membership functions are
shown in Figures 5 and 6. Fuzzy rules are shown in
Table 1.
30 ()
Load
The simulations are performed in MATLAB

environment. The schematic diagram of the proposed
MPP tracker is completely shown in Figure 7.
L
C dc
wind speed
Cuckoo
Optimization
Pm,optimal
Gain
optimal
Gain
Co
PMSG
Gain
Pm,actual
Fig.5. Input and output membership function of Pm and ,

respectively
Fuzzy logic
Gain
PID
PWM
Pm,actual
actual
Fig.7. Completely schematic diagram of the proposed MPP tracker

Fig.6. Input membership function of
A three- phase diode bridge rectifies the voltage

generated by PMSG, and the dc- link capacitor Cdc is
used to smooth the voltage ripple of the dc voltage
generated by the three- phase diode bridge rectifier.
Finally, the boost converter is used to control wind
turbine in the maximum power point by adjusting the
duty cycle.
To confirm the ability of the proposed MPP
tracker, in the first case, wind speed pattern will be
changed slowly as Fig.
Table.1.
Rules table of fuzzy logic applied to the WECS
++
++
++
++
++
5. Simulation results
In order to validate the performance of the
proposed approach, the proposed Fuzzy- Cuckoo
tracker is applied to the wind turbine with following
parameter values:
13
Wind speed (m/s)
12
11
10
9
8
7
Time (sec)
Fig.8. Assumed wind speed profile
10
199
Fig.9- a shows the output power of wind turbine

tracked by the proposed Fuzzy- Cuckoo controller,
and Figure 9- b shows the optimal rotational speed
calculated by cuckoo optimization according to the
wind speed pattern (as Fig.8).
As it is clearly observed, the proposed fuzzycuckoo outperforms PSO and fuzzy MPP trackers.
Figure 11 shows that fluctuations of the proposed
approach are also less than PSO and fuzzy, as well as
higher mean value of the output power.
One of the difficulties in MPPT trackers is the
fast variations of the wind speed. In this case, for
further demonstration, the fast variations of the wind
speed are applied to the MPP trackers. Figure 12
shows the fast wind variations, and Figure 13 shows
the power tracked by fuzzycuckoo, PSO, and fuzzy
trackers.
Output power (W)
x 10
ISSN: 2251-9246
EISSN: 2345-6221
10
14
Time (sec)
14
13
8
6
4
2
10
Time (sec)
Fig.9. Output maximum power captured by the proposed MPP
tracker and rotational speed calculated by cuckoo optimization
algorithm while the wind speed is slowly changed
As seen in Fig.9, the proposed fuzzycuckoo

tracker is capable of tracking the maximum power
point in normal wind changes. To prove the
superiority of the proposed fuzzycuckoo approach
than conventional MPP trackers such as PSO and
fuzzy trackers in normal wind changes, a comparison
with these trackers is provided in following figures.
Figure 10 shows the new assumed pattern of wind
speed.
Wind speed (m/s)
12
10
9
8
7
6
10
Time (sec)
Fig.12. Fast wind variations profile.
4
12
x 10
x 10
x 10
7
Fuzzy
PSO
COA- Fuzzy
10
10
8
6.5
5
4.2
x 10
6.6
6.7
4.8
4.25
4.6
9
x 10
9.2
6.5
6
2
0
2.2
2.4
2.6
10
Time (s)
Fig.13. Simulation results in fast variations of the wind speed
As it is shown in Fig.13, the proposed fuzzycuckoo controller outperforms the conventional MPP
trackers such as PSO and fuzzy trackers.
11
10
9
6. Conclusion
8
7
10
Time (sec)
Fig.10. Assumed wind pattern for comparison purpose
Fig.11 shows the maximum output power

tracked by fuzzycuckoo, PSO, and fuzzy trackers.
4
x 10
COA- Fuzzy
Fuzzy
PSO
Output power (W)
Wind speed (m/s)
11
10
Output power (W)
Rotational speed (rad/s)
12
12
x 10
6
7.1
7
6.9
4.9
4
x 10
x 10
4.6
4.55
4.5
3.9
3.85
4.45
2.9
3.8
0.9
1
0
5.1
x 10
6.9
3
5.35
5.3
5.25
5.2
x 10
7
3.9
3.8
8.9
1.1
7.1
3.85
3.1
9.1
Time (sec)
Fig.11. Maximum output power captured by the fuzzycuckoo,
PSO, and fuzzy MPP trackers
10
In this paper, an accurate MPP tracker called

Fuzzy-cuckoo tracker has been applied to the smallscale WECS with a PMSG and a three-phase diode
bridge rectifier. The proposed MPP tracker uses
Cuckoo optimization algorithm to detect the
maximum power point of the mechanical power
curve and its respective rotational speed. The wind
speed is measured by anemometer as an input of the
Cuckoo optimization algorithm. The difference of the
optimum output power that is calculated by the
Cuckoo optimization algorithm and actual output
power has been fed the Fuzzy as the first input, and
the difference of the rotational speed that belongs to
the optimum output power and actual rotational
speed has also been fed the Fuzzy as the second
input. Finally, the output of Fuzzy logic has been
used as a set- point. The implementation of the
200
proposed MPP tracker is very simple because only

the measurement of the wind speed and the rotational
speed are required. In comparison with other
conventional MPP trackers such as PSO, and Fuzzy
trackers, the proposed MPP trackers are more
accurate and robustness (Figures 11 and 13).
References
[1] Z. M. Dalal, Z. U. Zahid, W. Yu, and J. S. Lai, Design and
Analysis of an MPPT Technique for Smal- Scale Wind
Energy Conversion System, IEEE Transactions on Energy
Conversion, Vol.28, No.3, pp.756-767, 2013.
[2] A. Urtasun, P. Sanchis, I. S. Martn, J. Lpez and L. Marroyo,
Modeling of samll wind turbines based on PMSG with diode
bridge for sensorless maximum power tracking, Renewable
energy, Vol.55, pp.138-149, 2013.
[3] C. Ming, C. H. Chen and C. H. Tu, Maximum power point
tracking- based control algorithm for PMSG wind generation
system without mechanical sensors, Energy conversion and
management, Vol.69, pp.58-67, 2013.
[4] Y. Xia, K. H. Ahmed and B. W. Williams, A New Maximum
Power Point Tracking Technique for Permanent Magnet
Synchronous Generation Based Wind Energy Conversion
System, IEEE Transactions on Power Electronics, Vol.26,
No.12, pp.3609-3620, 2011.
[5] A. M. Eltamaly and H. M. Farh, Maximum power extraction
from wind energy system based on fuzzy logic control,
Electrical power systems research, Vol.97, pp.144-150, 2013.
[6] M. Narayana, G. A. Putrus, M. Jovanovic, P. S. Leung and S.
McDonald, Generic maximum power point tracking
controller for small-scale wind turbines, Renewable Energy,
Vol.44, pp.72-79, 2012.
[7] C. Y. Lee, P. H. Chen and Y. X. Shen, Maximum power
point tracking (MPPT) system of small wind power generator
using RBFNN approach, Expert Systems with Applications,
Vol.38, pp.12058-12065, 2011.
[8] V. Agarwal, R. K. Aggarwal, P. Patidar and C. Patki, C, A
Novel Scheme for Rapid Tracking of Maximum Power Point
in Wind Energy Generation Systems, IEEE Transactions on
Energy Conversion, Vol.25, No.1, pp.228-236, 2010.
ISSN: 2251-9246
EISSN: 2345-6221
[9] E. Koutroulis, and K. Kalaitzakis, Design of a Maximum

Power Tracking System for Wind- Energy- Conversion
Applications, IEEE Transactions on Industrial Electronics,
Vol.53, No.2, pp.486-494, 2006.
[10] M. E. Haque, M. Negnevitsky and K. M. Muttaqi, A Novel
Control Strategy for a Variable-Speed Wind Turbine With a
Permanent-Magnet
Synchronous
Generator,
IEEE
Transactions on Industry Applications, Vol.46, pp.331-339,
2010.
[11] A. Meharra, M. Tioursi, M. Hatti and S. A. Boudghene, A
variable speed wind generator maximum power tracking based
on adaptative neuro-fuzzy inference system, Expert Systems
with Applications, Vol.38, pp.7659-7664, 2011.
[12] W. M. Lin and C. M. Hong, Intelligent approach to
maximum power point tracking control strategy for variablespeed wind turbine generation system, Energy, Vol.35,
pp.2440-2447, 2010.
[3] C. H. Chen, C. M. Hong and F. S. Cheng, Intelligent speed
sensorless maximum power point tracking control for wind
generation system, Electrical Power and Energy Systems,
Vol.42, pp.399-407, 2012.
[14] I. Kortabarria, J. Andreu, I. M. Alegria and J. Jimenez, A
novel adaptative maximum power point tracking algorithm for
small wind turbines, Renewable Energy, Vol.63, pp.785-796,
2014.
[15] T. Senjyu, Y. Ochi, Y. Kikunaga, M. Tokudome and A. Yona,
Sensor-less maximum power point tracking control for wind
generation system with squirrel cage induction generator,
Renewable Energy, Vol. 34, pp.994-999, 2009.
[16] T. Senjyu, S. Tamaki, E. Muhando, N. Urasaki, H. Kinijo, T.
Funabashi, F. Hideki and H. Sekine, Wind velocity and rotor
position sensorless maximum power point tracking control for
wind generation system, Renewable Energy,Vol.31, pp.17641775, 2006.
[17] R. Rajabioum, Cuckoo Optimization Algorithm, Applied
Soft Computing, Vol.11, pp.5508-5518, 2011.
201
pp.201:208
ISSN: 2251-9246
EISSN: 2345-6221
Solving the Economic Load Dispatch Problem Considering

Units with Different Fuels Using Evolutionary Algorithms
Mostafa Ramzanpour1, Hamdi Abdi2
1
Electric Engineering Department, Science and Research branch, Islamic Azad University Kermanshah, Iran
Ramzanpour.mostafa@gmail.com
2
Electric Engineering Department, Faculty of Engineering, Razi University, Kermanshah, Iran
Abstract
Nowadays, economic load dispatch between generation units with least cost involved is one of the most important issues in
utilizing power systems. In this paper, a new method i.e. Water Cycle Algorithm (WCA) which is similar to other intelligent
algorithm and is based on swarm, is employed in order to solve the economic load dispatch problem between power plants.
In order to investigate the effectiveness of the proposed method in solving non-linear cost functions which is composed of
the constraint for input steam valve and units with different fuels, a system with 10 units is studied for more accordance with
literatures in two modes: one without considering the effect of steam valve and load of 2400, 2500, 2600 and 2700 MW and
the other one with considering the effect of steam valve and load of 2700 MW. The results of the paper comparing to the
results of the other valid papers show that the proposed algorithm can be used to solve in any kind of economic dispatch
problems with proper results.
Keywords: Economic load dispatch, water cycle algorithm, valvepoint effect.
1. Introduction
Most of the optimization problems in power
systems such as economic load dispatch have
complicated and non-linear features with equality
and inequality constraints. This causes it hard to be
mathematically solved [1]. Economic load dispatch is
an important issue in the field of management and
utilization of power system where the aim is to
determine the production value of each power plant
in a way the load of system is supplied with least cost
while all the constraints are fulfilled.
Obtaining the optimal solution is sometimes
difficult in solving the economic load dispatch
problem due to complication of fuel cost function of
power plants and also due to some limitation. One of
these complications is related to the actual form of
cost function. It has to be noted that in practice cost
function of a power plant has no smooth form. In
other words, this cost function has local maximum

and minimum and usually it is not derivable in these
points [2-3].
Different optimization methods and techniques
have been used to solve the problem of economic
load dispatch. Some of these methods are based on
mathematical methods such as linear and quadratic
programming [4-5]. Linear programming methods
are generally quick and use linear and step
approximation of fuel cost which in turn reduces the
accuracy of the problem. To overcome this problem,
non-linear programming methods is used. However,
non-linear programming method has difficulties in
convergence and also with an increase in the number
of units the required time and memory to solve the
problem would considerably increase.
To overcome such problems evolutionary
algorithms are proposed. These algorithms have
202
several distinct advantages such as using probable

search instead of explicit methods and also
effectively finding the general optimal points instead
of local optimal points. For instance, Artificial
Neural Network [6], Genetic Algorithm [7],
Simulated Annealing [8], Ant Colony Optimization
[9] are some of these algorithms. In ref [10], a new
algorithm named Artificial Life is proposed to
minimize a non-convex combinatorial function.
Interactive Honey Bee Mating Optimization [11] and
Artificial Immune Systems [12-13] are used to solve
economic load dispatch problem.
In this paper, a novel evolutionary algorithm
which is based on the water cycle algorithm and is
inspired from water cycle in the nature is proposed to
solve the economic load dispatch (ELD) between
power plants considering actual model of nonsmooth cost functions on a test system with 10 units
with and without the effect of steam valve. The rest
of the paper is organized as follow. Problem
description is presented in section 2. In section 3, a
comprehensive definition of the water cycle
algorithm (WCA) is expressed and the application of
WCA in solving the ELD is presented in section 4.
Simulation and conclusion are presented in section 5
and 6, respectively.
2. Economic Load Dispatch Problem
As mentioned before, ELD is defined as a
process of allocating the levels of generation for each
power plant in combinatorial form in a way the
demand of system is supplied completely and
economically [14].
In order to reach to optimal production for each
power plant, curve of fuel cost has to be modeled as
a mathematical relation. In classic case, this function
is modeled as quadratic function (Figure 1) but in
practical and developed cases; this model is modeled
as non-linear and discontinues form due to several
constraints.
ISSN: 2251-9246
EISSN: 2345-6221
(27)
=1
Where PD is the demand power, Pgi power

generated (output power) by ith generator, ng number
of generators in the system.
In more complex load dispatch problems, the
loss of transmission network (PL) has to be added to
the equation (1).
= +
(28)
=1
Equation (2) expresses that the sum of

generated power is equal to sum of the consuming
power including power consumed in loads and power
wasted in transmission line. This equation is a form
of the law of conservation of energy. The value of P L
is calculated by equation (3).

= + 0 + 00
=1 =1
(29)
=1
Where Bij, B0i and B00 are factors of loss

function for transmission network. Fuel cost of each
power plant is calculated from the following relation.
2
( ) =
+ +
(30)
Where F is fuel cost and ci, bi, ai are factors for

fuel cost function of ith unit.
In Figure 1, Pgimin is the minimum loading
range that below this range it would not be
economical (or technically impossible) for the unit
and Pgimax is output maximum range for unit.
Therefore, output power of generator has to be within
minimum and maximum ranges.
(31)
In steam plants, several steam valves are used

in turbine for controlling the output power of the
generators. Opening the steam valve would lead to a
sudden increase in loss and causes ripples in inputoutput curve and consequently causes cost function
non-smooth. If the effect of steam valve is
considered in power plants, cost function of their
generation would take a non-smooth form due to
related mechanical effects.
Fig.1. The curve of fuel cost for generators in smooth and
continues condition
The first relation in optimal load dispatch

problem is the law of conservation of energy. The
sum of generated power by power plants has to be
equal to the load demanded from grid.
203
ISSN: 2251-9246
EISSN: 2345-6221
In a random matrix with certain dimension, the

values of every variable X1 , X2 , X3 , , XN can be real
or complex. Cost function or cost of the problem is
defined as below:
= (1 , 2 , , ) , = 1,2,3, , (35)
Fig.2. Fuel cost curve for generators with 4 steam valves
This influence is usually modeled by adding a

sinusoidal term in cost function of power plants.
Therefore, equation (4) considering the effect of
input steam valve is expressed as equation (6) [15].
2
( ) =
+ + + | ( (
))|
(32)
Where egi and fgi are factors for the effect of

valve point on ith generator.
When generation units perform with different
kinds of fuels, cost function of each unit is expressed
by some quadratic equations where each one of them
corresponds to one kind of fuel. This is formulated
mathematically as in equation (7) [16-17].
2
1
+ 1 + 1 ,
1
, 1
2
2
+ 2 + 2 , 1 2
, 2
( ) =
:
:
2
{ + + , ,1 ,
(33)
After producing the initial matrix, N SR number

of them is considered as sea and rivers. Besides, a
drop with the best answer is chosen as sea. The other
members as rain drops would flow to the rivers or
directly to the sea. In fact, NSR is the sum of river
(user parameter) and sea.
= +
1
(36)
It would be understood which drop will go to

which river based on the stream intensity of each
river:
= {|
| }
(37)
=1
= 1,2, ,
After flowing the drop to the river, it has to be
known that how rivers are flown down to seas.
Movement of each stream to a river is known by a
line that joins them. This distance is calculated
randomly.
(, C ) , C > 1
(38)
, , express the factors for fuel cost

function of ith unit.
3. Water Cycle Algorithm
In this section, a new algorithm inspired by
water cycle in nature is presented which is not
employed for optimization on any power systems
[18]. Similar to other heuristic swarm algorithms, the
proposed method is started with an initial population
named rain drops (Npop). A matrix is produced as rain
drops with a dimension of Npop*Nvar for initializing
the optimization. Rows and columns of this matrix
are composed of population (Npop) and number of
design variables (Nvar) or generation units,
respectively.
1
2
Population of raindrops = 3
[ ]
(34)
1
11
21
31
=
12
[1
22
32
Fig.3. Schematic of flowing a stream to a river
Where C is a value between 1 and 2 (near to 2).

The current distance between stream and river is
shown by parameter d. X in equation (14) is a
random value between 0 and C*d. The values of C
greater than 1 enables stream to flow to the river in
different directions. Therefore, the best value for C
would be 2. This concept can be used in flowing
rivers to the sea. Hence, new positions for stream and
river can be defined as follow [19].
+1
=
+ (

) (39)
+1
=
+ (

) (40)
Where rand is a random number with uniform

distribution between 0 and 1. After updating the
204
position of each drop and investigating the objective

function, if the proposed solution by stream is better
than a river connected to it, position of stream and
river are changed (i.e. stream would be river and vice
versa). This displacement could be happened for
river and sea.
The evaporation criterion is investigated after
the abovementioned stages. When the position of a
stream or river is completely corresponding with the
position of a sea, it means that it has flown to sea. In
this condition, evaporation is done and new streams
and drops are again flown to the mountains by rain
and the above procedure is repeated.
|

| < , = 1,2,3, , 1 (41)
Where dmax is a small number near to zero and

controls the depths of search close to sea. Greater
values of dmax increase the search space and small
values if depth. dmax is decreased in any repetition.
+1
(42)
If the above condition is set, rain will come and

all the above procedure would be repeated. Equation
(19) is used for specifying the new position of newly
produced streams.
= + ( )
(43)
Where LB and UB are the lower and upper

range of the problem, respectively. Besides, there
may some streams that would flow directly to the sea
without flowing to a river.
= + (1, )
(44)
Where is the search domain near to sea. randn

is a random number of normal distribution. Great
values for increase the possibility of going out of
the possible area. Besides, small values for causes
the algorithm to search in a smaller area near to sea.
The best value for is selected 0.1.
Stop criterion in heuristic algorithm is usually
the best calculated answer in which stop criterion
would be defined as maximum number of iteration,
processing time or a non-negative small value as
tolerance between two previous results. Performance
of WCA is agreeable to maximum iteration as a
convergence criterion.
4. Application of WCA on ELD problem
In this paper, a new algorithm called Water
Cycle is used to optimize the overall fuel cost of
power plants. At first, the initial values of WCA like
Npop, Nsr and data related to generation units such
as factors of fuel cost function of generators, output
limitations of generators and the demanded load are
summoned by the system. After this, population
ISSN: 2251-9246
EISSN: 2345-6221
matrix of drops is produced in a random manner. At

third stage, constraints of the ELD problem are
investigated.
In order to investigate the power balance
condition, the value of is calculated for each
population like according to the following equation:
= ( + )
(45)
=1
If =0, it means that the inequality constraint is

met; otherwise, the calculated is added to a unit
randomly. Inequality constraint is then checked for
that unit. If the power is more than maximum power
of that unit, power is set to the maximum value.
Since can also be negative, so if the power is less
than the minimum power of that unit, then the power
is set to minimum value. Therefore, we have:
= {
>
<
(46)
Now we go back to the second stage and repeat

it until = 0.
For solving the ELD problem when there are
different kinds of fuels, the considered production
range for each unit is divided to some sub range
(usually 3 sub ranges) and if the production power is
put in a sub range, the fuel kind of that range is
introduced as the optimum fuel for that unit. In fact,
this constraint makes a piecewise quadratic cost
function and consequently difficult optimization.
There are a few studies in this field considering the
simultaneous effect of input steam valve.
At fourth stage, cost of one drop and at fifth
stage the intensity of stream for river and sea re
calculated. At sixth stage, flowing the streams to the
rivers and rivers to the seas are investigated. After
updating the position of each drop and investigating
the objective function, if the proposed solution by a
stream is better than a river connected to it, position
of stream and river are changed (seventh stage). This
stage could be happened for river and sea (eighth
stage). Evaporation condition which has an important
role in preventing the algorithm to be trapped in the
local minima is investigated. At next stage, dmax is
reduced. When the distance between river and sea is
less than dmax, it means that river has reached to the
sea. In other words, if the evaporation condition is
met, rain will occur based on equations (19) and (20).
And finally stop criterion is checked. Performance of
WCA is usually agreeable to maximum iteration as a
convergence criterion. If the stop criterion is met,
algorithm will stop; otherwise it goes back to the
third stage.
205
ISSN: 2251-9246
EISSN: 2345-6221
5. Simulation
In this kind of problem some different
consuming fuels are considered for different
production ranges of generators which make a
quadratic cost function.
5.1. System with 10 units without considering the
effect of steam valve
To investigate this kind of ELD problem, a
system with 10 units with loads of 2400, 2500, 2600
and 2700 MW without considering the effect of input
steam valve is taken into account as the first test
system. Parameters of the algorithm in this test is set
as follow
Npop = 100 , Nsr = 30 , dmax = 0,1 , C = 2 , U = 0.1
Figure 4 and Tables 1 and 2 show the

convergence diagram, optimum solutions and
comparison with other methods, respectively.
Table.1
Results of WCA for system with 10 units considering the
effect of input steam valve
Unit No 2400(MW) 2500(MW) 2600(MW) 2700(MW)
Output F Output F Output F Output F
1
189.73 1 206.52 2 216.58 2 218.27 2
2
202.35 1 206.45 1 210.88 1 211.67 1
3
253.89 1 265.74 1 278.51 1 280.72 1
4
233.05 3 235.95 3 239.07 3 239.63 3
5
241.85 1 258.01 1 275.53 1 278.49 1
6
233.04 3 235.95 3 239.18 3 239.63 3
7
253.26 1 268.87 1 285.70 1 288.60 1
8
233.04 3 235.95 3 239.13 3 239.63 3
9
320.38 1 331.49 1 343.52 3 428.47 3
10
239.40 1 255.05 1 271.97 1 274.87 1
Ftotal (S/h) 481.7216
526.2279
574.3791
623.798
Table.2
Comparison of WCA results by different methods without
considering the effect of input steam valve
Method
Fuel cost $/h
2400(MW) 2500(MW) 2600(MW) 2700(MW)
HNUM[20]
488.50
526.70
574.03
625.18
HNN[21]
487.87
526.13
574.26
626.12
AHNN[22]
481.72
526.23
574.37
626.24
ELANN[23]
481.74
526.27
574.41
623.88
IEP[24]
481.779
526.304
574.473
623.851
DE[25]
481.723
526.239
574.381
623.809
MPSO[26]
481.723
526.239
574.381
623.809
RCGA[27]
481.723
526.239
574.396
623.809
AIS[28]
481.723
526.240
574.381
623.809
HICDEDP[29]
481.723
526.239
574.381
623.809
EALHN[30]
481.723
526.239
574.381
623.809
WCA
481.7216
526.2279
574.3791
623.7980
Fig.4. Convergence diagram of WCA for system with 10 units and

loads of 2400 to 2700 MW
As illustrated in the results of system with 10

units without considering the effect of steam valve,
this system has the best cost in loads of 2400 to 2700
MW in comparison to the other methods. Besides,
Figure 4 shows that this system without considering
the effect of steam valve with load of 2400 MW
converges quicker than the same system with load of
2700 MW.
5.2. System with 10 units considering the effect of
input steam valve
There are no too many researches to combine
the effect of input steam valve and the constraint of
different consuming fuels in the ELD problem. For
this purpose, the second test system is considered
with adding the effect of steam valve to previous
system and load of 2700 MW. Parameters of
algorithm for this system are similar to the previous
system. Figure 5 and Table 3 present convergence
diagram and comparison of the results with other
methods, respectively.
Table.3
Comparison of WCA results for system with 10 units considering
the effect of input steam valve and load of 2700 MW
Unit No
1(MW)
2(MW)
3(MW)
4(MW)
5(MW)
6(MW)
7(MW)
8(MW)
9(MW)
10(MW)
Ftotal (S/h)
CGA_MU[31]
Output
222.0108
211.6352
283.9455
237.8052
280.4480
236.0330
292.0499
241.9708
424.2011
269.9005
624.7193
F
2
1
1
3
1
3
1
3
3
1
IGA_MU[31]
Output
219.1261
211.1645
280.6572
238.4770
276.4179
240.4672
287.7399
240.7614
429.3370
275.8518
624.5178
F
2
1
1
3
1
3
1
3
3
1
WCA
Output
220.0550
210.9169
279.6702
238.7458
280.1485
239.6610
287.7073
239.6864
427.4088
275.9998
623.8509
F
2
1
1
3
1
3
1
3
3
1
206
Fig.5. Convergence diagram of WCA for system with 10 units

considering the effect of input steam valve and load of 2700 MW
There are a few researches to simultaneously

study the effect of steam valve and units with
different fuels in system with 10 units with
considering the effect of steam valve due to
complications. Therefore, there was just a system
surveyed once with load of 2700 MW. Fuel cost of
this system is 623.8509 which are improved slightly
comparing to the other two algorithms. Figure 5
illustrates a comparison of system with 10 units and
load of 2700 MW with and without the effect of
steam valve. The effect of steam valve makes
convergence a little difficult and increases the
probability of converging to the local minimums. In
the first mode, convergence was achieved at about 10
iterations but the effect of steam valve delayed it to
25 iterations.
5.3. Sensitivity analysis of algorithm C
As mentioned before, this parameter controls
the way of flowing from streams to rivers and seas
and has a value between 0 and 2. Values higher than
1 causes the streams to be able to flow in two
directions. Table 4 and Figure 6 present the
comparison of results for 4 different values of this
test system.
Table.4
Results of WCA for system with 10 units and load of 2700 MW
and different values of C
Unit No
1(MW)
2(MW)
3(MW)
4(MW)
5(MW)
6(MW)
7(MW)
8(MW)
9(MW)
10(MW)
Ftotal (S/h)
C =0.1
Output
219.5905
207.1839
297.7758
242.3708
311.3986
245.5736
302.0291
244.5238
350.3311
279.1511
627.4213
F
2
1
1
3
1
3
1
3
3
1
C =0.8
Output
215.2694
222.9899
283.4461
241.0252
294.1461
244.8194
277.4023
243.4481
393.6467
283.6891
626.3751
F
2
1
1
3
1
3
1
3
3
1
C=1.4
Output
212.0377
215.1524
279.4089
240.6257
279.5790
240.0692
285.1218
236.9990
424.3429
286.5712
624.3440
F
2
1
1
3
1
3
1
3
3
1
C =2
Output
220.0550
210.9169
279.6702
238.7458
280.1485
239.6610
287.7073
239.6864
427.4088
275.9998
623.8509
F
2
1
1
3
1
3
1
3
3
1
ISSN: 2251-9246
EISSN: 2345-6221
Fig.6. Sensitivity analysis diagram of WCA for system with 10

units and load of 2700 MW and parameter C
5.4. Sensitivity analysis of Parameter dmax

As mentioned in this section, this parameter
prevents algorithm from quick convergence and to be
trapped in local minima. The value of this parameter
is updated after each iteration. Results and
convergence diagram for three different values of
dmax are shown in Figure 7 and Table 5, respectively.
6. Conclusion
In this paper, a new algorithm is employed in order
to solve the economic load dispatch problem. This
algorithm uses fewer factors to reach to the
optimum solution comparing to the other algorithms
which causes this algorithm to solve the problems
quicker while maintaining the accuracy. In order to
show the ability of WCA in solving non-linear
problems, a system with 10 units in two modes is
investigated i.e. with and without considering the
effect of steam valve.
Table.5
Results of WCA for system with 10 units and load of 2700 MW
and different values of dmax
dmax =2
dmax =0.1
dmax =0.01
Unit No
Output
Output
Output
1(MW)
218.5807
220.0550
216.3999
2(MW)
214.1351
210.9169
211.4120
3(MW)
280.6462
279.6702
284.4471
4(MW)
239.2832
238.7458
240.7613
5(MW)
273.7120
280.1485
278.8212
6(MW)
239.1236
239.6610
237.2423
7(MW)
287.7891
287.7073
289.7176
8(MW)
239.9551
239.6864
240.8957
9(MW)
427.5427
427.4088
425.7371
10(MW)
279.1428
275.9998
274.5656
Ftotal (S/h)
623.9863
623.8509
623.9657
In this system when the effect of steam valve is

not considered, as observed from the related Table,
the cost of this system is obtained as 481.7216,
526.2279, 574.3791 and 623.7980 for loads of 2400
207
to 2700 MW. These results are less than all the other
methods and more near to optimum value. Although
this reduction is not very noticeable, it is very
valuable considering the type of problem and other
performed researches up to now in ELD problem
with different kinds of fuels. In this system when the
effect of steam valve is considered, this effect makes
convergence a little difficult and increases the
probability of converging to the local minimums. In
the first mode, convergence was achieved at about 10
iterations but the effect of steam valve delayed it to
25 iterations. At last, Sensitivity analysis of
Parameters C and dmax was carried out on a system
with 10 units and load of 2700 MW. As shown by
the results and section 3, the best value for parameter
C is 2 which give the best results. The results of
parameter dmax shows that higher values of this
algorithm increase discovery range and search span
of the algorithm but too high values of this
parameter worsens the quality of the solution.
Besides, too low values of this parameter would also
lead to solution with little quality but a proper
reduction in this parameter increases the
convergence.
[8] M. Vanitha and K. Thanushkodi, Solving non-convex
[9]
[10]
[11]
[12]
[13]
References
[1] A. A. El-Keib, H. Ma, and J. L. Hart, Environmentally
[2]
[3]
[4]
[5]
[6]
[7]
constrained economic dispatch using the Lagrangian

relaxation method. IEEE Transactions on Power
Systems, Vol.9, No.4, pp.17231729, Nov. 1994.
N. Sinha, B. Purkayastha, PSO embedded
evolutionary programming technique for non-convex
economic load dispatch, Proc. Power Systems
Conference and Exposition, Vol.1, pp.6671, October
2004.
Z.-L. Gaing, Particle swarm optimization to solving
the economic dispatch considering the generator
constraints, IEEE Transactions on Power Systems,
Vol. 18, No.3, pp. 11871195, Aug. 2003.
Adler.R.B and Fischl.R, Security constrained
economic dispatch with participation factors based on
worst case bus load variations, IEEE Trans. on Power
Apparatus and Systems, Vol.96, No.2, pp.347356,
2006.
Bui.R.T.
and
Ghaderpanah.S,
Real
power
rescheduling and security assessment, IEEE Trans. on
Power Apparatus and Systems, Vol.101, No.8,
pp.29062915, 2007.
J. H. Park, Y. S. Kim, I. K. Eom, and K. Y. Lee,
Economic load dispatch for piecewise quadratic cost
function using Hopfield neural network, IEEE
Transactions on Power Systems, Vol.7, No.3, pp.1232
1237, 1993.
C.-L. Chiang, Improved Genetic Algorithm for Power
Economic Dispatch of Units with Valve-Point Effects
and Multiple Fuels, IEEE Transactions on Power
Systems, Vol. 22, No. 4, pp.16921699, Nov. 2005.
ISSN: 2251-9246
EISSN: 2345-6221
[14]
[15]
[16]
[17]
[18]
[19]
economic load dispatch problem by Efficient Hybrid

Simulated Annealing algorithm, 2212 IEEE
International Conference on Advanced Communication
Control and Computing Technologies (ICACCCT),
No.927, pp.362365, Aug. 2012.
N. A. Rahmat and I. M. M. Ieee, Differential
Evolution Ant Colony Optimization (DEACO)
Technique In Solving Economic Load Dispatch
Problem, IEEE Power Engineering and Optimization
Conference, pp.62, 2012.
T. Satoh, H. Kuwabara, M. Kanezashi, and K. Nara,
Artificial Life System and its Application to Mult
iple-Fuel Economic Load Dispatch Problem,
Proceedings of Evolutionary Computation, Vol.2,
pp.14321432, 2002.
B. Vanaja, S. Hemamalini, and S. P. Simon, Artificial
Immune based Economic Load Dispatch with valvepoint effect, TENCON 2227 - 2227 IEEE Region 12
Conference, pp.15, Nov. 2007.
R. Gonc, C. Almeida, and M. Goldbarg, Improved
Cultural Immune Systems to Solve the Economic Load
Dispatch Problems, IEEE Congress on Evolutionary
Computation, pp.621627, 2013.
A. Ghasemi, A fuzzified multi objective Interactive
Honey
Bee
Mating
Optimization
for
Environmental/Economic Power Dispatch with valve
point effect, International Journal of Electrical Power
& Energy Systems, Vol.49, pp.327321, Jul. 2013.
N. Amjady and H. Sharifzadeh, Solution of nonconvex economic dispatch problem considering valve
loading effect by a new Modified Differential
Evolution algorithm, International Journal of
Electrical Power & Energy Systems, Vol.32, No.7,
pp.793923, Oct. 2012.
T. Niknam, M. R. Narimani, and R. AzizipanahAbarghooee, A new hybrid algorithm for optimal
power flow considering prohibited zones and valve
point effect, Energy Conversion and Management,
Vol.57, pp.192226, Jun. 2012.
K. Vaisakh and a. S. Reddy, MSFLA/GHS/SFLAGHS/SDE algorithms for economic dispatch problem
considering multiple fuels and valve point loadings,
Applied Soft Computing, pp.111, Jul. 2013.
a. K. Barisal, Dynamic search space squeezing
strategy based intelligent algorithm solutions to
economic dispatch with multiple fuels, International
Journal of Electrical Power & Energy Systems, Vol.45,
No.1, pp.5259, Feb. 2013.
H. Eskandar, A. Sadollah, A. Bahreininejad, and M.
Hamdi, Water cycle algorithm A novel
metaheuristic optimization method for solving
constrained engineering optimization problems,
Computers & Structures, Vol.112111, pp.151166,
Nov. 2012.
H. Eskandar, A. Sadollah, A. Bahreininejad, and K.
Lumpur, Weight Optimization of Truss Structures
Using Water Cycle Algorithm, International Journal
of Optimization in Civil Engineering, Vol.3, No.1,
pp.115129, 2013.
208
ISSN: 2251-9246
EISSN: 2345-6221
[20] C. E. Lin and G. L. Viviani, Hierarchical economic
[27] S. Baskar, P. Subbaraj, and M. V. C. Rao, Hybrid real
dispatch for piecewise quadratic cost functions, IEEE

Transactions on Power Apparatus and Systems, vol.
PAS-123, No.62, pp.11221125, 1974.
J. H. Park, Y. S. Kim, I. K. Eom, and K. Y. Lee,
Economic load dispatch for piecewise quadratic cost
function using Hopfield neural network, IEEE
Transactions on Power Systems, Vol.7, No.3, pp.1232
1237, 1993.
K. Y. Lee, A. Sode-Yome, and J. H. Park, Adaptive
Hopfield neural networks for economic load dispatch,
IEEE Transactions on Power Systems, Vol.13, No.2,
pp.519526, 1997.
S. C. Lee and Y. H. Kim, An enhanced Lagrangian
neural network for the ELD problems with piecewise
quadratic cost functions and nonlinear constraints,
Electric Power Systems Research, Vol.62, No.3,
pp.162122, Jan. 2002.
Y. M. Park, J. R. Won, and J. B. Park, A new
approach to economic load dispatch based on improved
evolutionary programming, Eng. Intell. Syst. Elect.
Eng. Commun, Vol.6, No.2, pp.123112, 1997.
N. Noman and H. Iba, Differential evolution for
economic load dispatch problems, Electric Power
Systems Research, Vol.27, No.7, pp.13221331, Aug.
2007.
J.-B. Park, K.-S. Lee, J.-R. Shin, and K. Y. Lee, A
Particle Swarm Optimization for Economic Dispatch
with Non-smooth Cost Functions, IEEE Transactions
on Power Systems, Vol.22, No.1, pp.3442, Feb. 2005.
coded genetic algorithm solution to economic dispatch

problem, Computers & Electrical Engineering,
Vol.29, No.3, pp. 422419, May 2003.
B. K. Panigrahi, S. R. Yadav, S. Agrawal, and M. K.
Tiwari, A clonal algorithm to solve economic load
dispatch, Electric Power Systems Research, Vol.22,
No. 12, pp.13711379, Aug. 2002.
R. Balamurugan and S. Subramanian, Hybrid integer
coded differential evolution dynamic programming
approach for economic load dispatch with multiple fuel
options, Energy Conversion and Management, Vol.49,
No.4, pp.627614, Apr. 2007.
[21]
[22]
[23]
[24]
[25]
[26]
[28]
[29]
[30]
[31]

pp.209:216
209
ISSN: 2251-9246
EISSN: 2345-6221
Implementation of ROCOPS Method for Anti-Islanding

Protection of Wind Turbine
Behrooz Sobhani1, Mehdi Nooshyar2
1,2
Technical Engineering Department, University of Mohaghegh Ardabili, Ardabil, Iran. b.sobhany@gmail.com
Abstract
In recent years, based on the growing importance of clean energy in comparison with conventional energy production from
fossil fuels, DG systems are gradually becoming more popular all over the world. These resources solve many problem of
system. However, these resources create some other problems too. One of the most problem of DGs is unwanted islanding.
This paper addresses reliable passive islanding detection algorithm based on the change of positive sequence (ROCOPS) of
voltage signal analysis method. At first, all possible linear and nonlinear load switching, motor starting and capacitor bank
switching is simulated. The ROCOPS signal of these conditions is measured for all conditions. From of these data, the
reliable value of ROCOPS based on thresholding is calculated. The studies reported in this paper are based on time-domain
simulations using MATLAB, and the feasibility of the proposed method is evaluated with an experimental system. The
experimental system is a test system that impalement for islanding condition detection. The results show that the proposed
islanding detection method succeeds in detecting islanding both in the experimental and simulated systems with negligible
Non Detection Zone (NDZ).
Keywords: Islanding Detection, ROCOPS, Wind Turbines, Wind Turbine Simulator, Thresholding, Non-Detection Zone.
1. Introduction
Rapid technological developments in generation
and storage of energy, the growing concern on
environmental issues, increasing prices of oil and
natural gas and global warming led to great
affirmation on alternative energy sources such as
solar and wind [1]. Increasing of clean energy is
caused DG systems are gradually becoming more
popular all over the world. Further, the DG systems
increase energy efficiency and improve power
quality of the distribution network [2]. However,
installing DG in power systems may create some
problems [3-4]. One of these problems is unwanted
islanding. The islanding condition occurs when a
portion of the utility system that contains both load
and DG remains energized while it is isolated from
the remainder of the utility system [3]. IEEE standard
recommends disconnecting all DGs within 2 s after
the formation of unwanted island [5].
Islanding detection techniques can be classified

in three major groups include passive, active and
remote methods [6]. Remote detection techniques are
based on communication between the utility and the
DGs. These schemes include power line signaling [7]
and transfer trip [8]. Although these techniques may
have better reliability than active and passive
techniques, they are expensive to implement and
hence uneconomical.
Active techniques are based on applying a
disturbance into distribution systems and forcing the
islanded system to become unstable. Some active
islanding detection approaches include active and
reactive power variation [9, 10], sandia frequency
shift and sandia voltage shift [11, 12], impedance
measurement, frequency shift and active frequency
drift [13], current injection [14], reducing the
injected current periodically and monitoring the
voltage [5], high-frequency signal injection [15],
virtual capacitor [16] and virtual inductor [17],
210
phase-PLL perturbation [18], slide-mode frequency

shift [19], active frequency drift or frequency bias
[20, 21]. The most of these techniques are employed
to inverter-based distributed generations and
advantage of these methods is their relatively small
Non Detection Zone (NDZ). However, the main
challenges of these methods are power quality
problems and designing the complicated control
circuit.
In Passive techniques certain system parameters
such as voltage, frequency and etc are measured and
the islanding is detected through data processing.
One of the simplest passive methods in islanding
detection is over/under voltage and frequency. In
match condition of load and generation, the change
in system parameters might be very small and within
the thresholds, thus leading to an undetected
islanding situation [3]. Some other passive
techniques are total harmonic distortions [22], rate of
change of frequency [13], phase displacement
monitoring, vector surge [23], rate of change of
generator power output [13], rate of change of phase
angle derivation [24] and Wavelet energy [25] and STransform methods [26]. The main challenges of
passive techniques are large NDZ and threshold
selection. If the threshold for permissible disturbance
in these quantities is set to a low value, then nuisance
tripping becomes an issue, and if the threshold is set
too high, islanding may not be detected.
This paper presents a novel algorithm based on
ROCOPS of voltage signal for islanding detection of
wind turbine distributed generations. In order to
verify the proposed strategy an implemented wind
turbine simulator system is test using proposed
strategy. This method could be reduced the NDZ as it
possible. Also, in order to overcome to threshold
challenge, the thresholding method is used for
detection process. At first, all of ROCOPS signal of
positive sequence voltage is measured for any
switching and islanding condition of study system.
From of these values, threshold value is calculated
based on thresholding method. After that, for other
islanding and non-islanding conditions the proposed
strategy system is tested. Reduction of NDZ,
Avoidance of threshold selection, dominance of
power quality problem using signal processing and
test method on implemented system are mainly
ability of proposed strategy.
The proposed detection algorithm is applied on
wind turbine system which is simulated in
MATLAB/Sim Power System. Also, the feasibility
of the proposed method is evaluated with an
experimental system. The experimental system is a
simulator wind turbine system. The obtained
detection results for any islanding and switching
conditions, shows the efficiency of the proposed
approach.
ISSN: 2251-9246
EISSN: 2345-6221
2. Systems Description
Figure 1 shows a schematic diagram of a
simulated wind turbine unit. The DG unit is
represented by a wind turbine and induction
generator, a capacitor bank used in order to power
factor correction of DG unit. The local load is
represented by a three-phase parallel RL before of
circuit breaker (CB). A parallel RL is conventionally
adopted as the local load for evaluation of islanding
detection methods when the load inductance is tuned
to the system frequency. This system, as represented
in Fig.1, equally connected to point of common
coupling (PCC) with step-up transformer. In order to
obtain the experimental results, a wind turbine
simulator as Fig 2 was implemented. Fig 3 shows the
implemented simulator system. Parameters of the
implementation system are given in Table 1. and
motors saturation curves shown in Fig 4. In gridconnected condition the switches SW1 and SW2 are
closed. Island condition is occurred when SW2 is
open.
Fig.1. Single line diagram of study system
Fig 2.single line diagram of implementation system in order to

islanding condition detection
211
ISSN: 2251-9246
EISSN: 2345-6221
assumed that the DG unit and the local load are

balanced three-phase subsystems within the island. In
balanced three-phase subsystems within the island,
the state-space model of switching based DG system
in the -frame, dynamic model is:
Place of Fig.5
The state space model of the study systems in
islanding mode described as standard form:
.
X (t ) AX (t ) BU (t )
Y (t ) CX (t ) DU (t )
(1)
Which, the ABCD matrix of state-space model

can be defined as follow:
Fig 3: Implementation system in order to islanding condition
detection
Rt
L
t
0
A
1
C
0
-
Rl
L
- 20
1
BT
Lt
C 0
Rl
L
1
C
0 1
Voltage and Frequency of DG should be

remaining in admissible values in both gridconnected and islanded modes. In grid-connected
mode, the voltage magnitude and frequency of the
local load, at PCC, are regulated by the grid.
X T itd
itq
iLd
1
Lt
(3)
D 0
Fig 4.Motor and Generator saturation curves.
Rl C0 0
(
)
L
R (2)
1
( -02 C)
L
RC
vd
(4)
(5)
(6)
With considering u(t ) vtd , figure 6 shows the

step response of system in the islanding mode. The
response time constant of the islanding system is
selected as the analyzing time of wavelet analysis
window.
Table.1
Parameters of implemented system
Parameters
Induction Motors
Local Load
Capacitor
Value
Sn
Vn
f
PF
Rs, Rr
Lr,Ls
Lm
R
L
C
2 kVA
400 V
50 Hz
0.78 Lag
2.3541
0.01678 H
0.275 H
180
Inf
36.75 F
Fig 5: Simple model of case study
3. Mathematical model
This section gives a brief overview of state
space model, step time response of study systems.
By considering balanced condition for islanded
system (Fig. 1), the circuit model of self excited
induction generator with fixed wind speed simple
model model are shown in fig 1. From of this figure,
it is obvious that the model of two systems can be
described as same. Therefore in this section the
mathematical model of switching based DG is
described and generalized for other systems. It is
Fig. 6: Step response of wind turbine system
From of Fig. 6, the transient response time of

switching based DG is about 0.04 seconds, and wind
turbine about 0.2 second. We consider 0.2 seconds of
histogram of ROCOPS data for wind turbine to
achieve reliable detection from the step response
process.
212
4. Proposed Technique
ROCPS
Here a new algorithm for detection of islanding

is based on Rate of Change of positive sequence
(ROCOPS) of load voltage and thresholding method
is proposed.
4.1. Thresholding Method
We introduce a measure for calculating the
difference between two ROCOPS signals magnitude
for islanding and non-islanding conditions and call it
IROCOPS and NIROCOPS, respectively.
For each possible of the islanding cases the
ROCOPS values are calculated and reported.
Similarly for each possible of the non-islanding cases
the ROCOPS values are calculated and reported.
The average value of ROCOPS is greater than
the average value of all ROCOPS. This difference
between islanding and switching values is motivated
us to introduce an approach to classify an unknown
condition, based on the calculation of ROCOPS (for
known islanding or non-islanding cases), into
islanding cases or non-islanding cases.
Suppose that we have NI known islanding cases
and NN known non-islanding cases, these cases are
the training cases of the proposed method, and then
we can calculate (NINN) possible ROCOPS values.
All of the ROCOPS values are sorted and the
magnitude of them is considered. It is expected that
the magnitude of ROCOPS will be separated in
islanding and non-islanding conditions. Thus, a
proper threshold value, we call it k*, separate the
ROCOPS magnitudes into two groups.
The value of k* can be found by an appropriate
thresholding method such as Otsu method [?]. For an
unknown case if its ROCOPS is accessible, by
Comparing of the value of k* with threshold we can
determine the condition of the unknown case.
4.2. ROCOPS Method
Under balanced conditions, each three-phase
variable xabc (t) can be transferred to a stationary
reference frame system by applying the following
transformation [14]:
x p xa e j 0 xbe
2
3
xce
2
3
(7)
x p is positive sequence of x. Therefore, the
vabc in stationary reference frame is:

v p vae j 0 vbe
2
3
vce
2
3
(8)
We propose a new algorithm that employs

ROCOPS to detect the islanding event. Figure 7
shows algorithm of the proposed method. In the first
step, the voltage of load voltage is measured and
ROCOPS is calculated afterwards. The ROCOPS is
calculated as given in equations (9).
ISSN: 2251-9246
EISSN: 2345-6221
v p
t
(9)
In next step, ROCOPS have been compared

with their threshold values. If the value of any of this
parameter exceeds the threshold value, islanding has
been happened when the ROCOPS is higher from
threshold value.
In this study system in first, voltage signal of
the load is measured. And then this voltage signal
passed from the low-pass filter in order to expurgate
the high frequency harmonics of voltage signal.
Then, with ROCOPS given of this signal and
comparison with threshold value, it is detected that
the islanding is occurred or not. Fig 7 shows the
proposed algorithm of this method.
Fig 7. Proposed algorithm in order to islanding detection
5. Experimental and Simulation Results

In this study, the simulation is carried out in
four scenarios to illustrate the effectiveness of the
proposed method.
5.1 Islanding Scenario test (Case 1)
In this case, the load as shown in Fig. 1 is set to
the values given in Table 1. The DG is connected to
the grid and works in grid-connected mode. At t=2 s,
the CB is opened, and the system enters islanding
mode. Fig. 8 shows the dynamic response of the
system prior, during and subsequent to the islanding
event. Fig. 8(a) shows the frequency changing of the
load and demonstrates that it has no main change
prior to the islanding and after the islanding
condition. Fig. 8(b) shows the instantaneous voltage
of phase-a at the PCC and Fig 8(c) depict the
ROCOPS of load voltage. According to these figures,
at t=2.03 s, the ROCOPS increases from threshold
value. Therefore, the proposed method detects the
islanding. In Fig. 9, the experimental results for the
nominal load are depicted. In the experimental case
islanding occurred at t=3.8s. Fig. 9(a) shows the
frequency changing of the load and demonstrates that
is almost fixed prior to the islanding. Fig.9 (b) shows
the instantaneous voltage of phase-a at the PCC.
According to these figure the voltage of utility is
fixed before islanding, and after islanding. Fig. 9(c)
shows the ROCODPS. As shown, the ROCOPS is
exceed from the threshold value at t = 3.93s, which
leads to islanding detection after 0.13 s.
213
ISSN: 2251-9246
EISSN: 2345-6221
simulated system. According to Fig. 10(b), the value

of the voltage is decreased after islanding, and the
frequency of the system, as shown in Fig. 10(a), is
increased.
The results for the experimental system are
depicted in Fig. 11. Fig. 11(a) shows the frequency
changing of the load. Fig. 11(c) shows the ROCOPS
value. In this case, at t = 2.7s, islanding occurs and is
detected at t = 2.81 s. Fig. 11(b) shows the
instantaneous voltage of phase-a at the PCC. The
voltage of utility is decreased after islanding.
Fig 8: dynamic response of simulation system, a)frequency

deviation of PCC voltage b) instantaneous voltage c) rate of
change of positive sequence of voltage
Fig 10: dynamic response of simulation system, a) frequency

Fig 9: dynamic response of experimental system, a)frequency

5.2 Islanding Scenario test (Case 2)

The system shown in Fig. 1 operates in a gridconnected mode. The load absorbs 200 W of real
power from the grid and sends 140 var reactive
power to the grid and 600 W real powers from the
DG. The load parameters are R = 180.5 , L = 3H
and C = 40 mF. As shown, the ROCOPS is exceed
from the threshold value at t = 2.02s, which leads to
islanding detection after 0.02 s.. The islanding
detection time is shorter than in the previous case
study. The voltage magnitude at the PCC and the
islanded system frequency change rapidly, but these
values do not deviate from their acceptable limits.
Fig. 10 shows the results of this condition in the
Fig 11: dynamic response of experimental system, a) frequency

5.3 Motor Starting Condition

The starting of large induction motors may
cause a malfunction of the islanding detection
algorithm. To study the reliability of the proposed
214
ISSN: 2251-9246
EISSN: 2345-6221
algorithm, a 1.5 kW induction motor is connected to

the PCC via a switch in the non-islanding case. The
simulation results of the induction motor starting are
shown in Fig. 12, and the experimental results of this
condition are shown in Fig. 13.
At t = 1.5 s, the induction motor was started.
Then the RMS voltage of the PCC decreased the
values of frequency and voltage nearly fixed, as
shown in Fig. 12(a) and Fig. 12(b) respectively, Fig.
12(c) shows the ROCOPS value does change but the
value does not exceeds from threshold value at this
time. Therefore, the proposed method does not send
a trip and works in a reliable mode.
In the experimental mode, the switching of the
motor occurred at t = 1.2 s, and the voltage of the
PCC decreased at this time as shown in Fig. 13 (b).
The voltage of grid in this study is 375 V and, at the
instant of switching, decreased to 360 V. According
to Fig. 13 (c) it is obvious that the experimental
results and simulation results prove the reliability of
the proposed method.
5.4 Capacitor Bank Switching Condition

Large capacitor bank switching in distribution

power systems initiates disturbances. These
disturbances are propagated in the distribution
system and have some effects on the proposed
method. To test the proposed algorithm, a large 2
kvar capacitor bank was switched at the PCC in the
non-islanding case. This switching occurred at t = 1.5
s. The results for simulation system are shown in Fig.
14. At the switching time, the voltage and frequency
are almost constant. Fig. 14 (c) shows that the
ROCOPS does not change in this condition. The
results show that ROCOPS does not have any
sensitivity to the switching condition, and the
proposed method works perfectly.
The experimental results are described in Fig.
15. A 2kvar capacitor bank was switched at t = 2 s.
As shown in Fig 15(c), ROCOPS does change but
the value does not exceed from threshold value at
this time. Therefore, the results confirm the
simulation results.
215
ISSN: 2251-9246
EISSN: 2345-6221
technique for islanding detection of distributed

generation is proposed based on ROCOPS method.
The main emphasis of the proposed scheme is to
reduce the NDZ and this technique using
thresholding method can also overcome the problem
of setting the detection thresholds inherent in the
existing techniques. By case studies with numerical
simulations and implemented system, the proposed
approach was verified with feasibility, flexibility and
robustness.
References
[1] Jae-Hyung Kim, Jun-Gu Kim, Young-Hyok Ji, Yong-Chae
Jung, and Chung-Yuen Won, , An Islanding Detection Method
for a Grid-Connected System Based on the Goertzel
Algorithm, IEEE Trans. on Power Electronics, Vol.26, No.4,
pp.1049-1055, APRIL 2011.
[2] Soma Biswas, Swapan Kumar Goswami, Amitava Chatterjee.
Optimum Distributed Generation Placement with Voltage
Sag Effect Minimization, Energy Conversion and
Management, Vol.53, Issue.1, pp.163174, January 2012.

[3] Farid Hashemi, Noradin Ghadimi, Behrooz Sobhani,

Islanding Detection for Inverter-based DG Coupled with
Using an Adaptive Neuro-fuzzy Inference System, Electrical
Power and Energy Systems, Vol.45, pp.443455, 2013.
[4] M.R. Banaei, E. Salary, Application of Multi-stage
Converter in Distributed Generation Systems, Energy
Conversion and Management, Vol.62, pp.7683, October
2012.
[5] B. Indu Rani, M. Srikanth, G. Saravana Ilango, C. Nagamani,
An Active Islanding Detection Technique for Current
Controlled Inverter, Renewable Energy, Vol.51, pp.189-196,
2013.
[6] Behrooz Bahrani, Houshang Karimi, Reza Iravani,
Nondetection Zone Assessment of an Active Islanding
Detection Method and its Experimental Evaluation, IEEE
Trans. on Power Delivery, Vol.26, No.2, pp.517-525, April
2011.
[7] Mak ST, A new method of generating TWACS type
outbound signals for communication on power distribution
networks, IEEE Trans. Power App. Syst., PAS-103, No.8,
pp.213440, 1984.
[8] Xu W, Zhang G, Li C, Wang W, Wang G, Kliber J, A power
line signaling based technique for anti-islanding protection of
distributed generatorsPart I: scheme and analysis, [a
companion paper submitted for review].
[9] Hua Geng, Dewei (David) Xu, BinWu, Geng Yang, Active
Islanding Detection for Inverter-Based Distributed Generation
Systems With Power Control Interface, IEEE Trans. on
Energy Conversion, Vol.26, No.4, pp.1063-1072, Dec. 2011.

Conclusion
Following the increased number and enlarged
size of distributed generating units installed in a
modern power system, the protection against
islanding has become extremely challenging
nowadays. Islanding detection is also important as
islanding operation of distributed system is seen a
viable option in the future to improve the reliability
and quality of the supply. In this paper, A new
[10] Jun Zhang, Dehong Xu, Guoqiao Shen, Ye Zhu, Ning He, Jie
Ma, An Improved Islanding Detection Method for a GridConnected Inverter With Intermittent Bilateral Reactive
Power Variation, IEEE Trans. on Power Electronics, Vol.28,
No.1, pp.268-278, Jan. 2013.
[11] H. H. Zeineldin, and M. M. A. Salama, Impact of Load
Frequency Dependence on the NDZ and Performance of the
SFS Islanding Detection Method, IEEE Trans. on Industrial
Electronics, Vol.58, No.1, pp.139-146, Jan. 2011.
[12] John V, Ye Z, Kolwalkar A., Investigation of Anti Islanding
Protection of Power Converter Based Distributed Generators
Using Frequency Domain Analysis, IEEE Trans. on Power
Electronic,Vol.19, Iss.5, pp.1177-1183, 2004.
[13] J.A. Laghari, H. Mokhlis, A.H.A. Bakar, M. Karimi, A New
Islanding Detection Technique for Multiple Mini Hydro Based
216
on Rate of Change of Reactive Power and Load Connecting

Strategy, Energy Conversion and Management, Vol.76,
pp.215224, Dec. 2013.
[14] Hernndez-Gonzlez G, Iravani R., Current Injection for
Active Islanding Detection of Electronically-Interfaced
Distributed Resources, IEEE Trans. On Power Deliv., Vol.21,
Iss.3, pp.1698e705, 2006.
[15] David Daz Reigosa, Fernando Briz, Cristian Blanco Charro,
Pablo Garca, Juan Manuel Guerrero, Active Islanding
Detection Using High-Frequency Signal Injection, IEEE
Trans. on Industry App., Vol.48, No.5, pp.1588 -1597,
Sep./Oct., 2012.
[16] Wen-Jung Chiang, Hurng-Liahng Jou b, Jinn-Chang Wu,,
Active Islanding Detection Method for Inverter-based
Distribution Generation Power System, Electrical Power and
Energy Systems, Vol.42, pp.158166, 2012.
[17] H.-L. Jou, W.-J. Chiang and J.-C. Wu, Virtual Inductorbased Islanding Detection Method for Grid-connected Power
Inverter of Distributed Power Generation System, IET
Renew. Power Gener., Vol.1, No.3, pp.175-181, Sep. 2007.
[18] David Velasco, Cesar Trujillo, Gabriel Garcera, Emilio
Figueres, An Active Anti-Islanding Method Based on PhasePLL Perturbation, IEEE Trans. on Power Electronics,
Vol.26, No.4, pp.1056-1066, Apr. 2011.
[19] K. Jae-Hyung, K. Jun-Gu, J. Young-Hyok, J. Yong-Chae, and
W. Chung- Yuen, An Islanding Detection Method for a Gridconnected System Based on the Goertzel Algorithm, IEEE
Trans. Power Electron., Vol.26, No.4, pp.10491055, Apr.
2011.
ISSN: 2251-9246
EISSN: 2345-6221
[20] H. Guo-Kiang, C. Chih-Chang, and C. Chern-Lin, Automatic
Phase-shift Method for Islanding Detection of Grid-connected
Photovoltaic Inverters, IEEE Trans. Energy Convers.,
Vol.18, No.1, pp.169173, Mar. 2003.
[21] M. Roop, Design issues for grid-connected photovoltaic
system, Ph.D. dissertation, Sch. of Electr. and Comput. Eng.,
Georgia Inst. of Technol., Atlanta, GA, 1998.
[22] Kazemi Karegar H, Shataee A., Islanding Detection of Wind
Farms by Thd. IEEE Conf. DRPT2008 Nanjing China,
pp.2793-2797, Apr. 2008.
[23] Walmir Freitas, Zhenyu Huang, Wilsun Xu, A Practical
Method for Assessing the Effectiveness of Vector Surge
Relays for Distributed Generation Applications, IEEE Trans.
on Power Delivery, Vol.20, No.1, pp.57-63, Jan. 2005.
[24] Ankita Samui, S. R. Samantaray, Assessment of ROCPAD
Relay for Islanding Detection in Distributed Generation,
IEEE Trans. on Smart Grid, Vol.2, No.2, pp.391-398, JUNE
2011.
[25] H. Kazemi Karegar, B. Sobhani, Wavelet Transform Method
for Islanding Detection of Wind Turbines, Renewable
Energy, Vol.38, pp.94-106, 2012.
[26] Prakash K. Ray, NandKishor, Soumya R. Mohanty, Islanding
and Power Quality Disturbance Detection in Grid Connected
Hybrid Power System Using Wavelet and S-Transform,
IEEE Trans. on Smart Grid, Vol.3, No.3, pp.1082-1094, Sep.
2012.
217

pp.217:222
ISSN: 2251-9246
EISSN: 2345-6221
Efficient Parameters Selection for CNTFET Modelling Using

Artificial Neural Networks
Roya Abdollahzadeh Badelbo1, Fardad Farokhi2 and Alireza Kashaniniya3
1,2,3
Electrical Engineering Department, Islamic Azad University, Central Tehran Branch, Tehran, Iran. Email:
Roy.abdollahzadeh.eng@iauctb.ac.ir, f_farokhi@iauctb.ac.ir, ali.kashaniniya@iauctb.ac.ir
___________________________________________________________________________________
Abstract
In this article different types of artificial neural networks (ANN) were used for CNTFET (carbon nanotube transistors)
simulation. CNTFET is one of the most likely alternatives to silicon transistors due to its excellent electronic properties. In
determining the accurate output drain current of CNTFET, time lapsed and accuracy of different simulation methods were
compared. The training data for ANNs were obtained by numerical ballistic FETToy model which is not directly applicable in
circuit simulators like HSPICE. The ANN models were simulated in MATLAB R2010a software. In order to achieve more
effective and consistent features, the UTA method was used and the overall performance of the models was tested in MATLAB.
Finally the fast and accurate structure was introduced as a sub circuit for implementation in HSPICE simulator and then the
implemented model was used to simulate a current source and an inverter circuit. Results indicate that the proposed ANN model
is suitable for nanoscale circuits to be used in simulators like HSPICE.
Keywords: Artificial intelligence networks, CNTFET modelling, Fettoy, Fuzzy system, HSPICE.
_____________________________________________________________________________________________
based
on
numerical
piece-wise
non-linear
1. Introduction
approximation of the non-equilibrium mobile charge
With respect to the Nano scale devices, most of
density for CNTFETs [4], recently some new ways
them can be used in circuits. Due to the limitations
have been used to speed up the simulation in integrated
such as short channel effects, quantum effects and
circuits. Simulation with artificial intelligent networks
reduction of gate control on channel, performance of
is one of the powerful simulation methods. For Nano
Nano scale silicone MOSFET transistors will be
scale devices, because of some constraints and
declined [1]. Since the discovery of CNTs in 1991, due
quantum mechanical effects, the calculation of
to the suitable electrical and mechanical properties of
analytical equations are complex and time consuming.
carbon nanotubes, they were used in Nano electronic
Since the computation time in largescale circuits such
devices and they can be a good replacement for silicon
as VLSI must be done effectively, a model with fast
base devices [2]. Because of high current carrying
calculations is really needed. The artificial neural
capacity and low charge carrying scattering of CNTs,
network (ANN) model can be replaced with the Nano
CNTFETs are proper replacement for silicon
devices in simulators such as ADS, Hspice and
transistors and will have found many practical
Cadence. In previous studies, nanotransistors such as
applications in electronic industry in the near future
Double Gate FinFets, DG- MOSFETs, Nanoscale
[1].
MOSFETs. CNT- MOSFETs were simulated by
In addition to time-consuming numerical
various intelligent neural networks such as MLP
analytical methods such as NEGF (non-equilibrium
(Multi-Layer Perceptron), RBF and Neuro-fuzzy [1],
greens function) which is used to solve the
[5], [6], [7].
Schrdinger equations and finding the surface charge
and the density of states [3], or a simulation which is
218
In [1] an MLP network with two hidden layers (9,

8 hidden neurons) was used to simulate the CNTFET
drain current with mean relative error (MRE) of 1.09,
after 1000 epochs. Train and test data were obtained by
moscnt.1.0 model. To simulate the output drain current
in DGMOSFET in [6] with RBF network, the residual
error was 0.008. A Neuro-fuzzy structure with 15
neurons was used in drain current simulation of nano
scale MOSFET and 1.2705 mean square error (MSE)
was obtained [7].
CNTFET transistor doesnt have the nanoscale
silicon transistors limitations and many researchers
have been studied to find the suitable model for its
simulation. Ballistic effects are only included in some
models such as FETToy. Of course this model is not
applicable in circuit simulators like HSPICE, so we
need a model for elimination of time consuming steps
such as charge calculation process and the analytical
solution for Fermi Dirac integral [8].
In this paper we examine the simulation of
CNTFET transistor by MLP and neuro-fuzzy networks.
To obtain the training data, the CNTFETToy model in
Matlab R2010a is used. CNTFETToy, is a product of
Nanohub and Purdue University. Optimized geometry
of CNTFET consisting of a single carbon nanotube
semiconducting channel, completely surrounded by the
gate oxide, and perfectly contacted ends is taken into
consideration in this model (Fig.1) [9].
ISSN: 2251-9246
EISSN: 2345-6221
simulators such as HSPICE and it can be used in

nanoscale circuits.
2. Neural Networks Calculation

In this section, we will use an Artificial Neural
Network (ANN) to simulate CNTFET. Seven
CNTFETToy parameters and two voltages (vgs, vds)
are considered as inputs and Ids current as singleoutput for the network. The sufficient data for training
of the network are selected using CNTFETToy
environment. CNTFETToys seven parameters are
shown in Table 1.
Table.1.
Seven CNTFETToys input parameters.
parameter
Explanation
Nanotube diameter, nm
TOX
Oxide thickness, nm
Temperature
Dielectric constant, k
Ef
Source Fermi level
G, alphag
Gate control parameter
D, alphad
Drain control parameter
Fig.1. Optimized geometry assumed for CNTFETToy.
The basic parameters for CNTFETs will depend

on three groups including device, environment and
model. The device parameters are the diameter of the
nanotube, the gate dielectric thickness and dielectric
constant of the gate insulator. Parameters of the model
are specific to the underlying physics and are the Fermi
level of source, gate control parameter and drain
control parameter. Environmental variables are
temperature, voltage and values in the source/drain
series resistance [10].
In next section, MLP and Neuro-fuzzy networks
are introduced for modelling the device. Then feature
selection method is described. Finally, we show that
the proposed ANN model is applicable to run in
Fig.2. Simulation of CNTFET by ANN model.
Artificial intelligent networks (AINs) can be used

in the form of a block in circuit simulators for a device
and sub-circuits as shown in Fig.2.
219
2.1. Multilayer Perceptron (MLP)

MLP networks are composed of layers of neurons
that each one has an activation function and they are
connected through weight coefficients. By employing
an algorithm such as back propagation, training errors
are reduced. One of the most common activation
functions is hyperbolic tangent [11].
() = tanh = (
1 (2)
1+ (2)
2
) )
2
) )
=1 =1 ((
(5)
In above relation, t and s are free parameters for

center and width of the if part in fuzzy rules and w is
the free parameter for center in then part of the rule. N
is the number of training features and m is the number
of neurons.
(1)
2.3. UTA Method for Feature Selection
In MLP networks, several examples should

repeatedly be presented to the network, the network
must be trained to identify the desired output (2) by
modifying the weight coefficients using back
propagation algorithm through minimizing local
gradient of errors (
). Mean square error is recalled
in (4) and should be minimized. is the weight

m is the number of inputs applied to neuron j.
= (
=1 . )
1
= ( )2
2
The number of training samples is N.
=
=1
() =
=1 =1 ((
ISSN: 2251-9246
EISSN: 2345-6221
and
(2)
(3)
(4)
Feature selection methods are used for the

validation of networks features. Feature selection can
be used either before training or after that. For this
purpose, there are several ways. In this work UTA
method [12] which is done after training is used to
obtain the importance effect of CNTFET features. In
the UTA method, instead of each feature the constant
average value of that feature in all samples is replaced
and MSE error is re-measured each time and the results
impact on network features are determined. If a feature
is constant in all samples it cannot be effective in
identifying classes. Therefore the MSE error occurs in
three forms.
a) If the new MSE compared with the original
MSE unchanged, we can conclude that the
feature is not effective and the network will
not depend on it.
b)
If the new obtained MSE is greater than the

original MSE, it indicates that the feature is
effective.
c)
If the obtained MSE is less than the original

MSE the feature is not only ineffective but also
can be damaging.
Fig.3. Schematic diagram of MLP structure.
3. Implementation into HSPICE
Yj and Dj are MLP networks output and the

desired output respectively. The MLP neural network
structure to simulation is shown in Fig.3.
The proposed ANN model can be implemented as

a voltage-dependent current source depends on
CNTFET input parameters in HSPICE simulator. By
using the neural network weights and biases and simple
equations, the complex and time consuming quantum
equations can be simulated in HSPICE. We can state
the output current obtained by ANN as a sub-circuit as
follows:
2.2. Neuro-Fuzzy Network

Neuro-fuzzy networks consist of two parts of
neural network system and fuzzy logic system. Data
can be logged into system as fuzzy sets. To obtain the
optimal values of the free parameters which presented
in the fuzzy if-then rules and also the desired output,
the neural network is used. Parameters are modified
during learning process by gradient descent algorithm.
For this work, the center average method and Gaussian
is used for defuzzifier and membership function
respectively. The systems output is obtained as
follows (5):
.Subckt CNTFET nd ng ns
Ig ng ns 0
Gdrain nd ns cur=ANN equations of the proposed
model
.Ends CNTFET
220
Which nd, ng and ns are drain, gate and source nodes

respectively. Ig is the current between gate and source
and gdrain is the voltage-dependent current source
where the current Id is derived from the input
parameters of CNTFET (d, epsr, tox, T, alphag, alphad,
Ef, vgs and vds).
4. Results and Discussion
In order to get the ANN model, 12,500 data
obtained by CNTFET model is used, half of them for
training and the other half for testing. Seven
CNTFETToy input parameters and voltages of vgs and
vds are considered as input data and output current id
as single output. The range of minimum and maximum
values is shown in Table 2 [10].
ISSN: 2251-9246
EISSN: 2345-6221
neuro-fuzzy structure was implemented in HSPICE

simulator. Figure 4 compares the (i-v) properties
obtained from the numerical FETToy model, with the
neuro-fuzzy output which has been implemented in
HSPICE. According to figure 4, it can be concluded
that, for all bias points the neuro-fuzzy proposed model
is coordinate closely to numerical FETToy model.
The comparison between numerical FETToy
model and proposed ANN model, for testing data is
shown in Fig.5.
Table.2.
Input data ranges used in ANN.
PARAMETERS
Min
Max
d nm
4.0
51
5
04
04
54444
T Kelvin
544
014
EF
-4.45
-4.1
4.1
4.445
4.5
tox nm
K
Rs ohm-um
Fig.4. Current-voltage characteristics curve (Id-Vds) for CNTFET

using CNTFETToy model (circle) and proposed ANN model (solid
line).
Trying different structures, the MLP network

with two hidden layers (20, 10 neurons) with 0.1
learning rate was selected. Also, the back propagation
algorithm and tangent hyperbolic activation function is
used. For the Neuro-fuzzy network, from the other
hand, a structure consisting about 20 neurons as fuzzy
rule is selected. Comparison between these two
networks is shown in Table3.
Net.
Name
MLP
Neuro
-fuzzy
Table.3.
Comparison between MLP and neuro-fuzzy networks.
Max
Min
abs
abs
Train
Test
Layers
error
error
time
time
(in e- (in e6)
10)
20-10
6.077 20.664 0.36609 0.1583
20
7.53
1.937
0.26307
0.1574
Test
MSE
(ine-5)
9.935
6.133
By comparing the two structures, it seems the

neuro-fuzzy structure for low computation is relatively
fast and has somewhat lower error rate. According to
the results, to obtain the substitutive building block, the
Fig.5. The testing results of proposed ANN model.
In Table 4, after performing feature selection with

UTA method, test MSE for all CNTFET trained
features of CNTFETToy model, are shown for MLP
and neuro-fuzzy structures. Test MSE ratio for
CNTFET features after feature selection for both
structures are shown in Fig.6.
221
Table.4.
Test MSE after feature selection with UTA method.
network
MLP
Neuro-fuzzy
epsr
0.002
0.002
d
0.0017
0.0017
Tox(e-4)
3.56
4.89
T(e-4)
1.44
0.72
Ef
0.008
0.008
alphag
0.002
0.0023
Alphad(e-4)
2.2217
2.111
Vg
0.0237
0.0240
Vd
0.01
0.01
ISSN: 2251-9246
EISSN: 2345-6221
and obviously, the impact of all sources influence on

temperature has not been determined [10]. Some
experimental devices at low temperatures have been
reported [13], but we need to acquire devices which
work near the room temperature.
Fig.8. ANN CNTFET current source.
Fig.6. Test MSE ratio for CNTFET features.
By comparing the impact of various parameters on

the output current, it can be seen that the input voltages
has greatest effect and temperature parameter impact is
minimum. Although nanotubes have good effects even
at very high temperatures, but the effect of temperature
on the performance of CNFET is more complex and
unknown. In the survey, a large change based on
temperature, in simulation results have not been seen
To validate our proposed ANN model, we have

discussed CNTFET inverter and current source circuits.
In figure 7 the inputoutput signals for CNTFET
inverter with d=1.6e-9, tox=8e-9 and epsr=20 is shown.
The proposed model can be also used in expanded
circuits like current source (Fig. 8). In Fig. 9, the
output current and the reference current, in the current
source based on CNTFET circuit is shown.
Fig.7. The input-output signals for ANN CNTFET inverter.
222
ISSN: 2251-9246
EISSN: 2345-6221
[3] S. Datta, Nanoscale Device Modelling: the greens
[4]
[5]
[6]
Fig.9. Output characteristic for ANN CNTFET current source (IV) solid line for (Iref) and dashed line for (iout).
[7]
5. Conclusion
In this work, two types of neural networks were
employed to simulate the CNTFET. With comparing
the advantages and disadvantages of these two
structures, the appropriate structure is selected to be
implemented in HSPICE simulator. With respect to
accuracy and simulation speed in proposed model
against the numerical FETToy model, it can be
concluded that the Neuro-fuzzy network model is
suitable for modeling the CNTFET transistors.
Comparing the results with those obtained in [1], the
elapsed simulation time is 1.2 seconds for 100
samples in [1], whereas in this paper and for 6050
samples, the result takes the time of 0.15 seconds.
Feature selection algorithm running with UTA
method, we can see that the input voltages and
temperature are the most effective and the least
effective parameters respectively for determining the
output current. Finally with using our proposed ANN
model, we have discussed the analysis of nanoscale
circuits which can be also extended in large scale
circuits to save analysis time.
References
[1] M. Hayati, A. Rezaei and M. Seifi, CNT-MOSFET
Modelling Based on Artificial Neural Network:

Application to Simulation of Nanoscale circuits, SolidState Electronics, Vol.54, No.1, pp.5257, Oct 2010.
[2] M. Fakhrabadi, M. Samadzadeh, A. Rastgoo, M. Yazdi
and M. Mashhadi, Vibrational Analysis of Carbon
Nanotubes Using Molecular Mechanics and Artificial
neural network, Physica E, Vol.44, pp.565578, Oct
2011.
[8]
[9]
[10]
[11]
[12]
[13]
function method, Super lattices Microstruct, Vol.28,

No.4, pp.253278, 2000.
T.J. Kazmierski, D. Zhou, BM. Al-Hashimi, Efficient
Circuit-level Modelling of Ballistic CNT Using
Piecewise Non-linear Approximation of Mobile Charge
Density, IEEE int. Conf. Design, Automation, Test,
Munich, Europe, pp.14651, Mar 2008.
A. Abdollahi-Nohoji, F. Farokhi, M. Zamani,
Performance Comparison of Artificial Intelligence
Networks in Nanoscale MOSFET Modelling, IEEE
int. Conf. Natural Computation (ICNC), 26-28, PP.807
810, Jul 2011.
F. Djeffal, Z. Dibi, M.L. Hafiane and D. Arar, Design
and Simulation of a Nanoelectronic DGMOSFET
Current Source Using Artificial Neural Networks,
Materials Science and Engineering, Vol.27, pp.1111
1116, 2007.
A. Abdollahi-Nohoji, F. Farokhi, M. Shokouhifar, M.
Zamani, Efficent parameters selection for artificial
Intelligence Models of Nanoscale MOSFETs, IEEE
int. Conf. Electrical and Computer Engineering
(CCECE), Niagara Falls, Canada, Vol.24, pp.840844,
May 2011.
R. Yousefi and M. Shabani, A Model for Carbon
Nanotube
FETs
in
the
Ballistic
Limit,
Microelectronics Journal, Vol.42, No.11, pp.1299
1304, Sep 2011.
FETToy/matlab/CNTFET, FETToy-1.0, 2012. http://
www.nanohub.org/ resources/downloads.
J.L. HOFFA, Simulation of Carbon Nanotube Based
Field Effect Transistors, MSc, Thesis, Research and
Advanced Studies of the University of Cincinnati,
2007.
S. Haykin, Neural networks: A Comprehensive
foundation, NJ: Prentice-Hall, 1999.
M.F. Redondo, CH. Espinosa, A comparison among
Feature Selection Methods Based On Trained
Networks, Proc. IEEE Int. Neural Network for Signal
Processing, Madison, WI, USA, pp.205-214, Aug 1999.
S.J. Tans, A.M. Verschueren and C. Dekker, RoomTemperature Transistor Based on a Single Carbon
Nanotube, Nature, Vol.393, pp.49-52, 1998.

pp.223:230
223
ISSN: 2251-9246
EISSN: 2345-6221
Optimizing the Event-based Method of Localization in Wireless

Sensor Networks
Mohsen Jahanshahi1, Abdolvahed Moradpour2, Poorya Kiani3
1
Young Researchers and Elite club, Central Tehran Branch, Islamic Azad University, Tehran, Iran. Email: mjahanshahi@iauctb.ac.ir
2
Department of Information Technology, Qazvin Islamic Azad University, Qazvin, Iran. Email: a.moradpour@qiau.ac.ir
Department of Electrical Engineering, Shahreqods Islamic Azad University, Shahreqods, Iran. Email: pu.kiani@shahryariau.ac.ir
Abstract
A Wireless Sensor Network (WSN) is a wireless decentralized structure network consists of many nodes. Nodes can be
fixed or mobile. WSN applications typically observe some physical phenomenon through sampling of the environment so
determine the location of events is an important issue in WSN. Wireless Localization used to determine the position of nodes.
The precise localization in WSNs is a complex issue that requires consideration of many prominent aspects such as energy
consumption at the nodes as well as the algorithm execution time. In this article, we optimize a system called Spotlight. The
spotlight is a localization system that delivers high-location estimation accuracy at low cost. We propose several methods to
reduce execution time compared with previous methods in Spotlight. We proposed ILS, LAS and PAS methods that improve
execution time about 25%, 50% and 75%. Execution time of the proposed scheme is restricted by the size of deployment area.
Furthermore, in these methods, there is no need to equip the nodes with any special hardware.
Keywords: Localization, spotlight, linear scan, event base

1. Introduction
Recently WSNs received significant attention
from researchers due to its unlimited potential [20-22].
Determining location of the nodes in WSN is an
important issue. Different methods have been proposed
to solve this problem, including the use of GPS in
nodes that are not economically efficient. It also should
be noted that locations of the most sensor nodes are
fixed after deployment in the environment, and thus,
once localization is sufficient for each node. There are
many protocols reported in the literature that increase
energy consumption of the nodes due to exhaustive
messages exchanges among them.
In Spotlight method [1, 2], there is no need the

nodes to be equipped with GPS, only spotlight needs to
know its location. The core concept of the spotlight
localization system is the generation of controlled
events detectable by deployed sensor nodes. Events
like light and sound, with well-characterized
spatiotemporal properties which are detectable with
simple sensing hardware are performed well in this
system [1]. In this system, initially, the nodes start to
sync up their time. Then the spotlight will start to
spread events. Detection time of events is recorded by
the nodes, and then system can calculate the location of
all nodes.
224
Fig.2.The line scan method, tow spotlights do scan horizontally and

vertically
ISSN: 2251-9246
EISSN: 2345-6221
time difference of arrival (TDOA), angle of arrival

(AOA), and received signal strength (RSS). Many
algorithms have been proposed based on these methods.
Almuzaini and Gulliver propose a range-based algorithm
which is based on decision tree classification and the
density based spatial clustering of applications with
noise (DB SCAN) algorithm in [18]. Lei Zhang and
Binwei Deng in [19] proposed a novel range-based
distributed localization algorithm based on genetic
algorithm uses sampling method to obtain initial location
and refines the initial location based on genetic
algorithm.
In this article, we extend the idea proposed in [1]
that focuses on design paradigms in event-based
localization systems.
2.1 Spotlight system architecture
The architecture of the Spotlight system comprises
the following three functions:
Distribution function, or EDF, ()
Event detection function, ()
Localization function, ()
Since () is computed by the Spotlight device,
hardware requirements for the sensor nodes remain
minimal. Substantial algorithmic changes can be made
without requiring updates on deployed sensor nodes [1].
The EDF E (t) may be tuned to distribute events
optimally based on the limitations imposed by sensor
capabilities, the platform transporting the spotlight
system, terrain, and availability of detailed geographic
information. The Point Scan, Line Scan, and Area Cover
Event Functions each illustrate basic functionality of the
Spotlight localization system.
Fig.3.The area scan with events covering the shaded areas.
The rest of this article is organized as follows. In

section 2, we review related works. In section 3 we
propose improved methods and in section 4 their
performance is presented. Section 5 concludes the paper.
2. Related Works
Location information is crucial for many
applications in wireless sensor networks. Localization
algorithms in WSNs can be classified into two general
categories: range-based localization [6, 7, 8, 9, 10 and
17] and range-free localization [11, 12, 13, 14, 15, and
16]. Range-based method uses the measured
distance/angle to estimate the location. And the free
localization method uses the connectivity or pattern
matching method to estimate the location. Measuring the
range of wireless signal transmissions is a key for rangebased localization techniques. The common methods to
estimate the indoor location are time of arrival (TOA),
2.1.1. Point Scan Event Distribution Function

Spotlight device generates point events (e.g.,
light spots) in an area 2 along the x-axis (Figure.
1). It is assumed that the scanning speed is a constant s,
the deployment area is A=l*lr , the radius of the event is
r, and finally the EDF is given by:
() = { |(, ) = () ()
= [/]}
In this case, the resulting localization function is:
() = (1 ) = { () () , [ 1 /]}
(1)
225
Criterion
ISSN: 2251-9246
EISSN: 2345-6221
Table1.
Execution Cost Comparison Criterion
Point Scan
Line Scan
Area Cover
Localization Time
( / )/
(/)/
# Event Detection
Event Overhead
Fig.1.The Point Scan method. Spotlight scan environment by moving

on X-axis and distribute events.
3. Improved methods
2.1.2. Line Scan Distribution Function
In this proposed method, the environment is

divided into regions. Initially, the spotlight1 is in
position (/2, 0) as it is shown in Fig.5. In this Figure,
long arrows depict direction of the movement. Small
arrows also show direction of scan. Nodes in the scanned
area record detection time of events.
In this method, two spotlights do scan horizontally

and vertically (Fig.2). Therefore, each node should
records two events. = {1 , 2 }. Formula 1 is used to
calculate the position of each node on X-axis and
Formula 2 is used to calculate the position on Y-axis.
Based on formula 3, estimated location for any nodes is
calculated from Subscribe points that calculated by
formula 1 and 2.
3.1 Improved Line Scan (ILS)
() = {| = (, ) [0, ]} (1)
() = {| = (, ) [0, ]} (2)
Where 1 [0, /] 2 [/, 2/]
() = (1 ) (2 )
(3)
2.1.3. Area Cover Event Distribution Function

The Point Scan and Line Scan functions require
precise tracking of the event generator orientation. Area
Cover EDF lessens the precision required by using
devices, like light projectors, to generate events that
cover an area. It can be illustrated with a simple
example. As shown in Figure 3, the plane A is divided in
16 sections and each section k is assigned a unique code
k. The spotlight device distributes events according to
these codes: at time j, a section k is covered by an event
(illuminated, in the case of a visible light event) if j th bit
of k is 1. A node residing anywhere in a section k is
localized to that sections center. For example, nodes
within section 1010 detect the events at time T = {1,
3}. k can be determined at = 4 for any sensor in the
area covered by the event generator.
Assume that all sensor nodes are located in a
square with edge size D, and that the Spotlight device
can generate N events (e.g., Point, Line, and Area Cover
events) every second, and that the maximum tolerable
localization error is r. Table 1 compares the execution
cost of the three techniques [1].
Fig.5. ILS method-initial position of spotlight1
And as it is shown in Figure 6, initial position of

spotlight2 is(/2, ). After step1, the following times
were recorded by the nodes in each region (Figure 7).
Fig.7. ILS method - detected times after step1
Figure
226
Fig.8. ILS method -detected times after step2
In the second step, spotlight1 moves to the right

and spotlight2 moves to the left. After end of this step,
times detected by the nodes would be according to
Figure 8. Finally, the spotlight1 moves down and
spotlight 2 goes back to right (Fig.9).
Fig.9. ILS method detected times after step3
As it can be seen, each area has a unique diagnostic

time and by this times, we can determine which nodes
are located in one area by following algorithm:
First assume that t1, t2, t3 are:
1 = {| 0 < <= /3}
2 = {| /3 < <= 2/3}
3 = {| 2/3 < <= }
To determine the position of nodes we divide the area
into four regions.
ISSN: 2251-9246
EISSN: 2345-6221
If (dt1 t1 and dt2 3 )

//
zone 1
{// dt1 and dt2 are times detected by sensor
and t is total scan time
x=( dt2-(2t/3)) * linewidth
// linewidth is width of the scanning beam
y= (|(2t/3)-dt1|+1*linewidth
}
else if(dt1t1 and dt2 2 )
//
zone 2
{x=dt2 * linewidth
y= (|(2t/3)-dt1|+1*linewidth}
else if ( count(detected times) == 3 )
zone 3
{x= ( | dt2-(2t/3) | +1)* linewidth
y= dt1* linewidth}
else if (dt1 t2 and dt2 3 )
zone 4
{x=dt1*linewidth
y= ((t-dt2) +1)*linewidth}
//
//
The execution time of ILS method is (3/

2) /. Compared with the line scan method, execution
time reduced. The improved time can be used to
improve the location accuracy.
3.2 Hybrid methods
3.2.1 Point scan & Area scan (PAS)
PAS method is a combination of the point scan and
area scan. In this method, the environment is divided
into four regions as shown in Figure 10.
227
ISSN: 2251-9246
EISSN: 2345-6221
As it is depicted in Figure11, two devices scan the

field by moving across pre-specified paths. Recording
times are shown in Fig.12.
Fig.10. PAS method
Four spotlights simultaneously scan the regions. A

binary number can be assigned to each region:
Region 1: 00
Region 2: 01
Region 3: 10
Region 4: 11
Then, area scan method is used to distinguish
between regions. For example, the third region is
scanned at the second time and fourth region is scanned
at the first and second time. So each node can record 3
times. The second and third times indicates the region
and the first time can determine the position of nodes in
the area. Method's execution time is equal to (D2/4) +2.
Fig.12. Recording times in LAS method
Then, area scan method is used to distinguish

between regions as in the PAS method. Method's
execution time is equal to D +2.
4. Exprimental Results
In this paper, we have presented several methods
called ILS, LAS and PAS. In the following we analysis
their performance with other similar methods from the
perspective of execution time
3.2.2. Line scan & Area scan (LAS)
4.1. experiment 1
In this method, the environment is divided into four

regions as shown in Fig.11.
In this experiment we compared execution time for

ILS method with line scan method for r=3cm, r=13cm,
r=23cm and r=33cm where r is radius of the events.
Fig.13.a. execution time for ILS method compared with line scan method
where r=3cm
Fig.11. LAS method.
228
Fig.13.b. Execution time for ILS method compared with line scan
method where r=13cm
Fig.13.c. Execution time for ILS method compared with line scan
method where r=23cm
Fig.13.d. execution time for ILS method compared with line scan method
where r=33cm
ISSN: 2251-9246
EISSN: 2345-6221
Figure 14.a. Execution time for LAS method compared with line
scan method where r=3cm
Fig.14.b. Execution time for LAS method compared with line scan
method where r=13cm
Fig.14.c. Execution time for LAS method compared with line scan
method where r=23cm
As can be seen, the execution time of ILS method

is reduced by about 25% compared to the line scan
method. This reduction in execution time is due to the
simultaneously scanning by tow spotlights.
4.2. Experiment 2
In this experiment we compared execution time for
LAS method with line scan method for r=3cm, r=13cm,
r=23cm and r=33cm (figure14), where r is radius of the
events.
Fig.14.d. Execution time for LAS method compared with line scan
method where r=33cm
229
Execution time of LAS method is less than the line

scan method because in this method the two spotlights
do scan simultaneously and for distinguish between
different regions we use area scan method.
5. Conclusion
This paper improved the event-based localization
methods in wireless sensor networks. The proposed
method reduces the execution time of line scan method
from 2 to 3/2 where D is the width of environment.
It also presents two hybrid methods, PAS and LAS. The
former, reduces execution time of the point scan method
from 2 to (2 /4) + 2 and latter, reduces execution
time of the line scan method form 2 to + 2. As for
our future works, we will more concentrate on the hybrid
approaches and try to improve their accuracy.
References
[1] R.Stoleru, Member,S. Mathiharan,Stephen M. George, and John A.
Stankovic, "Asymmetric Event-Driven Node Localization in
Wireless Sensor Networks "
IEEE TRANSACTIONS ON
PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 23, NO. 4,
APRIL 2012.
[2] R. Stoleru, T. He, J.A. Stankovic, and D. Luebke, A HighAccuracy Low-Cost Localization System for Wireless
SensorNetworks, Proc. Intl Conf. Embedded Networked Sensor
Systems (ENSS), 2005.
[3] Z. Zhong, D. Wang, and T. He, Sensor Node Localization Using
Uncontrolled Events, Proc. Intl Conf. Distributed Computing
Systems (ICDCS), 2008.
[4] R. Stoleru, T. He, and J.A. Stankovic, Walking GPS: A Practical
Localization System for Manually Deployed Wireless Sensor
Networks, Proc. 29th Ann. IEEE Intl Conf. Local Area
Networks, 2004.
[5] J. Yick and B. Mukherjee, Wireless
survey,Computer Networks, 2008.
sensor
network
[6] P. Bahl and V.N Padmanabhan, "RADAR: An In-Building RFbased User Location and Tracking System,"INFOCOM 2000.
[7] T.C. Karalar and J. Rabaey,"An RF ToF Based Ranging
Implementation for Sensor Networks", Communications, 2006.
[8] Xiuzhen Cheng and Thaeler A, TPS: a time-based positioning
scheme for outdoor wireless sensor networks, Twenty-third
AnnualJoint Conference of the IEEE Computer and
Communications Societies, INFOCOM 2004.
[9]
D. Niculescu and B. Nath, Ad hoc positioning system (APS)

using AOA, Twenty-Second Annual Joint Conference of the
IEEE Computer and Communications, INFOCOM 2003.
[10] P. Kumar and L. Reddy, Distance measurement and error

estimation scheme for RSSI based localization in Wireless Sensor
Networks, Wireless Communication and Sensor Networks
(WCSN), 2009.
[11] D Niculescu and B Nath, DV based positioning in ad hoc
networks. Telecommunication Systems, Springer, 2003.
[12] Yi Shang and Wheeler Ruml, Localization from mere
connectivity,Proceedings of the 4th ACM international
symposium on Mobile ad hoc networking & computing, 2003.
ISSN: 2251-9246
EISSN: 2345-6221
[13] R. Nagpal and H. Shrobe, Organizing a Global Coordinate

System from Local Information on an Ad Hoc Sensor Network,
Information Processing in Sensor networks, 2003.
[14] Tian He and Chengdu Huang, Range-Free Localization Schemes
for Large Scale Sensor Networks, Proceedings of the 9th annual
international conference on Mobile computing and networking,
2003.
[15] N. Bulusu and J. Heidemann, "GPS-less low-cost outdoor
localization for very small devices", Personal Communications,
IEEE, 2000.
[16] L. Doherty and K.S.J. Pister, "Convex Position Estimation in
Wireless Sensor Networks", INFOCOM 2001.
[17] P.N. Pathirana, A. Savkin, S. Jha, and N. Bulusu, "Node
Localization Using Mobile Robots in Delay-Tolerant Sensor
Networks", IEEE Trans. Mobile Computing, vol. 4, no. 3, pp. 285296, May/June 2005.
[18] K.K ,Almuzaini ; T.A, Gulliver. ,Range-Based Localization in
Wireless
Networks
Using
the
DBSCAN
Clustering
Algorithm,Vehicular Technology Conference (VTC Spring),
2011
IEEE
73rd,
Digital
Object
Identifier:
10.1109/VETECS.2011.5956252 Publication Year: 2011
,
Page(s): 1 7
[19] L.Zhang ;B.Deng, A new range-based localization algorithm for
wireless sensor networks, Computing, Communication, Control,
and Management, 2009. CCCM 2009. ISECS International
Colloquium on
Volume: 1 Digital Object Identifier:
10.1109/CCCM.2009.5268137 Publication Year: 2009 , Page(s):
111 114.
[20] M. Jahanshahi, M. Maddah, N. Najafizadegan Energy Aware
Distributed Partitioning Detection and connectivity Restoration
Algorithm in Wireless Sensor Networks, International Journal of
Mathematical Modelling & Computations (IJM2C), Vol. 03, No.
01, 2013, 71- 82
[21] M. Jahanshahi, S. Rahmani, S. Ghaderi, An Efficient Cluster
Head Selection Algorithm for Wireless Sensor Networks Using
Fuzzy Inference Systems, The International Journal of Smart
Electrical Engineering (IJSEE), Vol.2, No. 2, 2013
[22] M. Jahanshahi, M. R. Meybodi and M. Dehghan, Cellular
Learning Automata based Scheduling Method for Wireless Sensor
Networks, 14th annual International CSI Computer Conference
(CSICC09), Tehran, Iran, July 1-2, 2009
231

pp.231:236
ISSN: 2251-9246
EISSN: 2345-6221
Power System Reliability Analysis with Emergency Demand

Response Program
Rahmat Aazami1, Nosratallah Mohammadbeigi1, Hadi Mirzaei1, Ali Mansouri1, Ehsan Mohamadian1
1
Department of Electrical Engineering, Science and Research Branch, Islamic Azad University, Ilam, Iran.
Abstract
With the development of restructured power systems and increase of prices in some hours of day and increase fuel price, demand
response programs were noticed more by customers. demand response consists of a series of activities that governments or utilities
design to change the amount or time of electric energy consumption, to achieve better social welfare or some times for maximizing
the benefits of utilities or consumers. In this paper the effect of emergency demand response program on composite system reliability
of a deregulated power system is evaluated using an economic load model, AC power-flow-based load curtailment cost function and
reliability evaluation techniques. In this paper for calculation the reliability indexes, the Emergency Demand Response Program
(EDRP) cost is considered and in each contingency state, the EDRP cost with the customer load curtailment cost is compared and the
load appropriate value is selected for load shedding or participating in EDRP. In the next stage, the system and nodal reliability
indexes are calculated. To investigate the impact of EDRP activity on composite reliability of restructured power systems the IEEE 6
bus Roy Billinton Test System is utilized. According to obtained results, EDRP using lead to increasing nodal and system reliability.
It can be said that solving problems such as congestion in transmission lines, power system reliability decrease at load network peak
hours, is impossible without customer interfering in power market. In other hand Consumer participation, makes the power markets
more competition and enhance its performance.
Keywords: Emergency demand response program (EDRP), power system deregulation, reliability.
1. Introduction
Participant of customers in electricity market increases
the competitiveness of electricity markets. When customers
see price volatility, they modify their demand which helps
the magnitude of price spikes be reduced. When consumers
can receive price signals and can respond to them, some of
them will shift their demand to cheaper hours when they face
high prices. Demand Response (DR) can be defined as
the changes in electric usage by end-use customers

from their normal consumption patterns in response to
changes in the price of electricity over time. [1-7].
DR is divided into two basic groups and several
subgroups [1-7]:
A- Incentive-based programs:
A-1- Direct Load Control (DLC)
A-2- Interruptible/curtail able service (I/C)
A-3- Demand Bidding/Buy Back
A-4- Emergency Demand Response Program
(EDRP)
A-5- Capacity Market Program (CAP)
A-6- Ancillary Service Markets (A/S)
B- Time-based programs:
B-1- Time-of-Use (TOU) program
B-2- Real Time Pricing (RTP) program
B-3- Critical Peak Pricing (CCP) Program
232
The benefits of DR include increased static and

dynamic efficiency, better capacity utilization, pricing
patterns that better reflect actual costs, reduction of
price spikes, decentralized mitigation of market power,
and improved risk management.
A recent study estimated the prospective benefits
of active demand response at $7.5 billion by 2010 (ICF
2002). Other studies, described in GAO (2004), give
further details of the benefits that have already been
generated because of demand response and active retail
choice [2-7].
Emergency Demand Response program (EDRP)
is the most usual demand response program when an
event occurs. EDRP provides incentives for customers
to reduce loads during reliability events, though the
curtailment is voluntary. No penalty is assessed if
customers do not curtail, and the rates are prespecified, though no capacity payments are received
[3] [7].
Emergency Demand Response Program (EDRP)
is a reliability-specific day-of interpretability contract
that is available for hours when there is a shortfall in
reliability reserves. Customers can choose to allow the
ISO to interrupt their service, for which the customer is
paid a price determined through a bidding process [3]
[7].
EDRP is an emergency DR program that provides
mechanisms where demand can be reduced on short
notice when reserve shortfalls are forecasted. EDRP is
a voluntary emergency program that pays customers an
incentive which, for example, is more than 500 $/MWh
in New York power market or can be the prevailing
real-time market price for curtailments of at least four
hours long when called by the ISO [3].
The New York Independent System Operator
(NYISO) calculated that its demand response program
provided substantial benefits to the market by helping
the power grid recover from the August 2003 Blackout.
Specifically, they estimated that on August 15, 2003,
the participating DR of 593.9 MW provided $50.8 M
(US) worth of economic benefits at a cost of $5.9 M
(US).
During August 2001, higher-than-normal
temperatures forced the NYISO to invoke emergencies
on August 7, 8, and 9 (18 hours in all zones) and on
August 10 (4.5 hours in New York City/Long Island
and Hudson River, Zones FK). On August 9th, a new
record peak load of 30,983 MW was established. Most
of the capacity shortfall occurred in the New York
City/Long Island area (Zones JK). During this time, a
ISSN: 2251-9246
EISSN: 2345-6221
variety of load management programs, including the

PRL1 programs (EDRP, DADRP2, and ICAP), were
deployed. At peak load, an estimated 1,580 MW was
curtailed, of which the PRL programs contributed 605
MW (38 percent), with the balance coming from other
sources. At the time the EDRP events were called, 292
participants had registered in the EDRP. Participants in
the EDRP provided 70 percent of all load curtailment
from all PRL programs. While 292 participants (712
MW) registered with the NYISO for EDRP, only 213
(617 MW) actually performed when emergencies were
declared. Those who performed delivered only an
average 418 MW per hour, or 68 percent of their
registered capability. A planning consideration for
future rounds of the EDRP, given that it is a voluntary
program, is that more loads have to be registered than
is actually required [3]-[4].
There is a growing concern about the reliability
of power systems under a market environment,
especially after the blackouts in North America and
Europe in 2003.
Bulk power system operators primarily rely on
adjustments in generation output (MW movements up
or down) to keep the system reliability.
In principle, changes in electricity demand could serve
as well as generator movements in meeting the
reliability requirements [5]. So, customer loads could
be able to participate in these markets. The
participation of these resources will either enhance
reliability or lower costs of maintaining reliability for
all customers and will save money for participating
customers.
This paper investigates the impacts of
emergency demand response program on system and
nodal reliability in a state enumeration approach. A
small reliability test system, RBTS, is studied for
which the simulation results show that, using
emergency demand response program, the system and
nodal reliability is improved.
This paper is organized in five sections. Section 2
defines the load economic model which is used to
evaluate the participation in emergency demand
response program and explains the economic analysis
formulation. Reliability Index Calculation is discussed
in section 3. Section 4 presents the numerical results
which have been tested on RBTS and finally section 5
is dedicated to the conclusions.
1
2
Price Responsive Load

Day-Ahead Demand Reduction Program
233
1. Demand Response Economic Modelling

In the beginning of the deregulation, usually
consumers did not have effective participation in the
power markets and therefore they were not able to
response to the prices effectively. However, the
development of the restructured power systems has
been accompanied by many problems, for example
reduced system reliability.
Fig. 2 shows how the demand elasticity could
effect on electricity prices [6-7].
ISSN: 2251-9246
EISSN: 2345-6221
the relative change in demand is smaller than the

relative change in price, the demand is said to be
inelastic. So the elasticity coefficients can be arranged
in a 24 by 24 matrix E [6-7].
The detailed process of modelling and
formulating how the EDRP program effects on the
electricity demand and how the maximum benefit of
customers is achieved, have been discussed in [7].
Accordingly the final responsive economic model is
presented by (3) [7]:
24
d (i)
d (i) d 0 (i) E0 (i, j ). 0 A( j )
0 ( j)
j 1
E (i)[ (i) 0 (i) A(i)]

i 1,2,...,24.
0 (i)
(3)
The above equation shows how much should be

the customer's demand in order to achieve maximum
benefit in a 24-hours interval [7].
Time period is assumed to be one hour. Variable
load curve for 24 hours within one day are considered
in the simulations [7].
Fig.1. Effect of demand variation on the electric energy price [6]
2. Modelling of EDRP
Elasticity is defined as the ratio of the relative change in

demand to the relative change in price [7]:
The final response of the economic model is

presented by (4). The modified model for showing the
effect of EDRP is as follows:
q 0 dq
.
q0 dp
(1)
Where:
d (ti ) : Demand changes in time interval t i
(ti ) : Price changes in interval t i
(t j ) : Price changes in time interval t j
For i = EDRP Non-Event Hours
According to equation (2), self elasticity ( ii ) and

cross elasticity ( ij ) can be written as:
d (t i )
ii
(t i )
d (t i )
ij
(t j )
(4)
For i = EDRP Event Hours
d0
18
d (i)
d (i) d 0 (i ) E0 (i, j ). 0 A( j )
0 ( j)
j 14
(2)
18
d (i)
E (i ) A(i ) (5)
d (i) d 0 (i) E0 (i, j ). 0 A( j )
(
j
)
0 (i )
j 14
0
d0
3. Reliability Index Calculation
Reliability assessment methodologies of bulk

power systems are clearly described in [8].
A composite system contains both generation and
transmission facilities and is sometimes designated as a
composite generation and transmission system.
According to the method of selecting system state,
there are two basic methods: state enumeration and
Monte Carlo sampling.
Reliability assessment of a composite system
generally involves the solution of the network
Where:
d (ti ) : Demand changes in time interval t i
(ti ) : Price changes in interval t i
(t j ) : Price changes in time interval t j
Self elasticity and cross elasticity are negative
and positive values, respectively. If the relative change
in demand is larger than the relative change in price,
the demand is said to be elastic, on the other hand, if
234
configuration under random outage situations

(contingencies). Various techniques, depending upon
the adequacy criteria used and the intent behind these
studies, are used in analysing the adequacy of a power
system. The three basic techniques used in network
solutions are as follows:
1) A network flow method
2) DC load flow method
3) AC load flow method
If the quality of power supply including
acceptable voltage levels and appropriate generating
unit MVAr limits is an important requirement, more
accurate AC load flow methods such as NewtonRaphson, Gauss-Seidel techniques must be utilized to
calculate the reliability indices.
The techniques of identifying and analysing
problems in a system state are the same. These include
power flow and contingency analysis for problem
recognition and optimal power flow for remedial
actions. In our following simulation, the enumeration
simulation method is adopted to select system states.
The formulation of load curtailment cost
determination under contingency S using AC load flow
and customer interruption load cost and EDRP cost and
generation cost can be depicted by the optimization of
Equation (6).
Min CostPGi ,QGi +Cost P EDRP ,Q EDRP +
CostPCLC , QCLC
for S 1,2,3,..., N
Pgimin Pgi Pgimax
i 1,..., n
Qgimin Qgi Qgimax
i 1,..., n
min
max
PEDRP
PEDRP PEDRP
min
max
QEDRP
QEDRP QEDRP
0 PCLC PLoad
0 QCLC QLoad
3
Cost PCLC , QCLC : is the customer interruption cost
buses under contingency S.
Cost P EDRP,Q EDRP : is the EDRP cost for ISO.
PG and PD are generation output and load power buses
vector.
Pgimin and Q gimin is the minimum output of power of
generators; Pgimax and Q gimax is the maximum output of

power of generators vector.
PEDRP and QEDRP are EDRP power values vector.
PCLC and QCLC are curtailment load values vector.
V is voltage buses vector.

T is power flow on a branch.
T max is maximum capacity limit of a line or

transformer.
This optimal model of load curtailment cost is a
problem that is solved by mat power software.
Indices of system reliability are:
EDLC: Expected duration of load curtailment of the
overall system:
24 NC
D j , i
i =1 j=1
EDLC
24
(hours / day )
(13)
Where
N C = Number of system interruptions in hour i.
(7)
Dj,i = Duration of the jth system interruption, in hour i.

LOLP: Loss of load probability of the overall system:
EDLC
(14)
LOLP =
(occurrence / day)
(8)
EDNS: Expected demand not supply of the overall

system:
(6)
P gi - Pdi P EDRPi PCLC i Pi ,V i 1,..., n

Qgi Qdi QEDRPi QCLC i Qi ,V i 1,..., n
ISSN: 2251-9246
EISSN: 2345-6221
24
24 NC
DNS Sys
j, i
j=
1
( MW / day )
(15)
EDNS =
24
= System demand not supplied in MW for the
DNS Sys
j, i
i =1
(9)
(10)
V min V V max
(11)
T T max
(12)
S: set of all failure system state.

N : Number of system failures.
c
Cost PGi ,QGi : is generators cost in power system.
jth interruption, in hour i.

EENS: Expected energy not supplied, in
[MWh/day], is the total amount of energy which is
expected not to be delivered to the loads.
24 NC
ENS Sys
j ,i
i =1 j=1
( MWh / day )
(16)
EENS =
24
= System energy not supplied in MWh for the
ENS Sys
j ,i
jth interruption, in hour i.
CLC: Customer Load Curtailment
235
Indices of load point reliability are:

LPEDLC: Load point expected duration of load
curtailment
24 NC, k
D kj , i
i =1 j=1
(hours / day )
(17)
LPEDLC
24
N C,k = Number of interruptions occurring in hour i, at
The amount of incentive in EDRP program

formulation is assumed to be equal to 500 $/MWh
(existing incentive in New York market). The elasticity
of the load is shown in Table1.
Table.1.
Self and cross elasticities
N C, k
DNS Busk
j,i
i =1 j=1
LPEDNS =
24
( MW / day )
(18)
= Demand not supplied in MW for the jth

DNS Busk
j ,i
interruption, in hour i at Bus k.
LPEENS: Load point expected energy not supplied
(MWh/day)
N c ,k
ENS Busk
j, i
i =1 j=1
LPEENS =
24
5.8
Low
0.0024
0.002
-0.02
x 10
5.4
5
4.6
( MWh / day )
(19)
= Energy not supplied in MWh for the jth
interruption, in hour i at Bus k.

4. Numerical Results
In order to show the effect of emergency demand
response program on system reliability of a deregulated
power system, a case study based on the IEEE 6-bus
system is presented in this section. Roy Billinton Test
System (RBTS) has 11 generating units, with the total
installed capacity of 240 MW and a total system peak
demand of 185MW spreading out among 5 system
buses. The single line diagram of RBTS is shown in
Fig. 2.
Demand (MW)
ENS
Off-Peak
0.0032
-0.02
0.002
Load curve of Mid-Atlantic region New York

network was selected for testing and analysing the
effect of EDRP program, Fig. 3 [4]. The load curve is
divided into three intervals: low load period (12.00
p.m. to 9:00 a.m.), off-peak period (10:00 a.m. to 13:00
p.m. and 19:00 p.m. to 12:00 p.m.) and peak period
(14:00 p.m. to 18:00 p.m.).
24
Busk
j ,i
Peak
-0.02
0.0032
0.0024
Peak
Off-Peak
Low
Bus k.
D kj , i = Duration of the jth interruption in hour i at Bus k.
LPEDNS: Load point expected demand not supply:
24
ISSN: 2251-9246
EISSN: 2345-6221
4.2
3.8
3.4
3
2.6
2.2
1.8
1.4
1
10
12
14
hour
16
18
20
24
Fig3. Load curve of Mid-Atlantic region New York network
The load curves before and after implementation

of emergency demand response program is represented
in Fig. 4. As it can be seen, by implementation of
emergency demand response program, based on the
difference between elasticities in different periods,
loads are transferred from peak periods to valley
periods. Without emergency demand response
program, the IEEE 6-bus system peak load is 315 MW;
considering demand response programs, however, the
IEEE 6-bus system peak load is 285.64MW.
x 10
Demand With EDRP

Demand Without EDRP
Load (MW)
4.8
4.6
4.4
4.2
4
8
Fig.2. Single line diagram of the RBTS
22
10 11 12 13 14 15 16 17 18 19 20 21 22
hour
Fig4. Effect of EDRP in Mid-Atlantic load curve
236
In order to show the effect of emergency demand

response program on the load curve and system and
nodal reliability of a deregulated power system, the
same test system, RBTS, has been simulated using the
reliability evaluation techniques. Above program
maximize the profit of customers moreover influencing
the system and load point reliability.
Two scenarios will be observed in this paper: 1Test of system without considering emergency demand
response program, 2- Test of system with considering
emergency demand response program.
The simulation results for the reliability of total
system have been shown in Table II.
Table.2.
System Reliability Indices of the RBTS
Without
Considering EDRP
Considering EDRP
LOLP
0.17875
0.135462
EDLC (h)
0.1282
0.0985
EDNS (MW)
942.673
629.626
EENS (MWh)
112.2254
28.1364
For system nodes the results can be seen in Table

III, IV, and V.
The results show that emergency demand
response program improves the reliability of the
system. Comparing the nodal reliability indices with
and without considering emergency demand response
program, it can be seen that the nodal reliability is also
improved when emergency demand response program
is considered.
Table.3.
Load Point Expected duration of load curtailment [Hour]
Without
Considering
Considering
EDRP
EDRP
Load 6
0.036547
0.031231
Load 3
0.036112
0.035024
Load 4
0.035101
0.035002
Load 5
0.035004
0.034987
Table.4.
Load Point Expected demand Not Supplied (MWh/day)
Without Considering
Considering EDRP
EDRP
Load 6
1199.3826
772.2243
Load 3
261.3892
113.3993
Load 4
114.2587
76.8648
Load 5
498.8546
416.0093
Table.5
Load Point Expected Energy Not Supplied (MWh/day)
Without Considering
Considering
EDRP
EDRP
Load 6
63.1254
59.7458
Load 3
29.5468
27.0125
Load 4
27.2548
25.6074
Load 5
34.8546
32.8425
ISSN: 2251-9246
EISSN: 2345-6221
5. Conclusion
This paper evaluated the effects of demand
response programs especially emergency demand
response program on system and load point reliability
of a deregulated power system using an economic load
model, AC power-flow-based load curtailment cost
function and reliability evaluation techniques.
In this paper for calculation the reliability
indexes, the Emergency Demand Response Program
(EDRP) cost is considered and in each contingency
state, the EDRP cost with the customer load
curtailment cost is compared and the load appropriate
value is selected for load shedding or participating in
EDRP. In the next stage, the system and nodal
reliability indexes are calculated.
From the simulation results it can be seen that
emergency demand response program improves the
system reliability and nodal reliability of a deregulated
power system.
According to obtained results, EDRP using lead
to increasing nodal and system reliability. It can be said
that solving problems such as congestion in
transmission lines, power system reliability decrease at
load network peak hours, is impossible without
customer interfering in power market. In other hand
Consumer participation, makes the power markets
more competition and enhance its performance.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
C. River, Primer on demand side management, Report for the

World Bank, February 2005.
U. S. Department of Energy, Energy policy Act of 2005,
section 1252, February 2006.
FERC report, Regulatory commission survey on demand
response and time based rate programs/ tariffs, August 2006.
Available at www.ferc.gov
www.nyiso.com
J. Wellinghoff, Collaborative dialog on demand response,
FERC report, 12 Nov. 2006, available at www.ferc.gov.
D. S. Kirschen, G. Strbac, P. Cumperayot, D. Mendes,
Factoring the elasticity of demand in electricity prices, IEEE
Trans. on power sys. , Vol.15, No.2, pp.612-617, May 2000.
H. Aalami, G. R. Yousefi and M. Parsa Moghadam, Demand
Response Model Considering EDRP and TOU Programs,
IEEE/PES Transmission and Distribution Conference &
Exhibition, 2008, Chicago, USA.
R. Billinton and R. N. Allan, Reliability Evaluation of Power
Systems, 2nd Edition, Plenum Press, New York, 1996.
237

pp.237:244
ISSN: 2251-9246
EISSN: 2345-6221
Design of Maximum Power Point Tracking in Solar Array

Systems Using Fuzzy Controllers
Marjaneh Hashemi1, Shahram Javadi 2
1
Electrical Engineering Department, Islamic Azad University Central Tehran Branch, Tehran, Iran. Email: marjaneh.hashemii@yahoo.com
2
Islamic Azad University Central Tehran Branch, Tehran, Iran. Email: sh.javadi@iauctb.ac.ir (Corresponding Author)
Abstract
In recent year's renewable energy sources have become a useful alternative for the power generation. The power of
photovoltaic is nonlinear function of its voltage and current. It is necessary to maintain the operation point of photovoltaic in
order to get the maximum power point (MPP) in various solar intensity. Fuzzy logic controller has advantage in handling
non-linear system. Maximum power point trackers are so important in photovoltaic systems to increase their efficiency.
Many methods have been proposed to achieve the maximum power that the PV modules. This paper proposed an intelligent
method for MPPT based on fuzzy logic controller. The system consists of a photovoltaic solar module connected to a
DC-DC Boost converter and the fuzzy logic controller for controlling on/off time of MOSFET switch of a boost converter.
The proposed MPPT controller for grid-connected photovoltaic system is tested using model designed by Matlab/Simulink
program. Comparison of different performance parameters such as: tracking efficiency and response time of the system
shows that the proposed method gives higher efficiency and better performance than the conventional perturbation and
observation method.
Keywords: Photovoltaic (PV) systems, MPPT, Fuzzy Logic Controller, Boost Converter, Perturb and Observe (P&O).
1. Introduction
Energy has the great importance for our life and
economy. The energy demand has greatly increased
due to the industrial revolution. Fossil fuels have been
started to be gradually depleted. The sustainability of
our civilization is seriously threatened. On the other
hand the greenhouse gas emissions are still increasing
due to the conventional generation of energy. It is a
really global challenge to reduce carbon dioxide
emissions and ensuring secure, clean and affordable
energy, and to achieve more sustainable energy
systems [1]. The photovoltaic (PV) energy as an
alternative Energy source has been widely used
because it is pollution Free, abundant, and broadly
available as a consequence the PV systems are
becoming more and more a relevant part Both in
power generation/distribution systems and in

Industrial and domestic plants [2].
There are many sources of renewable energy
such as solar energy, wind energy, etc. Photovoltaic
(PV) system has taken a great attention by the
researchers where it appears to be one of the most
promising renewable energy sources [1-3].
However, two important factors limit the
implementation of photovoltaic systems. These are
high installation cost and low efficiency of energy
conversion [3]. In order to reduce photovoltaic power
system costs and to increases the utilization efficiency
of solar energy, the maximum power point tracking
system of photovoltaic modules is one of the effective
methods [4]. Maximum power point tracking,
frequently referred to as MPPT, is a system used to
extract the maximum power of the PV module to
deliver it to the load, and the efficiency is increased
238
[5]. An important consideration in the use of PV

systems is to operate the system near maximum power
point (MPP) so as to increase the output efficiency of
PV. For any PV system, the output power can be
increased by tracking the MPP of the system by using
a controller connected to a boost converter between
the PV panel and load.
Different techniques have been developed to
maximize the output power of the photovoltaic
modules. Incremental Conductance Method is one of
these methods.
The other method is the constant voltage
tracking (CVT) method. This method compares the
measured voltage (current) of the PV module with a
reference voltage (current) to continuously alter the
duty cycle of the DC-DC converter and hence operate
the PV module at the predetermined point close to the
MPP [7]. Although the CVT method is very simple,
however, the constant voltage cannot track the
maximum power point under the temperature
changing. Perturbation and observation (P&O)
method is an alternative method to obtain the
maximum power point of the PV module. It measures
the voltage, current and power of the PV module.
Then it perturbs the voltage to encounter the change
direction. However, this method suffers of slow of
tracking speed and high oscillations around MPP
[1-5-6-7]. Fig.2 shows the flow chart of the P&O
MPPT algorithm.
This paper presents a new method based fuzzy
logic Controller to achieve maximum power point
tracking. The proposed method depends on measuring
the change in the PV voltage. The performance of the
FLC method is evaluated by Matlab/Simulink.
2. Method
2.1. Characteristics of Solar Module
In order to model a PV module, a PV cell model
must be initially accomplished. An electrical
equivalent circuit makes it possible to model the
characteristic of a PV cell. In a practical PV cell, there
are two resistances: series resistance and parallel
resistance. Series resistance is associated with the
losses in the current path due to the metal grid,
contacts, and current collecting bus. Parallel resistance
due to the loss path in parallel with the built-in device
and due to the p-n junction is not ideal [7]. Since
parallel resistance is larger than series resistance, this
paper neglects the parallel resistance. The equivalent
circuit of the PV cell is shown in Fig. 1.
The characteristic power curve for a PV array is
a non-linear function and, in order to obtain the
maximum efficiency, the array must operate at the
point where the product V*I is maximum (MPP). The
ISSN: 2251-9246
EISSN: 2345-6221
MPP depends on the temperature and insulation level

(G) therefore, in order to preserve the maximum
efficiency, the operating point of the PV array must be
repeatedly updated (maximum power point tracking,
MPPT).
To produce enough high power, the cells must
be connected in N series-parallel configuration on a
module. A PV array is a group of several PV modules
which are electrically connected in series and parallel
circuits to generate the required current and voltage.
The output current delivered to the load can be
expressed as
qv
(1)
I I
I
exp(
) 1
pv,cell
0 ,cell
akT
Where:
I is the output current of the solar module (A), v is
the output voltage of the solar cell (V), which can be
obtained by dividing the output voltage of the PV
module by the number of cells in series, I pv,cell is the
current source of the PV module by solar
irradiance(A), I o ,cell is the reverse saturation current
of a diode(A),
(n
=1~2),
( 1.6 e
is the ideality factor of the diode

is the electric charge of electron
19
is
c ), k
23
( 1.38 10
j/k ,
))
the
boltzmann's
constant
is the given temperature
( K )[9].
In order to model the PV module, the current
generated by the incident light should be first
calculated. This current is given as [9]:
I pv I scn K I (Ta Tn
I0
G
Gn
(3)
I SC, n K I T
Voc, n K v T
exp
aVt
(2)
Where I scn is the short circuit current at normal
conditions (25 C, 1000 w/m2), I pv is the short circuit

current at a given temperature of the cell ( Ta ), K I is
the temperature coefficient of I sc , Gn is the nominal
value of irradiance, which is normally 1000 w/m2. G
is irradiation on the surface of the device, Voc,n is the
open circuit voltage at normal conditions.
239
ISSN: 2251-9246
EISSN: 2345-6221
Fig1. Equivalent circuit of PV cell.
Fig4. I-V curves under changing the solar radiation.
2.2. DC-DC Boost Converter
Fig 2. Flow chart for P&O algorithm [10].
Equations are modeled using Matlab/Simulink in

order to set up the PV model module as well as to
simulate the P-V and I-V curves. Fig. 3 shows the P-V
curves of the PV module under changing solar
radiation from 200W/m2 to 1000W/m2 while keeping
the temperature constant at 25C. Fig. 4 shows the I-V
curves of the PV module.
The DC-DC converter is an electronics circuit

which is used to provide a loss less transfer of energy
between different circuits at different DC voltage
levels. There are many DC-DC converters. One of the
popular types of DC-DC converters is Boost
converter. A boost converter (step-up converter) is a
dc dc power converter with an output voltage greater
than its input voltage. Power for the boost converter
can come from any suitable DC sources, such as
batteries, solar panels, rectifiers and DC generators.
Since power must be conserved, the output current is
lower than the source current.
Boost converter is circuit that operating using
switching mode power supply. Boost converter used
to step up the DC voltage by changing the duty ratio
of the switch. When the duty ration is less than 0.3,
the output voltage is less than the input voltage and
vice versa. The Boost converter circuit is shown in
Fig. 5.
Fig. 5: The Boost converter circuit.
The relation between input voltage over the

output vaolate it is as:
Vo
1
Vd
1 D
Fig3. P-V curves under changing the solar radiation.
Where D it is the duty ratio.
(4)
240
2.3. The Proposed MPPT Fuzzy Login Base Method

Fuzzy logic controller has wide range of
applications in renewable energy applications. The use
of fuzzy logic controllers has been increased over the
last decade because of its simplicity, deal with
imprecise inputs, doesn't need an accurate
mathematical model and can handle nonlinearity [10].
FLC can be used as a controller to obtain the
maximum power that the PV modules capable of
producing under changing weather conditions.
The process of FLC can be classified into three
stages,
fuzzification,
rule
evaluation
and
defuzzification. These components and the general
architecture of a FLS are shown in Fig. 6.
ISSN: 2251-9246
EISSN: 2345-6221
In this paper, a new method based FLC is

proposed to achieve tracking the maximum power of
the PV module under changing solar radiation and
input current. The oscillation around MPP is
decreased and the response is faster in compared with
the conventional P&O method. The proposed fuzzy
logic based MPPT controller has two inputs and one
output. Fuzzy sets for each input and output variable
are defined as shown in Fig.7 and Fig. 8. Five fuzzy
subsets Negative Big (NB), Negative Small (NS),
Zero (ZE), Positive Small (PS), and Positive Big (PB)
are chosen for the input variable Pk . Eleven subsets
are used for the input and output variable Dk 1 . The
subsets are NB, Negative Medium (NM), Negative
Medium Medium (NMM), NS, Negative Small Small
(NSS), ZE, Positive Small Small (PSS), PS, Positive
Medium Medium (PMM), Positive Medium (PM),
and PB. Eleven fuzzy subsets were chosen for Dk 1
in order to smooth the control action. As shown in
Fig.7 and Fig. 8, triangular and trapezoidal shapes
have been adopted for the membership functions; the
value of each input and output variable is normalized.
The same membership function is used for the output
value Dk and the input value Dk 1 .
Fig. 6: Block Diagram of Fuzzy Logic Controller
The fuzzification step involves taking a crisp

input, such as the change in the voltage reading, and
combining it with stored membership function to
produce fuzzy inputs. To transform the crisp inputs
into fuzzy inputs, membership function must be first
assigned for each input. Once the membership
functions are assigned, fuzzification take a real time
inputs and compares it with the stored membership
function information to produce fuzzy input values.
The second step of fuzzy logic processing is the
rule evaluation in which the fuzzy processor uses
linguistic rules to determine what control action
should occur in response to a give set of input values.
The result of rule evaluation is a fuzzy output for each
type of consequent action.
The last step in fuzzy logic processing in which
the expected value of an output variable is derived by
isolating a crisp value in the universe of discource of
the output fuzzy sets. In this process, all of the fuzzy
output values effectively modify their respective
output membership function. One of the most
commonly used defuzzification techniques is called
Center of Gravity (COG) or centroid method.
Fuzzy logic controller has been used for tracking
the maximum power of PV systems since it has the
advantages such as it is robust, relatively simple to
design and does not require the knowledge of an exact
model [10-11].
The general way of representing human

knowledge is by forming natural language expression
given by IF antecedent THEN consequent. The fuzzy
rule-base includes 55 control rules. These rules are
implemented by a computer and used for the control
boost converter such that maximum power is achieved
at the output of the solar panel at all operating
conditions. The rule-base is given in the table 1. The
fuzzy inference of the FLC is based on the Mamdanis
method which is associated with the max-min
composition. The defuzzification technique is based
on the centroid method which is used to compute the
crisp output.
NB
Table 1.
Fuzzy controller rule base
NS
ZE
PS
PB
NB
PM
PM
NM
NM
NM
NM
PS
PMM
NM
NMM
NMM
NMM
PS
PS
NMM
NS
NS
NS
PSS
PSS
NS
NSS
NSS
NSS
PSS
PSS
NSS
NSS
ZE
ZE
NB
NS
ZE
PS
PS
PSS
NSS
NSS
PSS
PSS
PSS
PS
NSS
NSS
PM
PSS
PSS
PMM
NMM
NS
PMM
PS
PS
PM
NMM
NMM
PM
PMM
PMM
PB
NM
NM
PM
PM
PM
241
ISSN: 2251-9246
EISSN: 2345-6221
The Boost converter is designed under the

maximum power 84.67W, under the maximum power
point voltage 19.26V. The load is set to be 100.
Table 3 shows the parameters of the DC-DC boost
converter.
Fig.7: The Membership function of the input variable Pk .
Table 3.
Boost Converter Parameters.
200( H )
Inductance
Initial inductor current
4 (A)
10( F )
Capacitance
Initial capacitor voltage
Switching frequency
Fig.8: The Membership function for change in duty cycle.
Results
In order to verify the MPP tracker for the
photovoltaic simulation system, the proposed MPPT
method is compared with conventional controller
applied on the same system. This controller is
perturbation and observation controller. The proposed
method is by implementing a maximum power point
tracker controlled by fuzzy logic controller and using
boost DC-to-DC converter to keep the PV output
power at the maximum point all the time. This
controller was tested using Matlab/Simulink program,
and the results was compared with a perturbation and
observation controller applied on the same system.
The comparison shows that the fuzzy logic controller
was better in response and dont depend on knowing
any parameter of PV panel.
The model used for simulation is shown in
Fig. 10. The output of the MPPT control block is the
Duty cycle signal which is used to drive the
MOSFET. The principle of this controller is done by
changing the PWM duty cycle (D) and observing the
effect on the output PV power, this can be detailed as
follows:
-when dp/dv > 0, the voltage is increased, this is
done through D (k) = D (k 1) + C.
(C : incrimination step),
-when dp/dv < 0, the voltage is decreased
through D (k) = D (k 1) C.
Electrical characteristics of the modeled PV are given
in theTable 2.
Table 2.
Electrical Characteristics of PV Cell.
Maximum Power (Pmax) 85.14(W)
Short-circuit current
5.45(A)
Open-circuit voltage
22.2(V)
Current at Pmax
4.95(A)
Voltage at Pmax
17.2(V)
200 (V)
100(KHZ)
Maximum power point tracker (MPPT) tracks

the new modified maximum power point in its
corresponding curve whenever temperature and/or
insolation variation occurs. MPPT is used for
extracting the maximum power from the solar PV
module and transferring that power to the load. A
dc/dc (step up) converter acts as an interface between
the load and the module. The MPPT changing the
duty cycle to keep the transfer power from the solar
PV module to the load at maximum point [13].
The membership functions for Pk and Dk 1
are made denser at the center in order to provide more
sensitivity as the variation of power approaches zero.
The duty cycle is internally limited to a maximum
value of 90% to prevent operation at low efficiencies.
It is also limited to a minimum value of 10% to ensure
that the converter switching process does not stop as
operation at D = 0 will indicate a false maximum
power point. In the case when the maximum power is
used; there is no need to enter the duty cycle as an
input. Since, the duty cycle is changed by the
algorithm till the mentioned parameter reaches to the
appropriate value. On the other hand, if the maximum
power is not employed, the user should enter the
constant duty cycle as an input to the system.
When the duty cycle is in ON state, the diode
become as reversed biased and the inductor will
deliver current and switch conducts inductor current.
The current through the inductor increase, as the
source voltage would be greater. The energy stored in
inductor increased when the current increase, and the
inductor acquires energy. Capacitor will provides
smooth out of inductor current changes into a stable
voltage at output voltage. When the duty cycle is in
OFF state, the diode is ON and the inductor will
maintains current to load. Because of inductive energy
storage, inductor current will continues to flow. While
inductor releases current storage, it will flow to the
load and provides voltage to the circuit. The diode is
forward biased. The current flow through the diode
242
which is inductor voltage is equal with negative

output voltage.
In this paper, the fuzzy logic control
demonstrates good performance. Furthermore, fuzzy
logic offers the advantage of faster design, and
emulation of human control strategies. Also fuzzy
control worked well for nonlinear system and shown
higher efficiency over the covenantal controllers.
Comparing the tracking using FLC with that
obtained using P&O MPPT method, the P&O tracked
the maximum power but the oscillation around
maximum power is larger in compared with that
obtained in the FLC method.
Fig.10: Fuzzy MPPT Simulink Simulation Model.
mppt t k
Pachieved Dk , t k
Pmpp t k
(5)
Where, Pachieved Dk , t k is the power achieved
by the algorithm at duty cycle Dk , and Pmpp t k is

the actual maximum power possible. The average of
mppt t k overall t gives a good measure of the
effectiveness of the algorithm. The efficiency of
tracking using FLC MPPT method is over 98%. This
means that over 98% of the available power was
captured by the tracking algorithm.
The number of modules of array photovoltaic is
changed and the simulation is repeated. The results are
represented in Table 4. It should be mentioned that
entrance voltage of the exchanger is a function of
number of modules. For instance, the entrance
voltages are 19.26, 38.52, 115.56 and 231.12 (V) for
one, two, six and twelve modules, respectively. The
other parameters are changed as given in Table 4.
ISSN: 2251-9246
EISSN: 2345-6221
Table 4.
Input and output values of the fuzzy algorithm with changing the
number of photovoltaic modules.
contoller
Vin(V)
Vout(V)
Pin(W)
Pout(W)
iL(A)
Fuzzy
19.26
56.55
21.88
31.98
0.77
Fuzzy
38.52
88.43
59.07
78.2
1.05
Fuzzy
115.56
231.6
366.8
536.3
3.17
Fuzzy
231.12
463.2
1467
2145
6.34
When there are changes in solar radiation,

maximum power point changes. As a result, the Duty
Cycle required for the operation of the model is
changed. If a constant Duty Cycle is applied to, the
maximum power point cannot be tracked, so the
system becomes less efficient. Module voltage,
inductor current, switch control, and input and output
power is shown in Fig. 11. By reducing the time Fig.
12 and Fig 13 are obtained as the previous figure, but
in less time. According to Fig.12, the voltage
waveform module, the inductor current and output,
power input to the system is triangular. Control
switch is a step waveform.
From the simulation show that voltage input for
both controller is almost the same. Perturb and
Observe Controller shows a not stable condition.
From the simulation result is shows that controller that
connected with Boost converter which will give a
stable output is the FLC. Fuzzy Logic controller can
achieve maximum output value at 88.43(V) that better
than Perturb and Observe controller.
Table 5 shows a comparison between the two
controllers perturbation and observation and FLC
controller on the same PV power. The response of
FLC is better than the response of the perturbation and
observation controller since it take more settling time.
Other drawback point in perturbation and observation
controller is that it depends on knowing the value of
the voltage at the maximum power point (Vm).
Table 5.
Comparison Input and Output Value Between FLC MPPT and
Perturb & Observe in Boost Converter.
Controller Vin (V) Iout(A) Pout(W) Pin(W) Vout(V)
Fuzzy
38.52
0.96
84.67
59.07
88.43
P&O
38.79
1.9
72.18
73.7
37.99
243
ISSN: 2251-9246
EISSN: 2345-6221
Fig.11. Module voltage, inductor current, switch control, and input

and output power, the settling time of 0.08.
Fig.13. Module voltage, inductor current, switch control, and input
and output power, the settling time of 4 10 3 .
Furthermore, fuzzy logic offers the advantage of

faster design, and emulation of human control
strategies. Also fuzzy control worked well for
nonlinear system and shown higher efficiency over the
covenantal controllers. Other drawback point in
perturbation and observation controller is that it
depends on knowing the value of the voltage at the
maximum power point (Vm).
References
[1] Jancarle L. Dos Santos, Fernando L. M. Antunes and Anis
Chehab, A Maximum Power Point Tracker for PV Systems
Using a High Performance Boost Converter, Solar Energy,
Vol.80, Iss.7, pp.772- 778, 2005.
Fig 12. Module voltage, inductor current, switch control, and input
and output power, the settling time of 3 103 .
3. Conclusion
Photovoltaic model using Matlab/Simulink and
design of appropriate DC-DC Boost converter with a
maximum power point tracking facility are presented
in this paper. A new method for MPPT based fuzzy
logic controller is presented and compared with the
conventional P&O MPPT method.
The oscillation around MPP is decreased in
compared with the conventional methods. Comparing
the tracking efficiency of both methods indicates that
the proposed method has a higher efficiency than the
conventional P&O MPPT method. In this paper, the
fuzzy logic control demonstrates good performance.
[2] S. J. Chiang, K.T. Chang, C. Y. Yen, Residential Photovoltaic

Energy Storage System, IEEE Trans. On Ind. Electronics,
Vol.45, No.3, June 1998, pp.385-394 and M. Oliver, T.
Jackson, The market for solar photovoltaics, Energy Policy
Vol.27, pp.371-380, 1999.
[3] Ting-Chung Yu and Tang-Shiuan Chien, "Analysis and
Simulation of Characteristics and Maximum Power Point
Tracking for Photovoltaic Systems", Proceeding of power
Electronic and drive systems Conference, pp.1339-1344,
Taipei, 2009.
[4] Roberto Faranda, Sonia Leva, Energy Comparison of MPPT
techniques for PV Systems, Wseas Transctions on Power
System, Vol. 3, Iss.6, pp.446-455, June 2008.
[5] D. P. Hohm and M. E. Ropp, Comparative Study of Maximum
Power Point Tracking Algorithms using an experimental,
programmable, maximum power point tracking test bed,
Proceedings of Photovoltaic Specialists Conference, pp.16991702, USA,2000.
[6] Trishan Esram and Patrick L. Chapman, Comparison of
Photovoltaic Array Maximum Power Point Tracking
Techniques, Energy Conversion, Vol.22, Iss.2, pp.439-449,
May 2007.
244
[7] V. Salas, E. Olias, A. Barrado, A. Lazaro, Review of the

Maximum Power Point Tracking Algorithms for Stand-alone
Photo voltaic Systems, Solar Energy Materials and Solar Cells,
Vol.90, Iss.11, pp.1555-1578, July 2008.
[8] Castafier, Luis & Santiago Silvestre, Modelling Photovoltaic
Systems using PSpice, John Wiley & Sons Ltd, 2002.
[9] Villalva,M. Gazoli,J. and Ruppert ,E. "Modeling and CircuitBased Simulation of Photovoltaic Arrays, IEEE Transactions
on Power Electronics, Vol.24, No.5, May 2009.
[10] I. Laird, H. Lovatt, N. Savvides, Comparative Study of
Maximum
Power
Point
Tracking Algorithms
for
Thermoelectric Generators, Australasian Universities Power
Engineering Conference (AUPEC'08), Paper P-217, 2008.
ISSN: 2251-9246
EISSN: 2345-6221
[11] C. Larbes, S.M. Af't Cheikh, T. Obeidi, A. Zerguerras, "
Genetic algorithms optimized fuzzy logic control for the
maximum power point tracking in photovoltaic system",
Renewable Energy, Vol. 34, Iss.19, pp.2093-2100, 2009.
[12] Tavares, Leite, Suemitsu, Bellar, Performance Evaluation of
Photovoltaic solar system with different MPPT methods",
Proceedings of Industrial Electronics Conference, pp.719-724,
2009.
[13] At,M. Cheikh,S. C.Larbes, (Septembre 2007) Maximum power
point tracking using a fuzzy logic control scheme.

International Journal of Smart Electrical Engineering-8

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

International Journal of Smart Electrical Engineering-8

Hochgeladen von

Copyright:

Verfügbare Formate

187

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

Unit Commitment in Presence of Wind Power Plants and

Number of thermal units

Commitment status of unit i at time t

HPS generating mode decision variable

Startup cost of unit i at time t

Ramp up rate of unit i

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

Generation of unit i at time t

Absorbed power by HPS i at time t

Water reserve inventory at time t

the actual wind power would be different from its

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

high levels of wind power is examined and showed

2.1. Scenario Generation

Where the feasible set X Rn is typically

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

cost, start up and shutdown costs of thermal power

As a case study we have considered a single

Fig.1. Studied system

Pi ,t Pi ,(t 1) (2 Ii ,(t 1) Ii ,t ) Pi ,min (1 Ii ,(t 1) Ii ,t ) RUi

Pi ,(t 1) Pi ,t (2 Ii ,(t 1) Ii ,t ) Pi ,min (1 Ii ,(t 1) Ii ,t ) RDi

4.1. Applying on a Sample System

min Fc ,i ( Pi ,t )* I i ,t SU i ,t SDi ,t SGCh,t SPCh,t

Vh,min Vh,t Vh,max

IGH h,t IPH h,t IIH h,t 1

Each of the wind farms and required demand

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

GAMS24.1.3 and its CPLEX solver to solve the

Fig.2. Forecasted power and scenarios for wind1

Fig.3. Forecasted power and scenarios for wind2

Fig.4. Forecasted demand and scenarios

By means of the introduced system we solve the

In order to ensure system reliability and

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

transition peak load has occurred and thus prevents

It is sensible in fig.7 that the production cost at

Fig.5. Uphill reserve inventory changes

Fig.8. Cost change by varying initial water level in uphill reservoir

4.2. Sensitivity Analysis

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

pump storage unit for the system and power

[13] Venkata Swaroop Pappala, Istvan Erlich, Kurt Rohrig, Jan

[7] J. Liu, M. Salama, and R. Mansour, Identify the Impact of

[19] Amir Kalantari, Jose F. Restrepo, FranciscoD. Galiana,

[8] N. D. Sadanandan, D. W. Hilson, Impact Assessment of

[20] Chaoyue Zhao, Jianhui Wang, Jean-Paul Watson, Yongpei

[21] J. Dupacov, N. Grwe-Kuska, and W. Rmisch, Scenario

[10] Bart C. Ummels , Madeleine Gibescu, Engbert Pelgrum, Wil

[22] N. Grwe-Kuska, H. Heitsch, and W. Rmisch, Scenario

[11] Chun-Lung Chen, Optimal WindThermal Generating Unit

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

Maximum Power Point Tracking of Wind Energy Conversion

impossible [1]. The main disadvantage of the

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013

control. The many approaches have been proposed to

system. Section 3 introduces the Cuckoo

Fig.1. Wind energy conversion system

2.1. Wind Energy Conversion System Characteristics

Where P is the output power (in watt), is the

Where R is radius of blades (in m).

Where is the tip speed ratio (TSR), is the pitch

International Journal of Smart Electrical Engineering, Vol.2, No.4, Fall 2013