00 positive Bewertungen00 negative Bewertungen

0 Ansichten6 SeitenIEEE paper

Jan 23, 2020

© © All Rights Reserved

Sie sind auf Seite 1von 6

loads and seasons

Electrical Engineering Department Electrical Engineering Department

Indian Institute of Technology (ISM) Indian Institute of Technology (ISM)

Dhanbad, Jharkhand, India Dhanbad, Jharkhand, India

sonu.aec2007@gmail.com biplabrec@yahoo.com

Abstract—Optimal sizing of distributed generation sources for a microgrids. Some of the author uses artificial intelligence (AI)

microgrid is very essential for proper functioning of the microgrid techniques such as Genetic Algorithm (GA) and Particle Swarm

when minimization of the energy cost and pollutants emission are Optimization (PSO) while other uses the rule-based method and

of prime concern. This paper deals with three different cases of optimal global methods available. [3] uses iterative and AI

cost and emission minimization each for two microgrid test based methods for optimizing a hybrid PV system. Similarly in

systems consisting of fuel cell, micro-turbine, storage devices and [4] and [5] GA finds the optimal configuration of the hybrid

renewable energy sources. A proposed symbiotic organism search system as well as optimizes the operation strategy using each of

algorithm is used as the optimization tool to minimize the the optimal configuration. [6] uses PSO to minimize the

microgrid operating cost and emission abiding by the various

microgrid cost comprising of a micro-turbine fuel cell, PV,

equality and inequality constraints and considering load

wind turbine and battery storage. The forecasted value of PV

uncertainties and market bids. A comparative study is then made

to prove the efficiency of the proposed algorithm with various and wind turbine and also the real time market prices were

other techniques used in literature. considered while minimizing the microgrid cost in [7] and both

emission and microgrid cost in [8]. Likewise, [9] used hybrid

Keywords—Microgrid, dynamic load dispatch, PV, wind, symbiotic fireworks algorithm to minimize both microgrid costs and

organisms search emission consisting of RES effected by real time market price.

Organisms Search algorithm to minimize the operating cost and

Electric power grid is by far the largest and the most emissions of two test microgrids consisting of PV, Wind

complex innovation of mankind. The upliftment in the quality Turbine, Fuel cell and micro turbine. Section 2 of this paper

of human life and advancement of industrialization is especially discusses the formation of objective functions of the microgrid

because of the modern power system. High voltage alternating including the constraints involved. Section 3 gives a detailed

current (HVAC) transmission of power is used widely due to its knowledge about the Symbiotic organisms search algorithm

efficiency and reliability. Distributed Energy Resources used in the paper. Furthermore section 4 discusses the results

(DERs) like wind and solar are used to generate electric power obtained and a comparative study among the results of other

by integrating them into the utility distribution system [1]. They literatures is made. The paper concludes in section 5.

have many advantages such as energy efficiency, power quality

improvement, reliability and the utilization of transmission and II. ENERGY MANAGEMENT PROBLEM

distribution assets etc.[2]. The microgrid has been developed to FORMULATION

provide bidirectional and controlled power flow to the active

The microgrid considered in this work is a LV grid-

distribution networks.

connected microgrid. It consists of a micro-turbine, a wind

Microgrid comprises of a low voltage system along with turbine, a proton exchange membrane fuel cell and a PV system.

DERs, storage devices and flexible loads. The DERs such as The overall operating cost of a microgrid comprises of the sum

micro-turbines, fuel cells, renewable energy sources (RES) like of DG fuel costs only. Hence the cost objective function

wind turbines and photo voltaic (PV) system along with storage emphasizes on optimal sizing of power from DG sources. Such

devices such as flywheel, battery, energy capacitor etc. all are objective function can be defined as follows:

used in a microgrid. A microgrid has two modes of operation 24 n

Min F(x) = a P(t )

t =1 k =1

k

2

k + bk P(t ) k + ck (1.1)

to both the grid and customer. The primary microgrid control, 24 n

also known as the coordinated control, is used to optimize the Min E ( x) = Pk / Grid (t )ek / Grid (t ) (1.2)

allocation of power among DER, cost of producing the energy t =1 k =1

and emission. There are several methods for optimal sizing of where ek / Grid (t ) = CO2k /Grid (t ) + SO2k /Grid (t ) + NOxk /Grid (t )

where k = 1,2,3,….n and t = 1,2,3,…24 are the total number Case 2: Here the renewable sources (WT and PV) delivers

of DG sources and hours of operation respectively. P(t) is the their maximum available power outputs throughout the day

power output of the DG source at tth hour and ak, bk, ck are the whereas the rest of the DG sources including the MT, FC,

fuel cost coefficients of the kth DG source respectively. Eq. (1.2) battery and even the utility operates within their bounded limits.

is the emission objective where ek / Grid (t ) is the emission The significance of this case is to utilize maximum of the

available renewable energy sources and reduce the dependency

coefficient of the DG sources and grid. on the distribution grid (utility) thus helping in selling the

The above mentioned objective function in (1) is bound to excess power to the grid and maintain an economic balance.

some equality and inequality constraints as explained below: Case 3: The utility in this case is made unconstrained. In

other words all the DG sources shall be operating in their

A. Generation Limit maximum and minimum limits whereas the utility is allowed to

The hourly power outputs of all the DG sources including exchange any amount of power with the microgrid. This case is

the storage device and utility should lie between its maximum also known as “Infinite Energy Exchange”.

and minimum limits. The proposed SOS-based approach is implemented to find

the optimal sizing of the DG sources in all the three different

P k / Grid ,min (t ) ≤ Pk / Grid (t ) ≤ Pk / Grid ,max (2) cases.

III. SYMBIOTIC ORGANISMS SEARCH

Where Pk,min(t) are the minimum output powers and Pk,max(t) Symbiotic organisms search (SOS) is a relatively new,

are the maximum output powers of the DG sources at the tth powerful and meta-heuristic algorithm applied to optimize

hour respectively. many mathematical and engineering problems [10]. It works by

simulating the symbiotic strategies acquired by the organisms

B. Storage device limits

among themselves to survive and sustain in the ecosystem.

The charge and discharge rates of storage device of test system Almost all the meta-heuristic algorithms available in the

2 for each time interval are limited as follows literature share some common characteristics such as: they are

inspired from the nature, make use of random variables and they

1 make use of several parameters that need to be adjusted to the

Wess,t = Wess ,t −1 + η ch arg e Pch arg e Δt − Pdisch arg e Δt

η disch arg e problem. The fact that SOS doesn’t require any algorithm

(3)

specific tuning parameters makes it superior to many other

Wess ,min ≤ Wess ,t ≤ Wess ,max meta-heuristic algorithms. Symbiotic relationships provide at

least one of the participating species with a nutritional

Pch arg e ,t ≤ Pch arg e , max advantage. The symbiotic relationships that are found in nature

are of three types, viz. mutualism, commensalism and

Pdisch arg e ,t ≤ Pdisch arg e ,max (4) parasitism. These three relationships are mathematically

speculated and the SOS algorithm is developed as a result.

Where Wess,t and Wess,t-1 are the energy stored in the battery at tth

and (t-1)th hour respectively. Pcharge and Pdischarge are the Mutualism phase: In the mutualism phase of SOS, both the

allowable rate of charge and discharge during a definite period species involved are benefitted. One common example is the

say ∆t. The suffix ‘min’ and ‘max’ denotes the minimum and relationship between honey bees and flowers. This phase can be

maximum of the respective parameter. mathematically developed by the following equations:

C. Power Supply-Demand Balance Equation

Xi + X j

The total generation of power from the DGs in the microgrid Mutual_Vector= (6)

during every hour including the RES must fulfill the total load 2

demand of the grid. Transmission losses are neglected as the X inew = X i + rand(0,1) * ( X best − Mutual_ Vector* BF1 ) (7)

microgrid considered is a small LV one. Therefore X jnew = X j + rand (0,1) * ( X best − Mutual _ Vector * BF2 ) (8)

n where Xi is an organism of the ith member of the ecosystem and

P

k =1

k / Grid (t ) + PPV (t ) + PWT (t ) = Pload (t ) (5)

Xj is randomly selected from the ecosystem to interact with Xi.

rand(0,1) denote a vector of random numbers. BF1 and BF2

Three different cases for two different microgrid test denote the benefit factors and are kept either at 1 or 2.

systems are studied to exploit the availability and the Mutual_Vector represents the mutual relation between the

dependency on the renewable energy sources that are acting as organisms Xi an Xj.

DG sources for the microgrid. For microgrid test system 1,

these cases are nothing but the variation of load demand for Commensalism phase: Commensalism is a relationship

three different seasons viz. winter, summer and spring. The existing in nature between individual of two species where one

variation in the power generation of the RES were also species gather its food or other benefits from the other without

considered for these seasons. For microgrid test system 2, the harming or benefitting the latter. Similar to the mutualism

three cases are described as below: phase, Xj is selected randomly to interact with Xi and a new

Case 1: In this case it is taken into consideration that all of organism Xinew can be calculated as:

the DG sources including the storage device and the utility

operates within their maximum and minimum limits to satisfy X inew = X i + rand (−1,1)*( X best − X j )

(9)

the load demand throughout each hour of the day.

Where (Xbest – Xj) portrays the beneficial advantage Step 5: Parasitism phase: Organism Xj (Xj≠Xi) is randomly

provided by Xj to help Xi increasing its survival advantage in selected from the ecosystem. Parasite_Vector is formed by

ecosystem to the highest degree Xbest in current organism. mutating Xi in random dimensions using a random number

within a given range. Constraint checking is done and fitness

Parasitism phase: Parasitism is the name given to the value is calculated. If Parasite_Vector is found better than the

relationship between two organisms in the ecosystem where previously calculated fitness value, then the previous fitness

one is harmed and the other gets benefitted. The organism that value is replaced with the Parasite_Vector else the

gets benefitted is called ‘parasite’ and the one that faces the Parasite_Vector is rejected and we proceed to the next step.

harm is called the ‘host’.

In SOS, Xj is selected randomly to act as the host. Step 6: We proceed to step 2 if the current Xi is not the last

Parasite_Vector is an artificial organism created in the search member of the ecosystem; otherwise we proceed to the next

space. If fitness value of Parasite_Vector is better than Xj, it step.

will replace organism Xj. And if the fitness value of Xj is better, Step 7: We stop if one of the termination criteria i.e. the

it will have immunity and the Parasite_Vector will no longer maximum number of iterations is reached; otherwise we return

survive in that ecosystem. to step 2 and start the next iteration

The algorithm for optimal power management and sizing of IV. RESULTS AND DISCUSSION

DG sources with SOS are summarized as follows:- Description of the system: A strong and powerful Symbiotic

Step 1: Formation of Ecosystem – The parameters organisms search is used to evaluate its performance in optimal

considered for microgrid energy management include fuel cost- sizing of a microgrid and minimize its operating cost. The

coefficients of generators, power generation limits, power microgrid considered in this work is a LV islanded microgrid.

demand of various types of loads and limits of forecasted wind The microgrid test system 1 consists of three micro-turbines, a

power. Also the size of ecosystem i.e. the total number of wind turbine, fifteen proton exchange membrane fuel cells and

organisms in the ecosystem (eco_size) and maximum iteration a PV system. The time span for calculating the optimal sizing

(max_iter) is set in this step. is considered to be 24 hours. The complete system data which

includes the maximum and minimum capacity of the DGs, their

Step 2: The particles of the population is initialized in a fuel cost coefficients, load demand profile for 24 hours for

random manner according to the limits of each unit including various seasons and the RES output for those seasons are

individual dimensions. These initial particles must be feasible

gathered from [12]. MATLAB R2013a platform is used to code

candidate solutions that satisfy the practical operating

constraints. Let Vi be the trial vector designating the ith particle and execute the algorithm in a personal computer with 2.53GHz

of the initial population where Vi consists the power outputs of core i3 processor and 2GB RAM. The program is run with 30

micro-turbine, fuel cell, photo voltaic array and wind turbine population and 1000 iterations for 20 trials. The weightage

for 24 hours respectively. Hence Pi can be represented as factor (f_weight) is set at 0.7 and crossover probability constant

Vi=[Pk1, Pk2, Pk3….Pk24]; (f_cr) is set at 0.2 for the DE algorithm.

Now for k number of DG sources and m number of particles The microgrid test system 2 is a LV grid connected

i varies from i=1, 2, 3….m. Hence the population matrix can be microgrid consisting of MT, FC, PV, WT and a battery as

represented by equation (8) storage device. The system data which includes the DG

V = [ P1 , P2 , P3 ...Pm ]T (10) parameters, load demand and the real time market price are

taken from [7]. The personal computer configuration and the

Step 3: Mutualism phase: Here i is initially set at 1, organism tuning parameters of DE algorithm were maintained the same

X1 is matched to Xi and organism Xj is formed randomly from as of microgrid test system 1. The program is run with 30

the ecosystem. In this case, X2 is selected as Xj. Mutual_Vector population and 1500 iterations for 20 trials.

is calculated using (4). Benefit Factors (BF1 and BF2) are set Comparative Analysis:

at 2. Organism Xi and Xj are modified based on their mutual For Microgrid Test system 1: The optimal sizing of the DG

relationship using (5) and (6) and the constraints checking is sources is done to give a minimized microgrid cost for all the

done. Once it is found that Xi and Xj abides by the constraints, three seasons using a proposed SOS method. The costs are then

the fitness value is then accounted for, which if found better compared with three other algorithms such as Genetic

than the initial fitness value, we go to next step else we reject Algorithm, Particle Swarm Optimization and Differential

the modified organism and proceed to next step with the initial Evolution with local global neighborhood [11]. Table 1 does the

solution. detailed comparative analysis of the fuel costs obtained using

Step 4: Commensalism phase: Organism Xj (Xj≠Xi) is generated the four algorithms. It also highlights the statistical analysis of

from the ecosystem on random basis. New candidate solutions the algorithms and the computational time taken to obtain the

Xi,new are calculated using (7). Constraint checking is done and optimum value. It can be clearly seen that for all the three

fitness value is calculated. Like the previous step, if fitness seasons SOS yielded the least fuel costs. The fuel costs of the

value of the modified organism in this step is better than the DG sources was $6783.0102 in winter as evaluated by SOS.

previous value then we go to the next step else the modified This price is pretty less compared to $6799.4304 by GA,

organism is rejected and the previous solution is kept and $6790.0648 by PSO and $6787.8858 by DEGL. For summer

proceeded to the next step. too, SOS minimized the fuel cost to as low as $6836.8722

compared to $6878.7290 by GA, $6854.2558 by PSO and

$6841.5104 by SOS. Likewise the fuel costs for the loads and

RES of spring season was minimized by SOS to a low value of

$5924.3871. This value was much better than $5947.4138 by

GA, $5943.3537 by PSO and $5930.0844 by SOS. Figure 1, 2

and 3 shows the convergence characteristics of the proposed

SOS method along with the other algorithms used. The

steepness of the curves in the figures, the lowest values of

standard deviation and the decreasing order of computational

time for 1000 iterations in Table 2 clearly indicate about the fast

convergence criteria and robustness of the proposed algorithm

to attain such a minimal value.

For Microgrid Test system 2: Three optimization techniques

viz. DE, DEGL & SOS were implemented to minimize the

emission objective for microgrid test system 2. The

optimization techniques were run for 1500 iterations each to

match with the literature for comparative analysis. Table 2

gives a detailed statistical and comparative analysis for the Fig 2: Convergence characteristics of minimal micro grid fuel costs in

summer

minimized emission objective along with various algorithms

from literature. It can be clearly seen that for all the three cases

SOS yielded the least emission of pollutants among all the

algorithms used so far. The least value of standard deviation

from table 2 along with simulation time shows the fast

convergence property of SOS compared to DE and DEGL. SOS

gave the minimized result 18 to 19 times out of 20 trials thus

increasing its robustness to 90-95%. Figure 4 through 6 shows

the convergence characteristics of the emission objective for all

the three cases. Figure 7 through 9 are the hourly outputs of the

DG sources for the emission objective using SOS.

Best Solution

No. of hits to

optimum

Time (in

Solution

Average

secs) for

solution

Method

Worst

(in $)

(in $)

(in $)

1000

S.D

Winter

GA 6799.4304 6800.3939 6801.3574 1.36 301 17

PSO 6790.0548 6791.0279 6792.0011 1.37 287 17

DEGL 6787.8858 6788.0025 6788.1193 0.16 254 18

Fig 1: Convergence characteristics of minimal micro grid fuel costs in winter SOS 6783.0102 6783.5054 6784.0006 0.70 251 19

Summer

GA 6878.7290 6880.0482 6881.3674 1.86 300 16

PSO 6854.2558 6854.7347 6855.2137 0.67 289 19

DEGL 6841.5104 6841.7924 6842.0744 0.39 253 18

SOS 6836.8722 6836.9354 6836.9987 0.08 249 20

Spring

GA 5947.4138 5947.7079 5948.0021 0.41 297 18

PSO 5943.3537 5943.6685 5943.9834 0.44 293 17

DEGL 5930.0844 5930.5995 5931.1147 0.72 246 19

SOS 5924.3871 5924.5284 5924.6698 0.19 219 19

Fig 4: Convergence characteristics of minimal micro grid emission (Case 1) Fig 7: Hourly outputs of DGs for emission objective using SOS (Case 1)

Fig 5: Convergence characteristics of minimal micro grid emission (Case 2) Fig 8: Hourly outputs of DGs for emission objective using SOS (Case 2)

Fig 6: Convergence characteristics of minimal micro grid emission (Case 3) Fig 9: Hourly outputs of DGs for emission objective using SOS (Case 3)

TABLE 2: COMPARATIVE ANALYSIS OF MICROGRID EMISSION emissions with varying loads and RES outputs throughout the

seasons. The results obtained are then compared to a few other

secs/1000

optimization algorithms found in the literature and SOS

iteration

Solution

Solution

Average

Method

(in kg.)

(in kg.)

(in kg.)

Worst

Best

S.D

No. of

outperformed them all. Due to its better computational speed

hits

and exploitability SOS can therefore be considered as one of the

Case 1 strongest optimization tool to solve various power system and

GA [7] 435.2363 445.3862 457.4680 14.2299 - - microgrid problem.

PSO [7] 4358227 445.1072 454.5917 13.9708 - -

FSAPSO [7] 435.0830 443.4396 451.3821 11.3525 - - REFERENCES

CPSO-T [7] 434.9973 440.1036 444.9398 6.9950 - - [1] A. S. Safigianni, G. N. Koutroumpezis, and V. C.

CPSO-L [7] 434.9354 439.2369 443.6383 6.1538 - - Poulios, “Mixed distributed generation technologies in

AMPSO-T [7] 434.8611 434.9983 435.1126 0.1786 - - a medium voltage network,” Electr. Power Syst. Res.,

vol. 96, pp. 75–80, 2013.

AMPSO-L [7] 434.8193 434.9235 435.0099 0.0681 - -

[2] D. K. Nichols, J. Stevens, R. H. Lasseter, J. H. Eto,

FA [9] 485.6731 562.2740 605.6592 30.4118 - - and H. T. Vollkommer, “Validation of the CERTS

GSA [9] 513.3899 530.7464 544.5167 8.9494 - - microgrid concept the CEC/CERTS microgrid

FAGSO [9] 415.8571 415.8667 415.8661 0.0020 - - testbed,” 2006 IEEE Power Eng. Soc. Gen. Meet., pp.

DE 376.9918 377.5523 378.1129 0.7927 86 18 1–3, 2006.

DEGL 375.4463 375.7676 376.0889 0.4543 86 18 [3] T. Khatib, A. Mohamed, and K. Sopian, “A review of

SOS 371.8503 371.9235 371.9967 0.1035 78 19 photovoltaic systems size optimization techniques,”

Case 2

Renew. Sustain. Energy Rev., vol. 22, pp. 454–465,

GA [7] 435.1308 441.2402 448.7740 5.2689 - - 2013.

[4] A. K. Daud and M. S. Ismail, “Design of isolated

PSO [7] 435.5555 436.5928 438.2212 1.2666 - -

hybrid systems minimizing costs and pollutant

FSAPSO [7] 435.0037 436.0913 437.1788 1.5380 - -

emissions,” Renew. Energy, vol. 44, pp. 215–224,

CPSO-T [7] 434.9814 435.9408 436.9001 1.3567 - - 2012.

CPSO-L [7] 434.9064 435.6447 436.3830 1.0441 - - [5] J. C. Hernández, A. Medina, and F. Jurado, “Optimal

AMPSO-T [7] 434.8611 434.9357 435.0102 0.1054 - - allocation and sizing for profitability and voltage

AMPSO-L [7] 434.8161 434.8920 434.9690 0.0586 - - enhancement of PV systems on feeders,” Renew.

FA [9] 482.9904 511.9287 555.5889 17.9446 - -

Energy, vol. 32, no. 10, pp. 1768–1789, 2007.

[6] J. Radosavljević, M. Jevtić, and D. Klimenta, “Energy

GSA [9] 477.0860 489.7806 504.3866 9.0922 - -

and operation management of a microgrid using

FAGSO [9] 415.8571 415.8571 415.8571 4.7*10-8 - - particle swarm optimization,” Eng. Optim., vol. 273,

DE 375.4475 375.8704 376.2933 0.5980 85 18 no. May, pp. 1–20, 2015.

DEGL 375.2321 375.7686 376.3052 0.7587 86 18 [7] T. Niknam, F. Golestaneh, and A. Malekpour,

SOS 374.6560 374.8280 375.0001 0.2433 80 18 “Probabilistic energy and operation management of a

Case 3 microgrid containing wind/photovoltaic/fuel cell

GA [7] 435.9708 447.3231 458.6008 7.0154 - - generation and energy storage devices based on point

PSO [7] 434.8319 440.9284 448.7398 4.8683 - - estimate method and self-adaptive gravitational search

FSAPSO [7] 434.8287 436.0913 438.2267 2.3211 - - algorithm,” Energy, vol. 43, no. 1, pp. 427–437, 2012.

[8] A. A. Moghaddam, A. Seifi, T. Niknam, and M. R.

CPSO-T [7] 434.8263 435.9408 437.0801 1.5534 - -

Alizadeh Pahlavani, “Multi-objective operation

CPSO-L [7] 434.8204 435.6447 436.9937 1.5309 - - management of a renewable MG (micro-grid) with

AMPSO-T [7] 434.8190 434.9357 435.0100 0.1350 - - back-up micro-turbine/fuel cell/battery hybrid power

AMPSO-L [7] 434.8168 434.9038 434.9998 0.0604 - - source,” Energy, vol. 36, no. 11, pp. 6490–6507, 2011.

FA [9] 554.8896 617.6640 654.0684 28.3728 - - [9] Wang Z, Zhu Q, Huang M, Yang B. “Optimization of

GSA [9] 594.9877 635.3205 672.1839 19.3275 - - economic/environmental operation management for

FAGSO [9] 415.8571 415.8572 415.8574 5.3*10-5 - -

microgrids by using hybrid fireworks algorithm”. Int

Trans Electr Energ Syst. 2017;e2429

DE 398.7736 398.8681 398.9627 0.1337 83 19

[10] Cheng, M. Y., & Prayogo, D. “Symbiotic Organisms

DEGL 398.6660 398.8525 399.0391 0.2638 83 18 Search: A new metaheuristic optimization algorithm”.

SOS 397.6629 397.7649 397.8670 0.1443 81 19 Computers and Structures, vol. 139, pp. 98–112, 2014.

[11] S. Das, A. Abraham, U. K. Chakraborty, and A. Konar,

V. CONCLUSION “Differential evolution using a neighborhood-based

A SOS method was used in this paper for optimal sizing of mutation operator,” IEEE Trans. Evol. Comput., vol.

13, no. 3, pp. 526–553, (2009)

two typical LV microgrids which included as many as twenty [12] Maulik, A. and Das, D., Optimal operation of

numbers of DG sources including storage devices, PV and wind microgrid using four different optimization

turbine. Three cases per test system were studied for the optimal techniques. Sustainable Energy Technologies and

sizing of DG sources so that the microgrid functions in the most Assessments, 21, pp.100-120, 2017

efficient as well as economical way by releasing minimal

## Viel mehr als nur Dokumente.

Entdecken, was Scribd alles zu bieten hat, inklusive Bücher und Hörbücher von großen Verlagen.

Jederzeit kündbar.