Beruflich Dokumente
Kultur Dokumente
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents probabilistic coordination of distributed energy resources (DERs) operation in an
Received 22 January 2015 islanded microgrid with consideration of the associated uncertainties. In doing so, a comprehensive
Received in revised form 27 May 2015 stochastic mathematical model is developed which incorporates a set of valid probabilistic scenarios
Accepted 8 June 2015
for the uncertainties of load and intermittency in wind and solar generation sources. The uncertainty
Available online 3 July 2015
is addressed through a combination of a stochastic optimization model and additional reserve require-
ments. The model also includes hourly interruption costs for a variety of customer types as a means of
Keywords:
determining the optimal probabilistic interruptible load whose reliability-based value is low enough to
Distributed energy resources
Microgrid
enable it to be shed if necessary. A case study is carried out using a benchmark microgrid; numerical
Smart grid results demonstrate that coordinated operation of DERs brings notable benets in terms of expected
Uncertainty operation costs and system security. This probabilistic coordination further reduces the consequences of
the expected power dispatch of controllable generators and hourly unserved power.
Crown Copyright 2015 Published by Elsevier B.V. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
http://dx.doi.org/10.1016/j.epsr.2015.06.010
0378-7796/Crown Copyright 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-
nd/4.0/).
2 W. Alharbi, K. Raahemifar / Electric Power Systems Research 128 (2015) 110
is = di w
i
pi v =1 S =nqm (5) Down.R-C-I
Pk,s Down.R
= Pk,s Down.C
+ Pk,s Down.I
+ Pk,s s S (13)
i=1 i=1
where S is the total number of scenarios. To prevent the customers activities from being shifted to the
The scenarios extracted and their corresponding probabilities next day, the demand variation must be balanced within the 24-h
are then included in the formulation of the microgrid energy man- operating horizon, as in (14). The maximum demand that can be
agement model. shifted from 1 h to another is given by (15) [21]:
UP,R-C-I Down,R-C-I
3. Comprehensive mathematical model of microgrid Pk,s = Pk,s s S (14)
energy management kK kK
Down,R-C-I
R-C-I
Bdp Pk,s
The mathematical model is formulated as a multi-scenario D,R-C-I
Mixed Integer Linear Programming (MILP) model. The models Pk,s s S (15)
R-C-I UP,R-C-I
BUP Pk,s
objective is to minimize the expected total cost (ETC) over a period
4 W. Alharbi, K. Raahemifar / Electric Power Systems Research 128 (2015) 110
1400
1200
Power, kW
1000
800
600
400
200
0
0 3 6 9 12 15 18 21 24
Time, h
3.2. Spinning reserve requirements of a microgrid and must be linearized to preserve the MILP formulations. Thus
Eqs. (24)(26) are added to represent the linearization form of the
This constraint ensures that the microgrid spinning reserve security action to ensure any excess load is interrupted in order to
requirements are met in period k. maintain power balance.
D,RCI
Pjmax Vj,k Pj,k,s + DCH Ck,s Cmin Rk,s + D Pk,s 1 if W
Pk,s + Pk,s
PV
+ Pjmax Vj,k D,R-C-I
SRks < Pk,s
Int,R-C-I
Vk,s = (23)
jJ jJ
0 Otherwise
W
+ W Pk,s PV
+ PV Pk,s s S (16)
W
Pk,s PV
+ Pk,s + Pjmax Vj,k D,R-C-I
Pk,s Int,R-C-I
SRks (1 Vk,s ) s S (24)
where D , W , and PV are the forecasting error factors for load,
jJ
wind, and solar, respectively.
The rst term of Eq. (16) presents the net available capacity W
Pk,s PV
Pk,s Pjmax Vj,k D,R-C-I
+ Pk,s Int,R-C-I
+ SRks Vk,s s S (25)
of all committed controllable DGs, and the second term denotes
jJ
the available capacity of the ESS. The last four terms of the equa-
Int,R-C-I int,R-C-I
tion represent the spinning reserve requirements. In fact, the last Pk,s Vk,s s S (26)
three terms are the extra reserves added to mitigate the effects of
uncertainties on the load, wind, and solar power forecasts. 4. Model scenarios, results, and analysis
Each controllable generator is also subject to its own operating
constraints, such as ramp-up/ramp-down, minimum up and down 4.1. System under study
times, generation output limits, and logical and initial constraints
[20]. The model was validated using a benchmark microgrid consist-
ing of 12 controllable DGs, a wind turbine (WT), a PV system, and an
3.3. Energy storage systems ESS. The total installed capacity in the microgrid system is 4 MW,
with a renewable penetration level of 49%. The installed capacities
The following equations represent the operational constraints of controllable DGs, WT, and PV systems are 2.04 MW, 0.52 MW
of the ESS [15]: (4 0.14 MWp), and 1.44 MW (4 0.14 0.36 MWp), respectively
Power discharge and charge limits: [22]. The 24-h load prole for the microgrid, which includes the
DCH
Pk,s max
PES DCH
Vk,s sS (17) commercial, residential, and industrial loads derived from [23], is
shown in Fig. 1. Fig. 2 shows the hourly PV output power prole.
CH
Pk,s max
PES CH
Vk,s s S (18) The power output formulation of wind generation can be expressed
as follows [22,24]:
Discharging and charging dynamics:
DCH
Ck+1,s = Ck,s dk Pk,s CH
/DCH + dk Pk,s CH s S
0 0 Vm V1
(19)
aV 4 + b V 3 + cV 2 + dVm + e V1 Vm Vr
m m m
ESS operational end points and energy storage limits: out
Pw = (27)
Prated Vr Vm Vcut-out
C0,s = CSt , CK,s CE s S (20)
0 Vm Vcut-out
Cmin Ck,s Cmax s S (21)
where Vm , V1 , Vr , and Vcut-out are the wind speed, cut-in speed,
Coordination of charging and discharging operation:
speed at the rated power, and cut-out speed, respectively. Table 1
DCH
Vk,s DCH
+ Vk,s 1 s S (22)
16
14
12
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time, h
140 3.00 15.01 17.00 0.015 0.33 0.9 2.10 7.10 4.2. Results and analysis
2.5
1.5
0.5
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time, h
Fig. 4. Interruption cost hourly factors for the three types of customers during 1 day.
Table 3
Cost components and emissions for islanded microgrids.
Table 4
Unit commitment decisions for an isolated microgrid.
Hour 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Generator
DG1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
DG2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
DG3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1
DG4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1
DG5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 0
DG6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 0
DG7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0
DG8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0
DG9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0
DG10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0
DG11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0
DG12 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 0
changed from on-state (Case-1) to off-state (Case-4) during hours of controllable DGs during the worst scenario, in which the load
116. It can be noticed that probabilistic coordination of DERs oper- prole exhibits the highest positive values compared to the pre-
ations makes UC decisions of controllable DGs smooth and without fect forecasted values, while the wind and solar energy generation
any uctuation, and thus reduces the total expected operation cost proles have the highest negative values relative to their prefect
by reducing the starting up and shutting down costs of controllable forecasted values. It can be observed that coordinating the DERs,
DGs. as in Case-4, further attens the net demand and hence reduces
Fig. 5 illustrates the effects of the variability and uncertainty the impact of variability and uncertainty on the real-time dispatch
associated with wind and solar generation on the power dispatch of controllable DGs. During peak times, in Case-2 and Case-4, the
DGs are dispatched to their fully installed capacities since the ESS
is considered to be providing the required spinning reserve in the
100 system.
Expected power dispatch of controllable
90
DGs with DR DER Furthermore, the power dispatch of controllable DGs for 2 of the
80
125 scenarios, i.e., the best and worst scenarios, are compared with
70
that of the prefect scenario. The worst scenario has been dened
60
above. In the best scenario however, the load prole has the high-
50
est negative values compared to the prefect forecasted values and
40
30
the wind and solar energy generation proles exhibit the high-
20
est positive values relative to their perfect forecasted values. The
10 power dispatch of controllable DGs for the worst and best scenar-
0 ios and the perfect forecast for Case-1 and Case-4 are presented in
0 3 6 9 12 15 18 21 24
Figs. 6 and 7, respectively. Fig. 6 shows the impact of high variabil-
Time, h
ity and uncertainty of renewable power generation on the power
Fig. 5. Optimal dispatch of controllable DGs (worst scenario). dispatch of controllable DGs, mainly during the best and worst
W. Alharbi, K. Raahemifar / Electric Power Systems Research 128 (2015) 110 7
controllable DGs, kW
25
0
-25 0 3 6 9 12 15 18 21 24
-50
-75
-100
Time, h
180
150
120
Controllable DGs, kW
90
60
30
0
-30 0 3 6 9 12 15 18 21 24
-60
-90
-120
-150
-180
Time, h
scenarios. In the best scenario, the power dispatch of controllable over a period of 24 h. Fig. 9 shows that the ESS has no signicant
DGs reduces with respect to the prefect forecast. In the worst sce- impact on DR since the DR is derived primarily from its respon-
nario however, the power dispatch of controllable DGs increases. sive price, and hence occurs when the price is low. However, the
In contrast probabilistic coordination of DERs either shrinks the opposite is not true because the DR can play a signicant role in the
gap between the worst/best scenario and the prefect forecast in amount of discrepancy between the expected power generation
order to reduce the renewable variability and uncertainty effects and load and thus it will have a signicant effect on ESS charging
on power dispatch of controllable DGs, or widens the gap during and discharging behaviors.
off-peak periods to charge the ESS and/or induce optimal DR to Since the hourly interruption cost is not as high for residen-
use during on-peak hours (Fig. 7), and therefore further reduces tial customers as for commercial or industrial customers, when an
the total expected operation costs. It is worth noting that the gap excess demand occurs the residential loads are interrupted rst to
between the best/worst scenario and prefect scenario is created ensure system security. The expected hourly interruptible loads for
from forecasting errors of load, wind and solar energy generation residential customers during the worst-case scenario are shown
proles. in Fig. 10. The optimal probabilistic load interruption is clearly
Fig. 8 shows ESS charging/discharging behaviors and state of induced, mainly during peak times, and varies noticeably from case
charge of the ESS during the worst scenario, both alone and in coor- to case, depending on the condition and exibility of the system. It
dination with the DR. Similarly, DR behaviors with and without ESS is worth noting that, to ensure system security in Case-1, not only
are illustrated in Fig. 9. Fig. 8a shows the charging and discharging the residential load, but also the commercial load is interrupted
cycles of ESS which follows the load prole when DR is not consid- (Fig. 11).
ered. This means charging ESS during off-peak and discharging it A sensitivity analysis is carried out to assess the impact of higher
during on-peak periods. However, this behavior is changed when forecasting errors of renewable energy generation and load on
the ESS is in coordination with DR, which in this case follows the microgrid operations and system security. The higher uncertainties
original load plus the deferrable loads. Fig. 8b represents the state are obtained by assigning a large standard deviation to the probabil-
of charge of the ESS according to its charging/discharging behaviors ity density functions of the forecasting errors. The new distribution
Table 5
Cost components for islanded microgrids considering higher forecasting errors.
Charging/Dsichrging of the
75
50
25
ESS,,%
0
0 3 6 9 12 15 18 21 24
-25
b
100
SOC of the ESS, %
75
50
Fig. 8. (a) ESS charging/discharging behavior alone and in coordination with the DR during the worst scenario. (b) State of charge of the ESS alone and in coordination with
the DR during the worst scenario.
80
DR alone DR in Coordination with ESS
60
Expected responsive loads, kW
40
20
0
0 3 6 9 12 15 18 21 24
-20
-40
Time, h
-60
-80
Fig. 9. DR behavior with and without the ESS during the worst-case scenario.
functions are generated using Table 2, by multiplying the forecast- expected operation cost of the coordination case (Case-4) is still less
ing errors by 1.75 and maintaining the same probability. Table 5 than those of Case-2 and Case-3 (Table 3) where higher forecasting
presents the expected simulation results for the base case and coor- errors were not considered. This shows probabilistic coordination
dination case of the isolated microgrid. It can be observed that the of DERs operation reduces the risk in scheduling and increases the
450
Expected Interruptible Loads, kW
400
350
300
250
200
150
100
50
0
0 3 6 9 12 15 18 21 24
Time, h
Case-1 Case-2 Case-3 Case-4
Fig. 10. Optimal interruptible residential loads for each case during the worst-case scenario.
W. Alharbi, K. Raahemifar / Electric Power Systems Research 128 (2015) 110 9
150
90
60
30
0
0 3 6 9 12 15 18 21 24
Time, h
Fig. 11. Optimal interruptible commercial loads for Case-1 during the worst-case scenario.
[14] M. Rahimiyan, L. Baringo, A.J. Conejo, Energy management of a cluster of inter- [21] K. Dietrich, J.M. Latorre, L. Olmos, A. Ramos, Demand response in an isolated
connected price-responsive demands, IEEE Trans. Power Syst. 29 (March (2)) system with high wind integration, IEEE Trans. Power Syst. 27 (2012) 2029.
(2014) 645655. [22] T. Logenthiran, D. Srinivasan, A.M. Khambadkone, Multi-agent system for
[15] S. Chen, H.B. Gooi, M. Wang, Sizing of energy storage for microgrids, IEEE Trans. energy resource scheduling of integrated microgrids in a distributed system,
Smart Grid 3 (2012) 142151. Electr. Power Syst. Res. 81 (2011) 138148.
[16] A.Y. Saber, G.K. Venayagamoorthy, Resource scheduling under uncertainty in a [23] A. Tsikalakis, N. Hatziargyriou, Centralized control for optimizing microgrids
smart grid with renewables and plug-in vehicles, IEEE Syst. J. 6 (2012) 103109. operation, IEEE Trans. Energy Convers. 23 (2008) 241248.
[17] P.A. Ruiz, C.R. Philbrick, E. Zak, K.W. Cheung, P.W. Sauer, Uncertainty man- [24] T. Logenthiran, D. Srinivasan, Short term generation scheduling of a microgrid,
agement in the unit commitment problem, IEEE Trans. Power Syst. 24 (2009) in: IEEE TENCON, 2009, pp. 16.
642651. [25] K. Divya, J. stergaard, Battery energy storage technology for power
[18] I. Bae, J. Kim, C. Singh, Optimal operating strategy for distributed generation systemsan overview, Electr. Power Syst. Res. 79 (2009) 511520.
considering hourly reliability worth, IEEE Trans. Power Syst. 19 (1) (2004) [26] C. Volume II, Uncontrolled Emission Factor Listing For Criteria Air Pollutants,
287292. July 2001. Prepared by: Eastern Research Group, Point Sources Committee and
[19] S. Junlakarn, M. Ilic, Distribution system reliability options and utility liability, Emission Inventory Improvement Program, located at http://www.epa.gov/
IEEE Trans. Smart Grid 5 (September (5)) (2014) 22272233. ttnchie1/eiip/techreport/volume02/index.html
[20] M. Carrin, J.M. Arroyo, A computationally efcient mixed-integer linear for- [27] GAMS Development Corporation, General algebraic modeling system (GAMS)
mulation for the thermal unit commitment problem, IEEE Trans. Power Syst. Software, 2012 http://www.gams.com
21 (2006) 13711378.