Professional Documents
Culture Documents
Abstract--In recent years, a substantial amount of photovoltaic
(PV) generations have been installed in Japanese power systems.
However, the power output from the PV is random and
intermittent in nature. Therefore, the PV generation poses many
challenges to system operation. To evaluate impacts of the
behavior of PV on power supply reliability, we have developed
power supply reliability evaluation model considering a large
integration of PV generation into power system. As a result, the
power supply reliability is getting worse as the PV penetration
increases. To mitigate this issue, we have proposed that pumped
storage hydro power plant (PSHPP) is used to improve the
reliability. However, its operation may increase the power system
operational cost. In this paper a new method for scheduling the
effective operating pattern for PSHPP that makes it possible to
improve both reliability and economy is presented.
Index Terms--Genetic Algorithm, Monte Carlo Simulation,
Optimal Scheduling, Photovoltaic Generation, Power Supply
Reliability, Pumped Storage Hydro Power Plant, Surplus Power
Problem, Tabu Search.
I. INTRODUCTION
978-1-4244-8417-1/11/$26.00 2011
168 N
{u F ( P ) u (1 u
i
t
i
t
i
t 1
)S i}
(2)
t 1 i 1
Subject to:
Lt
PN PH PVt
(3)
i 1
i
i
Pmin
d Pt i d Pmax
(4)
Fi
ai Pi2 bi Pi ci
(1)
where a , b and c are fuel cost coefficients of generator i. Pi is
power output of generator i.
i
Ci
F i ( P i max )
P i max
(5)
PV Output [%]
where is the total fuel cost of thermal power plants for one
week. N is the number of thermal power plants. Pti is power
output of generator i at time t. uti is state variable of generator
i at time t (1: committed 0: stopped). Si is startup cost of
generator i. Lt is load demand at time t. PN is power output of
nuclear power plant. PH is power output of hydro power plant.
PVt is theoretical power output of PV at time t. Pimin is
minimum output of generator i. Pimax is maximum output of
generator i.
100
90
80
70
60
50
40
30
20
10
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
Generator Failure
Generators Re-dispatch
Weather
Average
Sunny
Cloudy
0.5
20
Rainy
0.1
0.03
PV Output [%]
G5
Supply Demand
G4
Load Demand
Load Demand
G2
G1
G1
G2
G4
G4
PSHPP
Second Step
Time t
Time t+1
G1
G1
TABLE I
RANDOM CHARACTERISTIC OF WEATHER CHANGES
Maximum Output
G3
G2
(7)
Load Demand
PVgt u F
PVdt
G4
Supply
Shortage
Supply Demand
Emergency
Control
First Step
Maximum Output
(6)
Load Demand
L dt Lgt u F
G2
LOLP
Sunny
Cloudy
LOSS
C u TIME
(9)
Rainy
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
Oi
1
, Pi
MTBFi
1
MTTRi
(8)
3) Generator Re-dispatch
During Monte Carlo simulation, if the generator failure
occurs at time t, the other generators are re-dispatched
immediately. Figure 4 shows how the generators are redispatched when generator G3 is tripped at time t, for example.
First, Generator G1, G2 and G4 are re-dispatched immediately.
If imbalance of power supply and demand cannot be avoided
by the re-dispatching, PSHPP is operated immediately to
avoid it as an emergency control. In the next time t+1, the
remaining generators are committed to compensate for the
B. Algorithm Overview
An algorithm overview of the proposed method is shown
in Fig. 5.
5000
64
Pareto Preservation Strategy
Roulette Selection
Selection
GA
Uniform Crossover
Mutation
Reproduction
Mutation Rate m
TS
0.05
10
64
Output of PSHPP
Pump
Mode Stop Generator Mode
C. Algorithm Details
1) Quantization of PSHPP operation pattern
The operation pattern of PSHPP is quantized by 6 steps at
each time t. Figure 6 shows quantization of PSHPP operation
pattern.
Time
Gene
2) Initial Solutions in GA
The initial genes of the PSHPP operation patterns are
generated by random number at each time period, which are
quantized as mentioned above. If the solution can not satisfy
the capacity constraint of the PSHPP, the solution is rejected
and another solution is generated.
Fig. 5. Algorithm Overview
3) Evaluation in GA
Generated genes are evaluated by Power Supply Reliability
Evaluation Model described in section II. The evaluation
indexes are power supply reliability and the fuel cost.
4) Determine the Pareto Optimal Solutions
The pareto optimal solutions are determined from
evaluated solutions. Solutions are labeled by pareto-ranking
method [4]. The ranking is numbered sequentially from
solutions closer to pareto optimal solutions
B. Simulation Condition
Figure 8 depicts two different load conditions for two
seasons, i.e. summer (August) and spring (May). The system
peak load is 3200 MW. In the simulation, the amount of PV
penetration is set to be 0MW and 1000MW.
6) Tabu Search
TS is conducted whenever GA is repeated 100 times. TS
can optimize only single objective function, TS is repeated
twice from each pareto optimal solutions. First, selecting one
of the pareto solution, TS is carried out from it. The objective
function of TS is LOLP improvement. Second, starting from
the same pareto solution, TS is conducted with objective
function for cost reduction. Figure 7 shows the searching from
pareto optimal solutions. This process is repeated until all
pareto optimal solutions are searched by TS and the pareto
optimal solutions are revised. Then the process goes back to
the GA. It is repeated until the number of generations reaches
Cmax.
3000
2500
2000
1500
1000
500
Summer (August)
Spring (May)
0
Sa t
Tu e
Day
We d
Thu
Fr i
V. SIMULATION RESULTS
3.7
3.6
PV 0MW
3.5
3.4
3.3
3.2
3.1
3
PV0 MW
PV1000 MW
2.9
0
0.05
0.1
0.15
0.2
0.25
0.15
0.2
0.25
LOLP
PV 0MW
1.6
Better
Mon
k 1
Sun
1.5
1.4
1.3
1.2
PV0 MW
PV1000 MW
1.1
0
0.05
0.1
LOLP
Best LOLP
Best Cost
2000
1500
1000
500
0
Sa t
Sun
Mon
Tue
Day
Wed
Thu
Fr i
Best LOLP
Best Cost
2000
1500
1000
500
0
Sa t
Sun
Mon
Tue
Day
Wed
Thu
Fr i
Best LOLP
Best Cost
2000
1500
1000
500
0
Sa t
Sun
Mon
Tue
Day
Wed
Thu
Fr i
Best LOLP
Capacity of Upper Reservoir [MWh]
Best Cost
2000
1500
1000
500
0
Sa t
Sun
Mon
Tue
Day
Wed
Thu
Fr i
7
Maximum Output of Thermal
Hydro
Nuclear
TABLE III
GENERATOR DATA OF IEEE RTS (MODIFIED)
4000
Cap. of
each
MTBF [h] MTTR [h]
Unit[MW]
12
5000
20
Generator
Group
Type
No. of
Unit
Thermal (Oil)
Thermal (Oil)
20
3000
50
Hydro
50
5000
40
300
Thermal (Coal)
76
3926
48
Thermal (Oil)
100
3926
48
Thermal (Coal)
155
4110
89
Thermal (Oil)
197
4110
89
Thermal (Coal)
350
5541
111
10
Nuclear
400
7000
168
3500
3000
2500
2000
1500
1000
TABLE IV
THERMAL POWER PLANT DATA
500
0
Sa t
Sun
Mo n
Tu e
We d
Th u
Day
Fig. 15. Generation Schedule (in Summer with PV 0MW, for Best Cost)
Maximum Output of Thermal
Hydro
Load Demand (w/o PV, w/o Pump)
Load Demand (with PV, w/o Pump)
Type
Fr i
No. of
Maximum
Minimum
Merit
Unit Output [MW] Output [MW] Order
Startup
Cost [$]
Thermal (Oil)
12
16-20
0.0646
16.8899
49.7108
Thermal (Oil)
20
10
21-24
0.025
43.5
200
181.6416
13.356
Thermal (Coal)
76
38
6-9
0.0533
9.2374
164.3492
1111.4208
Thermal (Oil)
100
50
13-15
0.0224
19.7128
287.4982
1511.8992
Thermal (Coal)
155
77.5
2-5
0.0067
10.2202
207.1786
1777.1544
Thermal (Oil)
197
98.5
10-12
0.0081
20.2909
378.5918
2070.18
Thermal (Coal)
350
175
0.0032
10.102
350.6398
8331.9264
4000
VIII. REFERENCES
3500
[1]
3000
2500
2000
[2]
1500
1000
500
[3]
0
Sa t
Sun
Mo n
Tu e
We d
Th u
Fr i
Day
Fig. 16. Generation Schedule (in Spring with PV 1000MW, for Best LOLP)
VI. CONCLUSION
This paper presents a new method for determining an
optimal PSHPP operation pattern which makes it possible to
improve both reliability and economy in the power systems
with a large integration of PV. The simulation results show
that the proposed method is very effective in terms of the
reliability of power system operation. The total fuel cost of
thermal power plants increases in order to secure the power
supply reliability using operation of the PSHPP in power
system with a large penetration of PV. In this paper, the hot
reserve capacity is not considered in scheduling of thermal
power plants. In the future work, it is necessary to develop a
cooperative scheduling method that optimizes both PSHPP
operation and hot reserve capacity of thermal power plants.
VII. APPENDIX
The generator data of IEEE RTS is tabulated in Table 3.
The cost curve and the startup cost of thermal power plants
are shown in Table 4.
[4]
[5]
[6]
IX. BIOGRAPHIES
Ryota Aihara was born in Tokyo, Japan, on
September 20, 1984. He received B.S. Eng. from
Sophia University, Tokyo, Japan in 2009 and M.S
from The University of Tokyo in 2011. Currently
he is pursuing the Ph.D. degree at The University
of Tokyo. He is a student member of IEEJ.
8
Akihiko Yokoyama was born in Osaka, Japan, on
October 9, 1956. He received B.S., M.S. and Dr.
Eng. from The University of Tokyo, Tokyo, Japan
in 1979, 1981 and 1984, respectively. He has been
with Department of Electrical Engineering, The
University of Tokyo since 1984 and currently a
professor in charge of Power System Engineering.
He is a member of IEEJ, IEEE and CIGRE.