Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: Wind, Solar photovoltaic and solar thermal power systems are emerging renewable energy technologies
Received 19 July 2011 and can be developed as viable options for electricity generation in future. In this paper, autonomous
Received in revised form 11 May 2012 hybrid generation systems consisting of wind turbine generators (WTGs), solar thermal power system
Accepted 14 May 2012
(STPS), solar photovoltaic (PV), diesel engine generators (DEGs), fuel cells (FCs), battery energy storage
Available online 23 June 2012
system (BESS), flywheel (FW), ultra capacitors (UCs) and aqua electrolyzer (AE) have been considered
for simulation studies. The power system frequency deviates for sudden changes in load or generation
Keywords:
or the both. The comparative performance of the controllers installed to alleviate this frequency deviation
Genetic algorithm
Aqua electrolyzer
for different hybrid systems, is carried out using time domain simulation. In practice, controllers (PI or
Fuel cell PID) are tuned manually which is difficult and time consuming. The computational intelligence has
Diesel engine generator opened paths to a new generation of advanced process control. Here, GA is used for optimization of con-
Battery energy storage system trollers’ gains of the proposed hybrid systems. The simulation results demonstrate the effectiveness of
Wind turbine generator the GA based controllers in terms of reduced settling time, overshoot and oscillations. The results are
compared with conventional controllers.
Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction energy storage in the form of batteries. A hybrid system can supply
power AC or DC or both [2]. Component or system control or both
In recent years the increasing concerns about the limited fossil is used to regulate the overall system operation.
fuel resources led to the awareness that the amount of energy im- Various optimization approaches, such as, genetic algorithm,
port should be decreased so as to become less dependent of oil particle swam optimization, artificial neural network, are applied
exporting countries. Further, the impact of fossil fuel on the envi- to optimize the gains of the controllers used in automatic genera-
ronment, especially the global warming and the harmful effects tion control. These techniques have not yet been reported to apply
of carbon emissions have created a new demand for clean and sus- in the field of hybrid energy systems for optimization of controller
tainable energy sources. Wind, sea, solar, biomass, geothermal gains. In [3,4] hybrid system studies proportional plus integral (PI)
powers are sustainable energy sources. Among these, wind and so- controller is used to regulate the output powers from distributed
lar have the potential to make significant contribution and hence generation system to achieve power balance condition due to sud-
assume great importance. Fuel cell (FC) also has the potential [1] den variations in generation and load. The gain values of PI control-
to be considered as one of the green power sources of the future. ler are chosen by trial and error method. In [5] the conventional PI
Off-grid electricity can be generated by single source system controller has traditionally been tuned by the method described in
such as using solar photovoltaic panels, wind turbine generators, Ziegler and Nichols. The controller gains once tuned for a given
micro-hydro plants, or fuel-powered combustion engine generator operating point are only suitable for limited operating point
sets, or by combining two or more of these electricity generating changes. Therefore, the use of the conventional PI controller does
sources in a so called hybrid system. The systems often include not meet the requirements of the robust performance [5]. More-
over, when the number of parameters to be optimized is large, con-
ventional technique for optimization is certainly not preferred one.
Abbreviations: GA, genetic algorithm; AE, aqua electrolyzer; DEG, diesel engine In this article, the genetic PI, PID controllers are tuned using GA
generator; WTG, wind turbine generators; FC, fuel cells; FW, flywheel; BESS, battery
technique. The GA is finding widespread applications in systems
energy storage system; PS, power system; PV, solar photovoltaic; STPS, solar
thermal power system. optimizations. As an intelligent control technology the GA can give
⇑ Corresponding author. Tel.: +91 9435172774; fax: +91 3842 233797. robust adaptive response with nonlinearity, parameter variation
E-mail addresses: dulal_nit@yahoo.co.in (D.Ch. Das), anjan_kumarroy@ and load disturbance effect [6–8]. Basics of genetic algorithm are
rediffmail.com (A.K. Roy), nidulsinha@hotmail.com (N. Sinha).
0142-0615/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2012.05.025
D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279 263
Nomenclature
illustrated in [9]. The results of applying the genetic PI, PID control- tions for wind turbine system, fuel cell, aqua electrolyzer, PV sys-
lers to the hybrid-power system are compared to those obtained tem, diesel engine generator are shown in this section [10].
by the application of a conventional PI, PID controllers respectively.
Simulated results show that the GA based controllers provide 2.1. Wind turbine model
improved dynamic performance than fixed gain conventional con-
trollers. The genetic controllers also show better transient perfor- The output of wind turbine generator depends on the wind
mance for load disturbances. speed at that instant. The characteristic of wind turbine generator
is illustrated in [3].
2. Proposed hybrid system The wind turbine system contains several nonlinearities. When
a wind turbine uses its pitch controller to counteract utility grid
The general block diagram of the proposed hybrid system is frequency oscillations, its output power varies between maximum,
shown in Fig. 1. Table 1 shows the parameters [3] of the proposed or rated power, and zero power. Hence, the pitch angle setpoint is
hybrid system. The system consists of wind turbine generators, nonlinearly limited by these boundaries. The pitch system, which
diesel generator, fuel cell, aqua electrolyzer, solar thermal and bat- turns the pitch angle according to wind speed, introduces a nonlin-
tery energy storage system. The power supplied to the load is the earity. The wind turbine can be simplified to a first order system.
sum of output powers from wind turbine generators, diesel gener- The transfer function of the WTG is represented by a first-order
ator, fuel cell and battery energy storage system. The aqua electro- lag [10] as
lyzer is used to absorb the fluctuations of wind turbine generator K WTG
and produce the hydrogen gas which is used as input to fuel cell GWTGðSÞ ¼ ð1Þ
sT WTG þ 1
generator. The mathematical models with first order transfer func-
Fig. 2a illustrates the wind farm model; it produces randomly
variable wind power. Random output fluctuation is derived from
white noise block with a low pass filter [11]. This model has been
Φ included in study case2.
Recently, trough solar power and solar power tower are two so-
lar thermal power systems (STPS) that are being explored and
Table 1
Parameters of proposed hybrid system.
1
1800 s + 1
_ + _
Band limited + +
x X
white noise +
+ _ %
Output
300s
300 s + 1
Initial
sqrt gain
load
Fig. 2a. Wind firm model.
Table 3
Simulation conditions for each case.
0 PID GA
-1
PID
-2
PI GA (dotted line)
-3 PI
0 1 2 3 4 5 6 7 8
Frequency deviation (Hz)
-0.1
PID
-0.2 PI GA (dotted line)
-0.3 PI
80 80.5 81 81.5 82 82.5 83 83.5 84 84.5 85
concede, is that hydrogen is still more expensive than other energy fluctuation is generated from white noise block, using low-pass
sources such as coal, oil and natural gas. Fuel cell generator is a and high-pass filters respectively [11]. This model has been in-
higher order model and has non linearity. For low frequency do- cluded in study case2.
main analysis it is represented by a first order lag transfer function
model given as [10] 2.8. Battery energy storage system
K FC
GFC ðSÞ ¼ ð7Þ The short time power fluctuation from wind, solar or solar ther-
T FC s þ 1
mal causes large problems for power systems operation. A possible
solution is storage of wind energy. Due to very good technical char-
2.7. Load model acteristics (large energy density, fast access time) the battery en-
ergy storage system has been an effective energy storage
To examine effects of loading power demand variation a vari- technology to store large amount of wind energy [22]. They can
able load model is developed as shown in Fig. 2c. The random supply the system with a large amount of the power in a short
D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279 267
1.2
PS (p.u)
Load demand at t=80s increased by 10%
0.8
At t=40s wind power changes from .5 p.u to .4 p.u and
0.6
solar thermal power changes from .36 p.u to .18p.u
0.4
-0.2
At t=40s wind power changes from .5 p.u to .4 p.u and
-0.4 solar thermal power changes from .36 p.u to .18p.u
-0.6
0 20 40 60 80 100 120
Time (s)
Fig. 4b. Total power generation, error between supply and demand.
-3
x 10
0
PAE (p.u)
-5 PID GA PI GA
PI
-10
PID
-15
0 20 40 60 80 100 120
0.4 PI PID
(p.u)
0.3
DEG
0.2 PI PI GA
P
0.1
0
0 20 40 60 80 100 120
0.4
(p.u)
0.2
P
0.1
0
0 20 40 60 80 100 120
(p.u)
0.3
PI
0.2 PI
FC
PID
P
0.1 PI
0
0 20 40 60 80 100 120
Time (s)
Fig. 4c. Power outputs of DEG, BESS, FC and input power to AE.
time, or large amount of energy for a longer period. The battery 2.9. Ultra capacitor
energy storage system (BESS) consists of a battery bank and a
power converter [23] that interfaces the battery bank to the auton- Ultra capacitor or super capacitor also called double layer
omous utility grid. A higher power capacity can be achieved by capacitor, is an emerging device for energy storage. It stores charge
connecting more modules. The transfer function model of battery in a double layer formed on a large surface area of micro-porous
energy storage system expressed by first order as [10] material such as activated carbon [24]. It has specific energy in
the range of 1–10 W h/Kg and high specific power in the range of
K BESS 1000–5000 W/Kg. The charge/discharge efficiency (85–98%) and
GBESS ¼ ð8Þ
T BESS s þ 1 rate of discharge (0.3–30 s) very high. It offers large capacitances
268 D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279
1.2
1
Power (p.u)
0.8 Load demand
Table 4
Gains of PI/PID controllers.
in the order of thousands of farads but at a low voltage of about rotation [28]. When the machine works as generator, the electrical
2.5 V [25]. The combination of ultra capacitors and batteries can energy is delivered to the load. The rotor of a flywheel made of
take the advantage of each kind of device to yield a power source composite materials instead of steel because of the composite’s
of high power capability and longer run time. Neglecting all the
non-linearities, the transfer function of ultra capacitor is given by
Best: 0.076383 Mean: 0.076491
first order lag: 10
Best fitness
K UC 9 Mean fitness
GUC ðSÞ ¼ ð9Þ
T CU s þ 1 8
7
2.10. Flywheel energy storage system
Fitness value
6
The few day power variation can be balanced by hydrogen 5 PI
based energy storage, while the short time turbulent power pulsa-
tion produced by wind turbine generator or solar thermal power 4
system can be balanced by flywheel energy storage drive [26]. It 3
is more suitable for repetitively absorbing and releasing electric Best fitness
2
energy for a short period of time than a battery energy storage sys-
tem [27]. The flywheel energy storage is rediscovered nowadays 1 Mean fitness
for wide range of powers due to its advantages over the other en-
0
ergy storage systems: reliability, long-life, cost, fast response [26]. 10 20 30 40 50 60 70 80 90 100
Flywheel energy storage systems operate by storing energy Generation
mechanically in a rotating flywheel. The generating motor is used
to rotate the flywheel and to generate electricity from flywheel Fig. 4e. Plot of fitness function value versus generation of PI controller.
D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279 269
Table 5
Gains of PI/PID controllers.
Z
ability to withstand the rotating forces exerted on the flywheel. In
I¼ X 2 dmx ð11Þ
order to store energy the flywheel is placed in a sealed container
which is then placed in a vacuum to reduce air resistance. The
The solution for a cylindrical flywheel of mass ‘m’ and radius ‘r’ is
amount of energy stored by flywheel may be given by
1 2 1
E¼ Ix ð10Þ I ¼ m r 2 and E ¼ m r 2 x2 ð12Þ
2 2
The amount of energy stored varies linearly with the moment of
where I is the moment of inertia and x is the angular velocity.
inertia of the flywheel and the square of its angular velocity. Fly-
The moment of inertia is the integral of the square of the dis-
wheels can be designed for low-speed or high-speed operation. A
tance ‘x’ from the axis of rotation to the differential mass ‘dmx’.
low-speed flywheel has the advantages of lower cost and the use
of proven technologies when compared to a high-speed flywheel
Best: 0.073716 Mean: 0.073822 system. The main disadvantages are less stored energy per volume,
10 higher losses, and increased volume and mass [29]. Since the en-
Best fitness
9 Mean fitness ergy stored is proportional to the square of angular velocity,
increasing the angular speed increases stored energy more effec-
8 tively than increasing mass. But increasing angular speeds results
7 in increased frictional losses and hence thermal problems. With
the help of magnetic bearing technology, the frictional losses due
Fitness value
0.995
0.99
0 50 100 150 200 250 300 350 400
0.2
(p.u)
0.18
STPS
P
0.16
0.36
0.34
0.32
0.3
0.28
0 50 100 150 200 250 300 350 400
Time (s)
Fig. 5a. Load demand, output power of STPS and WTG under. Randomly varying conditions.
Df 1 1
Gsys ðsÞ ¼ ¼ ¼ ð16Þ
DPe K sys ð1 þ sT sys Þ Ms þ D
PI GA
0
PID GA 2.12. Genetic algorithm
-1 PI
1.004
PI with gains optimized under varying
wind, solar thermal and load (broken line)
1.002
1
PI GA with gains optimized under varying
wind, solar thermal and load (solid line)
0.998
200 205 210 215 220 225 230 235 240 245 250
2
P (p.u)
1
S
0
0 50 100 150 200 250 300 350 400
Time (s)
1.004 PID with gains optimized under varying
wind, solar thermal and load (dotted line)
1.002
1
PID GA with gains optimized under varying
0.998 wind, solar thermal and load (solid line)
0.996
200 205 210 215 220 225 230 235 240 245 250
-3
x 10
0
PAE (p.u)
-20
0 50 100 150 200 250 300 350 400
0.5
PID
(p.u)
0.4
PI GA
P
0.35
0 50 100 150 200 250 300 350 400
PI GA
(p.u)
0.1
0.08
BESS
0.06
P
0.04 PID GA PI
PID
0 50 100 150 200 250 300 350 400
(p.u)
0.04
0.02 PID GA PI
FC
PID PI GA
P
0
0 50 100 150 200 250 300 350 400
Time (s)
Fig. 5d. Output power of DEG, BESS, FC and input power of AE.
Step 4: Generation of offspring: Offspring is a new chromosome of the fittest mechanism in nature. Following the selection of
obtained through the steps of selection, crossover and muta- parent population, crossover and mutation are performed to
tion. After fitness of each chromosome is computed, parent generate offspring population. The crossover and mutation are
solutions are selected for reproduction. It emulates the survival performed based on the probability of crossover and mutation.
272 D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279
Step 5: Replace the current population with the new Best: 0.056832 Mean: 0.057107
10
population. Best fitness
Step 6: Terminate the program if termination criterion is 9 Mean fitness
reached; else go to step2.
8
In PID controller design methods, the most common objective
7
function are integrated absolute error (IAE), the integrated of time
Fitness value
weight square error (ITSE) and integrated of squared error (ISE), 6 PID
etc. These three integral performance criteria have their own
5
advantages and disadvantages. In this paper, integral of the square
of frequency deviation is chosen as the objective function. 4
Z a
J¼ ðDf Þ2 dt ð17Þ 3 Mean fitness
0
2
Minimize J
1 Best fitness
Subject to
ð18Þ 0
K min
p K p K max
p
10 20 30 40 50 60 70 80 90 100
Generation
K min
i K i K max
i ð19Þ
Fig. 5e. Plot of fitness function value versus generation of PID.
K min
d K d K max
d ð20Þ
Typical range selected for each of Kp, Ki, Kd is [2000 to 2000]. Best: 0.09982 Mean: 0.10009
10
The minimization of the objective function or the fitness func- Best fitness
tion is implemented to determine the optimal parameters of PID 9 Mean fitness
controllers. The gains of PI controllers are optimized in the similar
manner. 8
Genetic algorithm (GA) is a computationally simple and power- 7
ful algorithm and has been employed successfully in the field of
Fitness value
3. Simulation results and analysis Fig. 5f. Plot of fitness function value versus generation of PI.
In this section, dynamic performances of the proposed hybrid justed by fuel cell, battery energy storage device and diesel
generation systems are analyzed using time-domain simulation. generator through the controllers since the wind and solar thermal
The responses of different combinations under various operating generator generate constant power. The power generation in this
points and disturbance conditions are presented with optimum case can be expressed by
gain settings of conventional and GA based PI and PID controllers
respectively. In the conventional optimization approach, a sequen- PS ¼ P DEG þ PWTG þ PSTPS þ P FC P AE PBESS ð21Þ
tial optimization method is used where one parameter is opti- The power system frequency fluctuates due to these sudden
mized at a time using ISE criterion keeping the other parameters changes in power generation by wind and solar thermal and load
fixed. Then this operation repeats for every other parameters in- demand. This deviation in frequency is controlled by the control-
tern to complete one iteration of optimization. The following four lers (PI/PID) and the outputs of system components are automati-
cases, as shown in Table 3 are considered for case studies. cally adjusted to corresponding values such that the error in supply
demand and the deviation in frequency are minimum. Fig. 4d
3.1. Time-domain analysis:case1 shows the step changes in wind, solar thermal power and load de-
mand. When wind power decreased from 0.5 p.u to 0.4 p.u, solar
In this case, during 0 < t 6 40 s, the average wind power and so- thermal power reduced from 0.36 p.u to 0.18 p.u after 40 s, and
lar thermal power is kept 0.5 p.u and 0.36 p.u respectively. Load the load increased from 1 p.u to 1.1 p.u suddenly at 40 s, the power
demand is 1 p.u during 0 < t 6 80 s. After 40 s wind power and so- outputs of the fuel cell, battery energy storage, and diesel genera-
lar thermal power is suddenly decreased to 0.36 p.u and 0.18 p.u tors are increased to different values for different controllers. Thus,
respectively. At t = 80 s load is suddenly increased from 1 p.u to the mismatch in generation and demand is alleviated. Finally the
1.1 p.u. The deviation in generation and load is automatically ad- frequency settles to a steady state value. The gain values of the
D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279 273
-3
x 10
6
PI-GA with gains optimized using
4 step change in wind, solar thermal and load
X: 86.69
2 Y: 0.003928
X: 86.53
Y: 0.002428
0
-2
Frequency deviation (Hz) -4 PI-GA with gains optimized undervarying wind, solar thermal and load
-5
0 -3
50 100 150 200 250 300 350 400
x 10 Time (s)
10
X: 89.27 PID-GA with gains optimized under
Y: 0.006324 varying wind, solar thermal and load
5 PID-GA with gains optimized using step
X: 89.12 change in wind, solar thermal and load
Y: 0.004954
-5
80 85 90 95 100 105 110 115 120
0.4
0.2 PI GA
0
-0.2
-0.4 PID GA
-0.6
-0.8
-1
Frequency deviation (Hz)
-1.2
0 0.5 1 1.5 2 2.5 3
1
-1
-2
0 20 40 60 80 100 120
Time (s)
PI GA(dotted line)
0
controllers obtained through conventional and GA technique and load demand. Input to this AE for the remaining period is zero.
are given in Tables 4 and 5. The power outputs of the fuel cell, bat- Fig. 4a shows the frequency deviation of the hybrid power system.
tery energy storage, diesel generators and input power to AE are It may be observed that that the response of PID controller is the
presented in Fig. 4c. During 4–8 s, AE, which uses GA based PID best amongst PI and PID controllers considered for study in terms
or PID controller, absorbs some power as generation is higher than of peak transient deviation and settling time. Further, it is observed
274 D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279
1.5
0.5
0
0 20 40 60 80 100 120
Time (s)
PS (p.u)
1.2
PI
1.1 PI GA
1
79.9 80 80.1 80.2 80.3 80.4 80.5 80.6 80.7 80.8 80.9
0.4
(p.u)
0.3
PID (broken line) PI PI GA
0.2
BESS
0
0 20 40 60 80 100 120
-3
x 10
0
PAE (p.u)
-2
-4 PI GA PID, PI
-6 PID GA
0 20 40 60 80 100 120
0.5 PID GA
(p.u)
PI
0.4
DEG
0.3
PI GA
0.2
P
PI PID GA
PI GA
0.2
FW
P
0
0 20 40 60 80 100 120
PI GA
(p.u)
0.2
PI PID
FC
0.1
PID GA
P
0
0 20 40 60 80 100 120
Time (s)
Fig. 6c. Output power of BESS, DEG, FW, FC and input power of AE.
that the performance of all the GA optimized controllers is better supply demand. The plots of fitness function value versus genera-
than their respective counterparts optimized using conventional tion for PI and PID are shown in Figs. 4e and 4f respectively. In all
method. Fig. 4b represents total power generation and error in the cases PID represents conventional PID.
D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279 275
1.2 X: 81
Load demand Y: 1.3
Power (p.u)
0.8
X: 39.96
0.6 Y: 0.5
Wind X: 77.28
Y: 0.4002
0.4
X: 39.96
Y: 0.2 Solar
X: 47.67
0.2 Y: 0.1014
0
0 20 40 60 80 100 120
Time (s)
8 8
7 7
Fitness value
Fitness value
6 6
5 5 PI
3.2. Time-domain analysis: case2 technique as well as GA optimization technique. It is observed that
the performance of all the GA optimized controllers is better than
In this study, simulation time considered is 400 s. In order to their respective counterparts optimized conventional method.
examine the effects of practically variable wind, solar thermal However, the GA based PID controller out-performed the conven-
and load power on dynamic performance of the hybrid system, tionally optimized PI, and PID and the GA optimized PI controllers.
randomly variable wind, solar thermal power and load model are The gain values of the controllers obtained through conventional
considered. These models are shown in Figs. 2a, 2b and 2c respec- and GA technique and are presented in Table 4. The initial and final
tively. The net power generation in this case also may be expressed values of GA optimized parameters are also presented in this table.
by Further, in order to examine the robustness of the GA optimized
PS ¼ PDEG þ PWTG þ PSTPS þ PFC PAE PBESS ð22Þ controllers, their performance with their gains optimized using
step change load, wind power, solar thermal power are compared
Fig. 5a shows the load demand, output power of solar thermal sys- with the controllers with their gains optimized under varying load,
tem and output power of wind generators under randomly varying wind and solar thermal power. Comparative performance of these
conditions. In order to eliminate the mismatch between generation controllers may be observed from the responses of frequency devi-
and load demand, the output power of the battery energy storage ation as shown in Fig. 5g. It may be observed that there is a negli-
system, diesel generators and the fuel cells are altered by installing gible frequency and power difference between the two. This shows
controllers. The power outputs of these generating systems are the robustness of the controllers.
changed to different values for different controllers. This in turn
alleviates the frequency fluctuation of system. The output powers
and input power to the AE are presented in Fig. 5d. It is observed 3.3. Time-domain analysis: case3
that AE, which uses PID controller, absorbs a small amount of power
as generation is higher than load demand. Input power to AE, which In this case, the hybrid system comprises of wind turbine gen-
uses any of the other controllers, remains zero. The plots of fitness erator, solar photovoltaic, aqua electrolyzer, diesel engine genera-
function value versus generation for PID and PI are shown in Figs. 5e tor, battery energy storage system and flywheel energy storage
and 5f respectively. Figs. 5b, and 5c shows the frequency deviation, system. The combination of photovoltaic power system, wind gen-
and total power generation for conventionally optimization erator, diesel engine and energy storage systems increases the
276 D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279
1
PI (broken line)
0.5
PID GA(solid line)
0
PID
-0.5
PI GA (dotted line)
Frequency deviation (Hz) -1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
0.2
-0.2
-0.4
0 20 40 60 80 100 120
Time (s)
0.2
PI GA (dotted line) PID GA(solid line)
-0.2 PID
PI (broken line)
-0.4
80 80.5 81 81.5 82 82.5
the power outputs of the fuel cell, battery energy storage system,
flywheel, diesel generators and input power to AE are also in-
2 creased to different values for different controllers respectively.
This in turn, alleviates the mismatch in generation and demand. Fi-
1.5
nally the frequency settles to a steady state value. Figs. 6a and 6b,
1 represents respectively, the frequency deviation of system and to-
tal generation. The plots of fitness function value versus generation
0.5 for PID and PI are shown in Figs. 6e and 6f respectively. The power
P (p.u)
Time (s) PI
1.5 PI GA
PID 3.4. P. Time-domain analysis: case4
1.4 PID GA
1.3 In this case the hybrid system considered is same as case3 ex-
1.2 cept that flywheel is replaced by ultra-capacitor. In order to exam-
ine the fact that ultra capacitor delivers high power within a short
1.1
duration of time during peak load demand it is incorporated in hy-
1 brid system. The net power generation in this case may be ex-
80 80.2 80.4 80.6 80.8 81
pressed as
Fig. 7b. Total power generation.
PS ¼ P DEG þ PWTG þ PPV þ PFC PAE PBESS PFW ð24Þ
The power system frequency fluctuates due to these sudden
reliability [34]. The flywheel energy storage system is incorporated changes in power generation by wind and solar photovoltaic and
because it is capable of mitigating the short time power fluctuation load demand. This deviation in frequency is controlled by the con-
due to intermittent generation of wind generator and PV system. trollers (PI/PID) and the outputs of system components are auto-
The net power generation in this case may be expressed as matically adjusted to corresponding values such that the error in
supply demand and the deviation in frequency are minimum.
PS ¼ PDEG þ P WTG þ PPV þ PFC PAE PBESS PFW ð23Þ
Simulation results are shown in Figs. 7a–7f. Fig. 7d shows the
The power system frequency fluctuates due to these sudden step changes in solar photo voltaic, wind power and load. When
changes in power generation by wind and solar photovoltaic and wind power decreased from 0.5 p.u to 0.4 p.u at 40 s, solar photo
load demand. This deviation in frequency is controlled by the con- voltaic power reduced from 0.2 p.u to 0.1 p.u at 40 s and the load
trollers (PI/PID) and the outputs of system components are auto- increased from 1 p.u to 1.3 p.u suddenly at 80 s, the power outputs
matically adjusted to corresponding values such that the error in of the fuel cell, battery energy storage system, ultra capacitor, and
supply demand and the deviation in frequency are minimum. diesel generators are increased to different values for different con-
Fig. 6d shows the step changes, solar photo voltaic, wind power trollers respectively. This in turn, alleviates the mismatch in gener-
and load. When wind power decreased from 0.5 p.u to 0.4 p.u at ation and demand. Finally the frequency settles to a steady state
40 s, solar photo voltaic power reduced from 0.2 p.u to 0.1 p.u at value. Figs. 7a, and 7b, represents respectively, the frequency devi-
40 s and the load increased from 1 p.u to 1.8 p.u suddenly at 80 s, ation of system, total generation. The plots of fitness function value
D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279 277
1
PI GA (dotted line)
(p.u)
PID GA (solid line)
0.5
UC
P
PI (broken line) PID
0
0 20 40 60 80 100 120
-4
x 10
(p.u) 0
PI (solid line)
-10 PID GA, PID
AE
P
PI GA
-20
0 20 40 60 80 100 120
0.06
PID
(p.u)
0.04
PI (dotted line)
BESS
0.02 PID GA
P
0
PI GA(solid line)
0 20 40 60 80 100 120
PI GA (solid line)
0.05 PID GA
FC
P
0
0 20 40 60 80 100 120
(p.u)
0.1
PI GA (dotted line)
P
0
0 20 40 60 80 100 120
Time (s)
Fig. 7c. Output power of UC, BESS, FC, DEG and input power of AE.
1.4
1.2 X: 81
Y: 1.3
Load demand
1
Power (p.u)
0.8
X: 11.2
0.6 Y: 0.4993
X: 45.98
Y: 0.4021 Wind
0.4
X: 10.39
Y: 0.1994 Solar photovoltaic X: 96.91
0.2 Y: 0.1
0
0 20 40 60 80 100 120
Time (s)
versus generation for PID and PI are shown in Figs. 7e and 7f 4. Conclusion
respectively. The power outputs of the fuel cell, battery energy
storage system, ultra capacitor, diesel generators and input power The autonomous hybrid generation/energy storage system re-
to AE are shown in Fig. 7c. quires an automatic generation control system to eliminate the
278 D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279
Best: 0.08786 Mean: 0.088605 method. However, in all cases, the GA based PID controller out-per-
10
Best fitness
formed the conventionally optimized PI, and PID and the GA opti-
9 Mean fitness mized PI controllers. It can be concluded from the simulation
results that GA-based optimization technique is much better to en-
8 able automatic generation control.
7
Fitness value
6 Acknowledgement
5 PID
Authors wish to thank Electrical Engineering Department, NIT
4
Best fitness Silchar, for providing the necessary facilities for completing this
work.
3
2
Best fitness References
1
[1] Hajizadeh Amin, Aliakbar Golkar Masoud. Control of hybrid fuel cell/energy
0 storage distributed generation system against voltage sag. Int J Electr Power
10 20 30 40 50 60 70 80 90 100
Energy Syst 2010;32(5):488–97.
Generation [2] Kurohane K, Senjyu T, Toshihisa Yona A, Urasaki N, Goya T, Funabashi T. A
hybrid smart AC/DC power system. IEEE Trans Smart Grid 2010;1(2):199–204.
Fig. 7e. Plot of fitness value versus generation of PID controller. [3] Senjyu T, Nakaji T, Uezato K, Funabashi T. A hybrid power system using
alternative energy facilities in isolated islands. IEEE Trans Energy Convers
2005;20(2):406–14.
[4] Kumar BS, Mishra S, Senroy N. Agc for distributed generation. Proc Int Conf
Best: 0.25855 Mean: 0.25922 Sustain Energy Technol 2008:89–94.
10 [5] Cetin Elmas, Tuncay Yigit. Genetic Algorithm based on-line tuning of a PI
Best fitness
Mean fitness controller for a switched reluctance motor drive. Electr Power Comp Syst
9 2007;35(6):675–91.
[6] Fleming PJ, Purshouse RC. Evolutionary algorithms in control system
8 engineering: a survey. Control Eng Pract 2002;10:1223–41.
[7] Krohling RA, Rey JP. Design of optimal disturbance rejection PID controllers
7
using genetic algorithms. IEEE Trans Evol Comput 2001;5:78–82.
Fitness value
[8] Karr C, Gentry E. Fuzzy control of ph using genetic algorithms. IEEE Trans Fuzzy
6
Syst 1993;1(1):46–53.
[9] Kim Jin-Sung, Kim Jin-Hwan, Park Ji-Mo, Park Sung-Man, Choe Won-Yong, Heo
5
Hoon. Auto tuning PID controller based on improved genetic algorithm for
reverse osmosis plant. Int J Intell Syst Technol 2008;3(4).
4 PI
[10] Lee Dong-Jing, Wang Li. Small-signal stability analysis of an autonomous
3 hybrid renewable energy power generation/ energy storage system – Part I:
Best fitness time-domain simulations. IEEE Trans Energy Convers 2008;23(1):311–20.
2 [11] Matsubara M, Fujita G, Shinji T, Sekine T, Akisawa A, Kashiwagi T, et al. Supply
and demand control of dispersed type power sources in micro grid. In: Proc.
1 Best fitness 13th Int. Conf. ISAP’05. p. 67–72.
[12] Letcher TrevorM. Future Energy: Improved. Sustainable and clean options for
0 our planet. UK: Elsevier; 2009.
10 20 30 40 50 60 70 80 90 100 [13] Liu DY, Wang J, Feng XQ, Guo S, Xu C. Investigation and analysis on the
combined operation of solar thermal power and conventional thermal power.
Generation In: IEEE international conference on sustainable power generation and supply,
SUPERGEN ’09; 2009. p. 1–6.
Fig. 7f. Plot of fitness value versus generation of PI Controller. [14] Tai Lü, Nan Li, Zhen Zhang, Qi-chun Zhu. Study on the continuous and stable
running mode of solar thermal power plant, sustainable power generation and
supply, 2009. SUPERGEN ‘09. International conference; April 2009. p. 1–4.
mismatch in supply and demand under varying condition of load [15] www.Newtechpapers.com.
[16] Todd Woody. In: California’s Mojave desert, solar-thermal projects take off.
and generation. This paper presents models of an autonomous hy- Yale Environment 360; 27 October 2010.
brid generation/energy storage system. The studied hybrid system [17] Ucilia Wang. The rise of concentrating solar thermal power, renewable energy
consists of wind turbine generators, diesel generators, aqua elec- world; 6 June 2011.
[18] Blaabjerg F, Chen Z, Kjaer SB. Power electronics as efficient interface in
trolyzers, solar thermal power system, photovoltaic, battery energy dispersed power generation systems. IEEE Trans Power Electron 2004;19(5).
storage system, flywheel, ultra-capacitors and fuel cells. The gener- [19] Wang C, Nehrir MH. Power management of a stand-alone wind/photovoltaic/
ated hydrogen from the aqua electrolyzer is used as fuel for the fuel cell energy system. IEEE Trans Energy Convers 2008;23(3):957–67.
[20] Kim S-K, Jeon JH, Cho CH, Ahn J-B, Kwon SH. Dynamic modeling and control of
fuel cell. The models for wind power, solar thermal and load are
a grid-connected hybrid generation system with versatile power transfer. IEEE
also properly selected to simulate the important performance of Trans Ind Electron 2008;55(4):677–1688.
the studied hybrid system. [21] Sedaghat B, Jalilvand A, Noroozian R. Design of a multilevel control strategy for
integration of stand-alone wind/diesel system. Int J Electr Power Energy Syst
In order to reduce the frequency deviation i.e., eliminate the
2012;35(1):123–37.
mismatch in supply and demand under varying condition of load [22] Danesh Shakib A, Balzer G. Optimization of battery application for wind energy
and generation, the output power from the sources is regulated storage. International conference on renewable energies and powet quality,
using controllers (PI/PID). The gains of controllers are designed Granada, Spain; 23th to 25th March, 2010.
[23] Sebastian R. Modelling and simulation of a high penetration wind diesel
by using conventional method and GA. Performance of each con- system with battery energy storage. Int J Electr Power Energy Syst
troller is examined from the dynamic behaviour in time-domain 2011;33(3):767–74.
simulations of the system in isolated mode of operation. It has [24] Gao L, Dougal RA, Liu S. Power enhancement of an actively controlled battery/
ultracapacitor hybrid. IEEE Trans Power Electron 2005;20(1):236–43.
been observed that the response of PID controller is the best [25] Rajakaruna RMAS. Small-signal transfer functions of the classical boost
amongst PI and PID controllers considered for study in terms of converter supplied by ultracapacitor banks. Second IEEE conference on
peak transient deviation and settling time. Further, it is observed industrial electronics and applications; 2007. p. 692–7.
[26] Veszprémi K, Schmidt I. Flywheel energy storage drive for wind turbines. In:
that the performance of all the GA optimized controllers is better Proc. 7th international conference on power electronics and drive systems,
than their respective counterparts optimized using conventional PEDS’2007, Bangkok; November 2007. p. 916–23. ISBN:1-4244-0645-5.
D.Ch. Das et al. / Electrical Power and Energy Systems 43 (2012) 262–279 279
[27] Akagi H, Sato H. Control and performance of a doubly-fed induction machine [31] Fogel DB. Evolutionary computation toward a new philosophy of machine
intended for a flywheel energy storage system. IEEE Trans Power Electron intelligence. New York: IEEE; 1995.
2002;7(1). [32] Eberhart RC, Shi Y. Comparison between genetic algorithms and particle
[28] Subkhan M, Komori M. New concept for flywheel energy storage system using swarm optimization. In: Proc IEEE Int Conf Evol Comput, Anchorage, AK; 1998.
SMB and PMB. IEEE Trans Appl Superconduct 2011;21(3):1485–8. p. 611–6.
[29] Samieni S, Johnson BK, Hess HL, law JD. Modeling and analysis of a flywheel [33] Gaing Z-L. A particle swarm optimization approach for optimum design of PID
energy storage system for voltage sag correction. IEEE Trans Ind Appl controller in AVR system. IEEE Trans Energy Convers 2004;19:384–91.
2006;42(1):42–52. [34] Kaldellis JK, Zafirakisa D, Kondili E. Optimum sizing of photovoltaic-energy
[30] Shivakumar R, Lakshmipathi R. Implementation of an innovative bio inspired storage systems for autonomous small islands. Int J Electr Power Energy Syst
GA and PSO algorithm for controller design considering steam GT dynamics. 2010;32(1):24–36.
IJCSI Int J Comput Sci Issues 2010;7(1):18–28. No. 3.