You are on page 1of 10

Electrical Power and Energy Systems 75 (2016) 205–214

Contents lists available at ScienceDirect

Electrical Power and Energy Systems


journal homepage: www.elsevier.com/locate/ijepes

Automatic generation control of multi source power generation under


deregulated environment
P.K. Hota ⇑, Banaja Mohanty
Department of Electrical Engineering, Veer Surendra Sai University of Technology (VSSUT), Burla 768018, Odisha, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a novel scheme for automatic generation control (AGC) of interconnected two area
Received 15 January 2014 multi source power generation under deregulated environment. The conventional two-area AGC system
Received in revised form 3 September 2015 is modified to take into account the effect of bilateral contracts on the dynamics. In the considered two
Accepted 11 September 2015
area power system, each area contains two GENCOs; first GENCO is a reheat steam turbine with
appropriate generation rate constraint nonlinearity and second GENCO is a gas turbine generation. The
performances of integral (I), proportional-integral (PI), integral-derivative (ID) and proportional-
Keywords:
integral-derivative (PID) are evaluated for the proposed AGC system in the deregulated environment.
Automatic Generation Control (AGC)
Multi-source power generation system
The gain of the controllers and speed regulation parameters are optimised using differential evolution
Proportional-integral-derivative (PID) (DE) algorithm. Differential evolution algorithm is used because of its convergence superiority and easy
controller to implement. The performance of DE algorithm applied to the proposed problem is compared with that
Differential evolution (DE) algorithm of Genetic Algorithm (GA) to establish its optimisation superiority of the former.
Ó 2015 Elsevier Ltd. All rights reserved.

Introduction in deregulated power system. Under the supervision of ISO, the


DISCO can contract any amount of power from the GENCO. These
Electric power utilities throughout the world are currently are implemented through DISCO participation matrix (DPM), the
undergoing major restructuring processes and are adopting the concept of DPM and area participation factors (apf) are illustrated
deregulated market operation. Competition has been introduced in [9]. The authors in [9] used trajectory sensitivity to find out opti-
in power systems around the world based on the premise that it mal parameters of the system using gradient Newton algorithm.
will increase the efficiency of the industrial sector and reduce the Many intelligent techniques such as Genetic Algorithm bacteria
cost of electrical energy of all customers. In order to control electric forging optimisation algorithm (BFOA), are used to optimise con-
power industry, government has set up some restructured rules troller gains of AGC under deregulated power system. In [10] the
and economic incentives where the collection of those restructured author has reported GA based integral controller in multi-area
rules called deregulation. Deregulated system consists of genera- power system in deregulated environment considering
tion companies (GENCOs), distribution companies (DISCOs), trans- hydro–thermal power generating units. In [11], BFOA is used in
mission companies (TRANSCOs) and independent system operator multi-area thermal system under deregulated environment using
(ISO). Those entities like GENCO, TRANSCO, DISCO, ISO and many non-integer control. Decentralised controller [12] is also imple-
ancillary services do have different roles to play and therefore have mented in deregulated power system. Ghoshal et al. [13] have used
to be modelled differently. The more challenging issue that has a Genetic Algorithm (GA) to optimize controller gains of a multi-
come up after deregulation is the ancillary service which is essen- area hydro–thermal AGC system. In another paper, Ghoshal [14]
tial for maintaining the electrical system security and reliability has proposed a scheme of GA/GA-SA based fuzzy control for AGC
together. One of them is Automatic Generation Control (AGC) that of a multi-area thermal generating system. He has reported better
restores mismatches between generation and load and keeps the results in comparison to his previous method. The premature con-
system stable. vergence of GA degrades its efficiency and reduces the search capa-
Several authors have reported AGC in deregulated power sys- bility. Differential evolution (DE) is a branch of evolutionary
tem [1–8]. The authors [7,8] represented price based simulation algorithms developed by Rainer Stron and Kenneth Price in 1995
for optimization problems [15]. It is a population-based direct
⇑ Corresponding author. search algorithm for global optimization capable of handling
E-mail addresses: p_hota@rediffmail.com (P.K. Hota), banaja_m@yahoo.com non-differentiable, non-linear and multi-modal objective func-
(B. Mohanty). tions, with few, easily chosen, control parameters. DE differs from

http://dx.doi.org/10.1016/j.ijepes.2015.09.003
0142-0615/Ó 2015 Elsevier Ltd. All rights reserved.
206 P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214

other Evolutionary Algorithms (EA) in the mutation and recombi- The steady state power flow on the tie-line is given by:
nation phases. DE uses weighted differences between solution vec-
DPtie12;scheduled ¼ ðDemand of DISCOs in area
tors to change the population whereas in other stochastic
techniques such as Genetic Algorithm (GA), perturbation occurs  2 from GENCOs in area  1Þ
in accordance with a random quantity. DE employs a greedy selec-  ðDemand of DISCOs in area
tion process with inherent elitist features. DE has been successfully
applied to SSSC-based damping controller design [16] and also to
 1 from GENCOs in area  2Þ ð1Þ
robust coordinated design of multiple and multi-type damping At any given time, the tie-line power error is given by:
controllers [17] where the performances are found to be
promising. DPtie12;error ¼ DPtie12;actual  DP tie12;scheduled ð2Þ
In the above literature, all authors have studied either thermal This error in tie-line power is used to generate ACE signal as in the
or hydro power plants. Keeping in view the present power scenar- normal AGC system.
io, the combination of multi-source generators is more realistic for
the study of AGC. The control area may have the combination of e1 ðtÞ ¼ ACE1 ¼ B1 Df 1 þ DPtie12;error ð3Þ
thermal, hydro, gas, nuclear, renewable energy sources, etc. [18]. e2 ðtÞ ¼ ACE2 ¼ B2 Df 2 þ DPtie21;error ð4Þ
It is quite apparent from literature survey that hardly any author DPtie21;error ¼ a12 DPtie12;error ð5Þ
has reported multi-source generation system in deregulated
power system. In view of the above discussion, the authors have where a12 ¼  PPr1
r2
; Pr1 and Pr2 are the rated power of area-1 and 2,
taken a maiden attempt to study the automatic generation control respectively.Accordingly,
of multi-source two-area power system under deregulated envi-
ACE2 ¼ B2 Df 2 þ a12 DPtie12;error ð6Þ
ronment where, each area consists of thermal–gas generating
units. For this study, classical controllers such as integral, The nominal system parameters are given in Appendix B.
proportional-integral, integral-derivative, and proportional- Since, in the considered two-area system, there are two GENCOs
integral-derivative are considered to reveal the performances, and two DISCOs in each area. The corresponding DPM is
where the gains of these controllers are optimised by using GA 2 3
..
and DE algorithms. Further, to investigate the performances of 6 cpf11 cpf12 .cpf13 cpf14 7
proposed controllers, three different cases namely base case, 6 .. 7
6 7
bilateral transaction case and contract violation cases have been 6 cpf21 cpf22 .cpf23 cpf 24 7
6 7
studied. Finally, the dynamic performances obtained both by GA DPM ¼ 6 . . . . . . . . . . . . . . . . . . . . . . . . 7
6
7 ð7Þ
6 . 7
and DE algorithms are compared for the proposed deregulated 6 7
6 cpf31 cpf32 ..cpf33 cpf34 7
AGC system. 4 5
.
cpf41 cpf42 ..cpf43 cpf44

System investigated From, the above equation, the block diagonals of DPM refers to
local demands whereas, off diagonal blocks correspond to the
The AGC system considered is two equal area systems consist- demands of the DISCOs in one area to the GENCOs in another area.
ing of thermal and gas generation units. Each area consists of
two numbers of GENCOs and two numbers of DISCOs. The thermal Design of controllers
area is provided with a single reheat turbine having appropriate
generation rate constraint of 3% per min [19]. The gas generating The proportional integral derivative controller (PID) is the most
unit is considered with a gas turbine whose parameters are popular feedback controller used in the process industries. It is a
adopted from [18], details of which is given in Appendices A and robust and easily understood controller that can provide excellent
B. The transfer function model of two-area thermal–gas system is control performance despite the varied dynamic characteristics of
shown in Fig. 1. As there are more than one GENCO in each area, process plant. As the name suggests, the PID algorithm consists
area control error (ACE) signal has to be distributed amongst them of three basic modes, the proportional mode, the integral and the
in proportion to their participation in the AGC. Coefficients that derivative modes. A proportional controller has the effect of reduc-
distribute ACE to several GENCOs are termed as ‘‘ACE participation ing the rise time, but never eliminates the steady-state error. An
Pn
factors” (apf). Note that i¼1 apfi ¼ 1 where, n is the number of integral control has the effect of eliminating the steady-state error,
GENCOs. A DISCO in each area demands a particular GENCO or but it may make the transient response worse. A derivative control
GENCOs for load power. As there are more than one GENCOs and has the effect of increasing the stability of the system, reducing the
DISCOs in the deregulated structure, a DISCO has freedom to have overshoot, and improving the transient response. Proportional
a contract with any GENCO for transaction of power. A DISCO may integral (PI) controllers are the most often type used today in
have a contract with a GENCO on another control area also. These industry. A control without derivative (D) mode is used when: fast
demands must be reflected in the dynamics of the system. Since, a response of the system is not required, large disturbances and
particular set of GENCOs are required to follow the load demanded noises are present during operation of the process and there are
by a DISCO, information signals must flow from a DISCO to that large transport delays in the system. Derivative mode improves
particular set of GENCOs specifying corresponding demands. This stability of the system and enables increase in proportional gain
is achieved using the concept of DPM i.e., DISCO participation and decrease in integral gain which in turn increases speed of
matrix [9]. DPM helps to visualise the contracts easier. As the name the controller response. PID controller is often used when stability
suggests, DPM shows the participation of DISCO in a contract with and fast response are required. In view of the above, I, PI, ID and
GENCO. In DPM, number of rows is equal to the number of GENCOs PID structured controllers are considered in the present chapter.
and the number of column equal to the number of DISCOs of the Design of PID controller requires determination of the three
system. Thus, each ij entry of the matrix called as ‘‘contract partic- main parameters, Proportional gain ðK P Þ, Integral gain ðK I Þ and
ipation factor” (cpf) corresponds to the fraction of a total load Derivative gain ðK D Þ. Similarly, for PI controller ðK P Þ and Integral
contracted by a DISCO j from a GENCO i. The sum of all entries in gain ðK I Þ are to be determined. For design of ID controller K I and
a column in a matrix is unity. K D are to be determined. The controllers in both the areas are
P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214 207

Fig. 1. Transfer function model of multi-source two-area system under deregulation.

considered to be different, so that proportional gains are K P1 ; K P2 ; where Df 1 and Df 2 are the system frequency deviation in area-1 and
integral gains are K I1 ; K I2 and derivative gains are K D1 ; K D2 . The area-2, respectively; DPTie is the incremental change in tie line
integral square error (ISE) criterion is considered as the objective power and t sim is the time range of simulation. The problem con-
function for the present work which is described in Eq. (8). straints are the controller parameter bounds. Therefore, the design
Z tsim problem can be formulated as the following optimization problem.
2 2
J ¼ ISE ¼ ðDf 1 Þ þ ðDf 2 Þ þ ðDP Tie Þ2  dt ð8Þ Minimize J ð9Þ
0
208 P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214

maximum values of controller parameters are chosen as 0 and


2.0, respectively.

Results and analysis

Implementation of DE algorithm

The model of the system under study has been developed in


MATLAB/SIMULINK environment and DE and GA programs are
written in .m files. Implementation of DE algorithm requires the
determination of six fundamental issues: DE step size function also
called scaling factor (F), crossover probability (CR), the number of
population (NP), initialization, termination and evaluation function.
The flow chart of differential evolution algorithm employed in the
present study is shown in Fig. 2. The scaling factor is a value in the
range (0, 2) that controls the amount of perturbation in the muta-
tion process. Crossover probability (CR) constants are generally
chosen from the interval (0.5, 1). If the parameter is co-related,
then high value of CR works better and the reverse is true for no
correlation. DE offers several variants or strategies for optimization
denoted by DE/x/y/z, where x = vector used to generate mutant
vectors, y = number of difference vectors used in the mutation pro-
cess and z = crossover scheme used in the crossover operation.
Optimization is terminated by the pre-specified number of gener-
ations for DE.
Simulations were conducted on an Intel, core 2 Duo CPU of
2.4 GHz based computer in the MATLAB (R2010a) environment
for DE and GA both. A series of simulation has been performed to
properly tune the DE parameters to reduce the objective function.
Table 1 shows the outcomes of DE parameters variation where, 50
independent runs are performed for each parameter variation.
Based on the results obtained from Table 1, the parameters for fur-
ther simulation studies considered in the present paper are: a pop-
ulation size of NP = 50, generation number of G = 90, step size of
F = 0.7 and crossover probability of CR = 0.3. The strategy employed
Fig. 2. Flow chart for DE algorithm. is DE/best/1/exp. Optimization is terminated when the pre-
specified number of generations for DE is reached. One more
important factor that affects the optimal solution more or less is
Subject to
the range for unknowns. In the very first run of the program, i.e.,
K P min 6 K P 6 K P max ; K I min 6 K I 6 K I max and K D min 6 K D 6 K D max first iteration, a wider solution space is explored and after getting
the initial solution the solution space is shortened nearer to the
ð10Þ
values obtained in the previous iteration. Here, the lower and
where J is the objective function and K P min , K I min ; K P max , K I max and upper bounds of the gains are chosen as 0 and 2, respectively. Sim-
K D min , K D max are the minimum and maximum value of the control ilarly, based on the Ref. [13], the parameters of GA for simulation
parameters. As reported in the literature, the minimum and studies considered in this paper are: a population size of

Table 1
Study of tuning of DE parameters.

Parameters Average Max Min Std Other parameters


NP 10 1.3858 1.7136 1.0803 0.225074 Gen(G) = 90; F = 0.1; CR = 0.9
20 1.37755 1.7485 1.0342 0.234971
30 1.44094 1.9153 1.1125 0.235593
40 1.4976 1.9484 1.0481 0.247296
50 1.4239 1.8391 0.9952 0.210464
F 0.1 1.34788 1.7485 1.0813 0.22114 Gen(G) = 90; NP = 50; CR = 0.9
0.3 1.469253 1.9259 1.0342 0.251433
0.5 1.497607 1.9484 1.0481 0.247296
0.7 1.44342 1.8391 0.9952 0.204003
0.9 1.4144 1.9153 1.1125 0.232506
CR 0.1 1.478147 1.9259 1.0342 0.259239 Gen(G) = 90; NP = 50; F = 0.7
0.3 1.381753 1.6205 1.032 0.169625
0.5 1.37976 1.6359 1.0768 0.183741
0.7 1.537327 1.8691 1.2188 0.221018
0.9 1.344693 1.7136 1.0803 0.217211
P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214 209

NP = 100, generation or iteration number of G = 80, crossover prob- apf1 ¼ 0:5; apf2 ¼ 1  apf1 ¼ 0:5; apf3 ¼ 0:5; apf4 ¼ 0:5:
ability = 0.1 and mutation probability = 0.1.
Assume that the load change occurs only in area 1. Thus, the load is
demanded only by DISCO1 and DISCO2. DISCO3 and DISCO4 do not
Base case demand any load from GENCOs, thus corresponding cpfs in DPM
matrix are zero. Let this load demand for DISCO1 and DISCO2 be
Consider a case where the GENCOs in each area participate 0.1 p.u. MW for each of them. The DPM matrix is given by
equally in AGC; i.e., ACE participation factors are,
2 3
..
Table 2
6 0:5 0:5 . 0 0 7
6 7
6 0:5 0:5 ...0 0 7
Gain values of different controllers.
6 7
Type of controller DE optimised GA optimised 6 7
DPM ¼ 6 . . . . . . . . . . . . 7
I controller KI1 0.3604 0.3341 6 7
6 .. 7
KI2 0.3107 0.3799 6 0 0 . 0 0 7
4 5
PI controller KI1 0.2997 0.4072 ..
KI2 0.1999 0.2799 0 0 .0 0
KP1 0.0341 0.0791
Simulation has been carried out by considering 4% regulation
KP2 0.1226 0.2126
(R), the gains of the controllers are optimised using GA and pro-
ID controller KI1 0.973 0.8883
posed DE algorithms and are given in Table 2. Overshoot, under-
KI2 0.649 0.8922
KD1 0.3673 0.3191 shoot and settling time of Df1, Df2 and DPtie for I, PI, ID and PID
KD2 0.7083 0.5866 controllers are given in Table 3. As seen from this table, the objec-
PID controller KI1 1.3065 0.9819 tive function values are improved with proposed DE optimised
KI2 0.3903 0.7219 PID/ID/PI/I controllers by 42.5%, 37.13%, 2.55% and 0.44%, respec-
KD1 1.0959 0.7899 tively, compared to GA optimised controllers. The overshoot and
KD2 0.4993 0.7163
KP1 0.8495 0.169
KP2 0.0767 0.1727
0.2
0.1
Table 3
0
Undershoot (US), overshoot (OS) and settling time (ST) for base case with different
Δf1

controllers using GA and DE algorithm. -0.1


-0.2 PID controller
Type of controllers Parameters DE optimised GA optimised ID controller
-0.3
I controller Df1 OS 0.2301 0.2242 I controller
US 0.425 0.4252
-0.4 PI controller
ST 17.43 19.43 0 5 10 15 20 25 30
Df2 OS 0.1619 0.1603
US 0.2153 0.2163
Time in sec
ST 18.58 20.59
DPtie OS 0.0164 0.0164 Fig. 3. Frequency deviation of area-1 for base case with GA.
US 0.0694 0.0695
ST 13.49 13.52
OBJ 0.2698 0.271
PI controller Df1 OS 0.2203 0.2488
US 0.4237 0.421 0.1
ST 17.4 19.18
Df2 OS 0.1515 0.1706 0
Δf2

US 0.2137 0.2126
ST 18.53 20.3 PID controller
-0.1
DPtie OS 0.0135 0.0183 ID controller
US 0.0691 0.0677 I controller
ST 13.34 13.3 -0.2 PI controller
OBJ 0.2627 0.2695 0 5 10 15 20 25 30
ID controller Df1 OS 0.1477 0.1559 Time in sec
US 0.3494 0.3576
ST 5.79 6.21
Fig. 4. Frequency deviation of area-2 for base case with GA.
Df2 OS 0.0563 0.0616
US 0.1522 0.1582
ST 5.2 8.34
DPtie OS 0.0085 0.0091 0.02
US 0.0516 0.0533
ST 1.88 1.88 0
OBJ 0.1028 0.1635
ΔPtie

PID controller Df1 OS 0.0409 0.0503 -0.02


US 0.2574 0.2916
PID controller
ST 5.52 6.11 -0.04
ID controller
Df2 OS 0.012 0.0138
I controller
US 0.0946 0.1191 -0.06
PI controller
ST 4.51 6.79
DPtie OS 0.0057 0.0086 0 5 10 15 20 25 30
US 0.0315 0.0397
ST 1.88 2.4
Time in sec
OBJ 0.0548 0.0935
Fig. 5. Change in tie line power for base case with GA.
210 P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214

0.1

0.08 0

0.06 -0.02

ΔPtie
ΔPG1

0.04 PID controller -0.04 PID controller


ID controller ID controller
0.02 I controller I controller
-0.06
PI controller PI controller
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30

Time in sec Time in sec

Fig. 6. Generation of GENCO1 for base case with GA. Fig. 10. Change in tie line power for base case with DE.

0.2
0.1
0.15
0.08
ΔPG2

0.1
0.06

ΔPG1
PID controller
0.05 ID controller 0.04 PID controller
I controller
ID controller
0 PI controller 0.02 I controller
0 5 10 15 20 25 30 PI controller
0
Time in sec 0 5 10 15 20 25 30
Time in sec
Fig. 7. Generation of GENCO2 for base case with GA.
Fig. 11. Generation of GENCO1 for base case with DE.

0.2
0.1
0.2
0
Δf1

-0.1 0.15
-0.2 PID controller
ΔPG2

ID controller 0.1
-0.3 I controller
PID controller
-0.4 PI controller 0.05 ID controller
0 5 10 15 20 25 30 I controller
0 PI controller
Time in sec
0 5 10 15 20 25 30
Fig. 8. Frequency deviation of area-1 for base case with DE.
Time in sec

Fig. 12. Generation of GENCO2 for base case with DE.

0.1

In Figs. 6 and 7, the change in generations, i.e., DPG1 and DPG2 of


0
Δf2

GENCOs for different controllers optimised using GA is shown. It is


PID controller observed from these two figures that both the GENCOs generate
-0.1
ID controller same power. Similarly, dynamic performances of different con-
I controller
trollers with DE algorithm are shown in Figs. 8–10. The change
-0.2 PI controller
in generations, i.e., DPG1 and DPG2 of GENCOs for different con-
0 5 10 15 20 25 30 trollers optimised using DE algorithm is shown in Figs. 11 and
Time in sec 12. In both the cases, generation of GENCOs generate the same
power 0.1 p.u. MW. With regard to the performances of different
Fig. 9. Frequency deviation of area-2 for base case with DE.
controllers it is observed that, PID controller offers lesser value of
objective function and also reduces the settling time of frequency
undershoot of Df1, Df2 and DPtie with PID, PI and ID controllers
deviations of both the control areas and change in tie-line power
optimised with DE are improved as compared to those obtained
than those obtained by I, PI and ID controllers, as shown in Table 3.
by GA technique. However, for I controller, overshoot and under-
So, for further investigation the PID controller is only considered.
shoot are found to be less in GA optimisation compared to DE algo-
rithm. The improvement in settling times for Df1 are found to be
10.29%, 6.76%, 9.28% and 9.66%, respectively, by I, ID, PI and PID Bilateral transaction case
controllers. Similarly, for Df2 these are 9.76%, 37.65%, 8.72% and
33.58% by I, ID, PI and PID controllers, respectively. The improve- In this case, all DISCOs are in contract with all GENCOs for trans-
ment in settling time for DPtie is found to be 21.67% for PID con- action of power. DISCOs contract with the GENCOs as per the fol-
troller. The dynamic performances of I, ID, PI and PID controllers lowing DPM matrix. It is assumed that each DISCOs demand
are shown in Figs. 3–5 using GA optimised values. 0.1 p.u. MW power from GENCOs as defined by cpfs in DPM matrix.
P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214 211

Table 4 Table 5
Gain values of PID controller and speed regulation (Ri) for bilateral transaction and Overshoot, undershoot and settling time by both DE and GA for bilateral transaction
contract violation cases both in DE and GA. and contract violation cases.

For bilateral transaction For contract violation Parameters Bilateral transaction case Contract violation case
optimised optimised
Parameters DE GA DE GA
DE GA DE GA
KI1 1.2251 0.9934 1.3424 0.8693
KI2 1.4309 1.0955 1.4304 0.5797 Df1 OS 0.0988 0.1478 0.1093 0.1177
KD1 0.990 0.9509 1.0421 0.5499 US 0.4664 0.4949 0.6192 0.7542
KD2 0.8277 0.947 0.919 0.145 ST 10.87 12.5 11.71 17.19
KP1 0.7795 0.2963 1.1908 0.853
Df2 OS 0.0683 0.1036 0.0641 0.0474
KP2 0.9013 0.7447 1.0861 0.6221
US 0.2158 0.2076 0.2089 0.3856
R1 3.0784 3.8896 3.1743 3.5095
ST 11.47 12.64 12.49 17.43
R2 4.9809 6.8678 8.1454 5.1325
DPtie OS 0.05 0.05 0.05 0.05
US 0.0186 0.0309 0.015 0.0449
ST 3.08 3.2 3.47 4.25
0.1
OBJ 0.4945 0.7348 0.7605 1.2967
0
-0.1
Δf1

-0.2
-0.3
-0.4 DE tuned PID controller
0.1
GA tuned PID controller
-0.5

ΔPG1
0 5 10 15 20 25 30
0.05
Time in sec

Fig. 13. Frequency deviation of area-1 for bilateral transaction case. DE tuned PID controller
0 GA tuned PID controller
0.1 0 5 10 15 20 25 30
0.05 Time in sec
0
Fig. 16. Generation of GENCO1 for bilateral transaction case.
Δf2

-0.05
-0.1
-0.15
DE tuned PID controller DE tuned PID controller
-0.2 GA tuned PID controller GA tuned PID controller
0 5 10 15 20 25 30 0.1
ΔPG2

Time in sec
0.05
Fig. 14. Frequency deviation of area-2 for bilateral transaction case.

DE tuned PID controller 0


0.04
GA tuned PID controller 0 5 10 15 20 25 30

0.02 Time in sec


ΔPtie

Fig. 17. Generation of GENCO2 for bilateral transaction case.


0

-0.02

0 5 10 15 20 25 30 0.15
Time in sec
ΔPG3

0.1
Fig. 15. Change in tie line power for bilateral transaction case.

2 3 0.05
..
6 0:5 0:25 0:3 7.0 DE tuned PID controller
6 7
.. GA tuned PID controller
6 0:2 0:25 0 7 .0
0
6 7 0 5 10 15 20 25 30
6 7
DPM ¼ 6 . . . . . . . . . . . . 7 Time in sec
6 7
6 . 7
6 0 0:25 ..1 0:7 7
4 5 Fig. 18. Generation of GENCO3 for bilateral transaction case.
.
0:3 0:25 ..0 0
Let each GENCO participates in AGC as defined by following apfs: With the above apf values, GENCOs participate in AGC. The apfs
only affect the transient behaviour of the system, not the steady
apf1 ¼ 0:75; apf2 ¼ 1  apf1 ¼ 0:25; apf3 ¼ 0:5;
state behaviour. The optimum gain values, speed regulation (Ri)
apf4 ¼ 1  apf3 ¼ 0:5 for the PID controller are obtained using GA and DE optimisation
212 P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214

0.12 DE tuned PID controller DE tuned PID controller


0.04
GA tuned PID controller GA tuned PID controller
0.1
0.08 0.02
ΔPG4

ΔPtie
0.06 0
0.04
-0.02
0.02
0 -0.04
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time in sec Time in sec
Fig. 19. Generation of GENCO4 for bilateral transaction case. Fig. 22. Change in tie line power for contract violation case.

algorithms and the values are given in Table 4. Dynamic perfor-


0.2
mances of PID controller using GA and DE algorithms are shown
in Figs. 13–15. From these figures, the overshoot of Df1 and Df2 0.15
are found to be improved by 33.14% and 34.07% using DE algorithm

ΔPG1
as compared to GA. Overshoot, undershoot and settling time 0.1
obtained for PID controller using both GA and DE algorithms are
0.05
given by Table 5. As given in this table, the settling times of Df1, DE tuned PID controller
Df2 and DPtie are improved by 13.04%, 9.26% and 3.75%, respec- 0 GA tuned PID controller
tively, using proposed DE tuned controller compared to GA tech- 0 5 10 15 20 25 30
nique. Also, the objective function is improved by 32.7% using DE
Time in sec
algorithm compared to GA. Figs. 16–19 shows the generations of
GENCOs for bilateral transaction case using DE and GA algorithms Fig. 23. Generation of GENCO1 for contract violation case.
with PID controller.

0.2 DE tuned PID controller


Contract violation case
GA tuned PID controller
0.15
DISCO may demands more power than that of the specified con-
ΔPG2

tract. The excess of power must be supplied by the GENCOs of the 0.1
same area as the DISCOs. Let us consider DISCO1 demands 0.1 p.u.
MW of excess power, the extra power reflects as local load of the 0.05
area. So the local load of area-1 is
0
DPL1;loc ¼ load of DISCO1 ð0:1Þ þ load of DISCO2 ð0:1Þ þ 0:1 0 5 10 15 20 25 30
¼ 0:3 p:u: MW Time in sec

Fig. 24. Generation of GENCO2 for contract violation case.

0
0.2
-0.2
Δf1

0.15
-0.4
ΔPG3

0.1
-0.6 DE tuned PID controller
GA tuned PID controller
0.05
0 5 10 15 20 25 30 DE tuned PID controller
GA tuned PID controller
Time in sec 0
0 5 10 15 20 25 30
Fig. 20. Frequency deviation of area-1 for contract violation case. Time in sec

Fig. 25. Generation of GENCO3 for contract violation case.

0.15
0 DE tuned PID controller
GA tuned PID controller

-0.1 0.1
ΔPG4
Δf2

-0.2
0.05
-0.3
DE tuned PID controller
GA tuned PID controller
-0.4 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time in sec Time in sec

Fig. 21. Frequency deviation of area-2 for contract violation case. Fig. 26. Generation of GENCO4 for contract violation case.
P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214 213

-0.2
Δf1

-0.4

-0.6 0.3 C.V

-0.8 0.4 C.V


0.5 C.V Fig. A1. Load frequency model of gas turbine power plant.

0 5 10 15 20 25 30
Time in sec unit considering reheat turbine and appropriate value of GRC is
taken. Performances of different controllers are compared with
Fig. 27. Frequency deviation of area-1 for different values of contract violation.
GA and DE algorithm for base case. The controller parameters are
optimized using differential evolution (DE) optimization tech-
nique. Initially the control parameters of DE algorithm are tuned
0.05
by carrying out multiple runs of algorithm for each control param-
0 eter variation. The best DE parameters are found to be: step size
-0.05 F = 0.3, crossover probability of CR = 0.7, Population size of
-0.1
Δf2

NP = 50 and Generation of G = 90. The parameters of integral (I),


-0.15 integral derivative (ID) and proportional integral derivative (PID)
-0.2 0.3 C.V are optimized employing tuned DE algorithm and GA. The superi-
0.4 C.V
-0.25 ority of the proposed approach has been shown by comparing the
0.5 C.V
results with GA technique for the same power systems by using
0 5 10 15 20 25 30
various performance measures like overshoot, settling time and
Time in sec standard error criteria of frequency and tie-line power deviation
for base case. The critical study of the dynamic responses reveals
Fig. 28. Frequency deviation of area-2 for different values of contract violation.
that PID controller is superior keeping in view of settling time
and reduced oscillations than other controllers. Controller gains
and speed regulation (Ri) parameters are optimised using DE and
0.06 GA algorithms for bilateral transaction and contract violation cases
0.5 C.V
0.04
also. For contract violation case, higher values of (Ri) is obtained in
0.4 C.V
0.3 C.V case DE, which results into economical governor. With references
0.02
to obtained values settling time, overshoot and objective function
ΔPtie

0 values by both the algorithms, DE is found to perform better than


-0.02 GA for all the three cases, i.e., base case, bilateral transaction and
contract violation case. Furthermore, it is also observed that the
-0.04
proposed system is robust and is not affected by change in the con-
-0.06 tact violation condition, system parameters and size of contract
0 5 10 15 20 25 30
violation.
Time in sec

Fig. 29. Change in tie power for different values of contract violation. Appendix A. Modelling of gas turbine power plant

The local load of area-2 remains same as the second case i.e., A gas turbine power plant usually consists of valve positioner,
0.2 p.u. MW. The DPM matrix remains same as the second case. speed governor, fuel system & combustor and gas turbine. The
The power generation of area-2 i.e., GENCO3 and GENCO4 remains load-frequency model of gas turbine power plant is shown in
same as before. The un-contracted load of DISCO1 is reflected in Fig. A1, where DPCGref is reference power setting of the gas plant
generation of GENCO1 and GENCO2. The dynamic performances and DPGT is the gas turbine output power. The system frequency
of PID controller using GA and DE algorithms are given in Figs. 20– deviation and governor speed regulation parameters are repre-
22. Overshoot of Df1 is improved by 7.45% using DE technique sented by Df in p.u. and R2 in Hz/p.u. MW respectively. The transfer
compared to GA. Generation of all GENCOs are shown in Figs. 23– function representation of valve positioner is shown in Fig. A1,
26 when computed by both in GA and DE algorithms. Overshoot, where, cg is the gas turbine valve positioner, bg is the gas turbine
undershoot and settling time obtained by DE and GA algorithms constant of valve positioner. The speed governing system is repre-
for PID controller is also given in Table 5. As given in this table, sented by a lead-lag compensator as shown in Fig. A1, where, XG is
the settling times of Df1, Df2 and DPtie are found to be improved the lead time constant of gas turbine speed governor in sec, YG is
by 31.88%, 28.38% and 18.35% with proposed technique as com- the lag time constant of gas turbine speed governor in sec. the fuel
pared to GA. To show the robustness of proposed controllers, con- system and combustor is represented by a transfer function with
tract violation (C.V) is increased from 10% to 30% in steps of 10% appropriate time constants as shown in Fig. A1, where, TF is the
and the dynamic responses are shown in Figs. 27–29 from which gas turbine fuel time constant in sec and TCR is the gas turbine com-
it is clear that the designed controllers are robust and perform sat- bustion reaction time delay in sec. The gas turbine is represented
isfactorily for different contract violation. by a transfer function, consisting of a single time constant i.e. the
gas turbine compressor discharge volume-time constant (TCD) in
Conclusion sec.

AGC of interconnected multi-source two-area system under Appendix B


deregulated environment is considered in this chapter. Thermal
and gas generation units are considered in the two area. Thermal Nominal parameters of the system investigated are:
214 P.K. Hota, B. Mohanty / Electrical Power and Energy Systems 75 (2016) 205–214

B1 ; B2 ¼ 0:4312 p:u: MW=Hz; R1 ¼ R2 2:4 Hz=p:u:; [9] Donde V, Pai MA. Simulation and optimization in an AGC system after
deregulation. IEEE Trans Power Syst 2001;16(3):481–8.
T G1 ¼ 0:06 s; T t1 ¼ 0:3 s; K r1 ¼ 0:3; T r1 ¼ 10 s; [10] Demiroren A, Zeynelgil HL. GA application to optimization of AGC in three-
area power system after deregulation. Electr Power Energy Syst
2007;29:230–40.
K PS1 ¼ K PS2 ¼ 120 Hz=p:u: MW; T P1 ¼ T P2 ¼ 20 s; [11] Debbarma S, Saikia LC, Sinha N. AGC of a multi-area thermal system under
T 12 ¼ 0:2712; a12 ¼ 1; X G ¼ 0:6 s; Y G ¼ 1:1 s; cg ¼ 1; deregulated environment using a non-integer controller. Electr Power Syst Res
2013;95:175–83.
bg ¼ :049 s; T F ¼ 0:239 s; T CR ¼ 0:01 s; T CD ¼ 0:2 s; [12] Tyagi B, Srivastava SC. A decentralized automatic generation control scheme
for competitive electricity markets. IEEE Trans Power Syst 2006;21:312–9.
[13] Ghoshal SP, Goswami SK. Application of GA based optimal integral gains in
References fuzzy based active power frequency control of non reheat and reheat thermal
generating systems. Electr Power Syst Res 2003;67:79–88.
[1] Chritie RD, Bose A. Load frequency control issues in power system operation [14] Ghoshal SP. Application of GA/GA-SA based fuzzy automatic generation
after deregulation. IEEE Trans Power Syst 1996;11(3):1191–200. control of a multi-area thermal generating system. Electr Power Syst Res
[2] Bakken BH, Grande OS. Automatic generation control in a deregulated power 2004;70:115–27.
system. IEEE Trans Power Syst 1998;13(4):1401–6. [15] Stron R, Price K. Differential evolution – a simple and efficient adaptive scheme
[3] Meliopoulos APS, Cokkinides GJ, Bakirtzis AG. Load-frequency control service for global optimization over continuous spaces. J Glob Optim 1995;11:341–59.
in a deregulated environment. Decis Supp Syst 1999;24:243–50. [16] Panda S. Differential evolution algorithm for SSSC-based damping controller
[4] Bevrani H. Reduced l-based load frequency controller in a deregulated power design considering time delay. J Franklin Inst 2011;348(8):1903–26.
system environment. In: Proc of 14th Int Power Syst Conf. Iran; 1999. p. 1–9. [17] Panda S. Robust coordinated design of multiple and multi-type damping
[5] Bevrani H. Robust load frequency controller in a deregulated environment: a controller using differential evolution algorithm. Electr Power Energy Syst
l-synthesis approach. In: Proc IEEE Int Conf Control Appl; 1999. p. 616–21. 2011;33:1018–30.
[6] Bevrani H, Teshnehlab M, Bevrani H. Load frequency controller design in a [18] Singh Parmar KP, Majhi S, Kothari DP. Load frequency control of a realistic
deregulated environment using flexible neural networks. In: Proc 15th Int power system with multi-source power generation. Electr Power Energy Syst
Power Syst Conf; 2000. p. 1–6. 2012;42:426–33.
[7] Kumar J, Ng Kah-Hoe, Sheble G. AGC simulator for price-based operation: Part- [19] Hassan Bevrani. Robust power system frequency control. New York: Springer;
I. IEEE Trans Power Syst 1997;12(2):527–32. 2009. pp. 15–31.
[8] Kumar J, Ng Kah-Hoe, Sheble G. AGC simulator for price-based operation: Part-
II. IEEE Trans Power Syst 1997;12(2):533–8.

You might also like