G Model

ASOC 3686 1–9 ARTICLE IN PRESS

Applied Soft Computing xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Applied Soft Computing

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

1 Multi area AGC scheme using imperialist competition algorithm

2 in restructured power system
3 Q1 Nagendra Kumar ∗ , Vishal Kumar, Barjeev Tyagi
4 Q2 Department of Electrical Engineering,Indian Institute of Technology, Roorkee, India
5

6
20 a r t i c l e i n f o a b s t r a c t
7
8 Article history: This paper is focused on optimization based design methodology and application of PID controller in
9 Received 28 May 2014 restructured, competitive electricity market environment, for AGC problem. The paper compares two
10 Received in revised form 24 October 2015 search algorithms for designing of PID controller used for AGC in multiarea power system. The optimal
11 Accepted 3 July 2016
parameters of PID controller have been determined with the use of Imperialist Competitive Algorithm
12 Available online xxx
(ICA). A deregulated scenario has been considered to develop the model of the multiarea AGC scheme. This
13
paper presents that the ICA tuned PID (ICA-PID) controller can optimally regulate the generators output
14 Keywords:
and can provide the best dynamic response of frequency and tie-line power on a load perturbation.
15 Automatic generation control
16 Deregulation
The performance of proposed controller has been checked on 2-area thermal power system and 3-area
17 Imperialistic competition introduction thermal-hydro power system with the consideration of generation rate constraint (GRC). The results
18 PID controller obtained by ICA-PID controller and genetic algorithm tuned (GA-PID) controller have been compared on
19 Optimization the basis of performance parameters (settling time and oscillations). It is seen that ICA-PID controller
shows the better performance as compared to GA-PID controller.
© 2016 Published by Elsevier B.V.

21 1. Introduction Various load frequency control issues related to deregulation is 39

reported in [9]. Q4 40

22 AGC is used to maintain zero steady state errors in deviation of The increasing complexities in power system encouraged 41

23 system frequency and tie-line power exchange to other areas. The researchers to explore appropriate control algorithms for AGC. Var- 42

24 conventional AGC scheme for isolated and interconnected power ious control strategies, such as optimal control [10,11], variable 43

25 system has been discussed in [1–3]. For the improvement in the effi- structure control [12,13], adaptive and self-tuning control [14,15], 44

26 ciency of the operation of power system, it has been restructured robust control [16–18] etc. have been discussed in the literature for 45

27 according to open market system, which consists Gencos (gen- AGC scheme. PID controller has a simple structure and widely used 46

28 eration companies), Transcos (transmission companies), Discos in AGC scheme to control the frequency and the tie-line power. The 47

29 (distribution companies), and ISO (independent system operator). literature reveals that the different methods have been proposed to 48

30 ISO is an independent agent which procures various ancillary ser- determine the parameters of a PID controller. A variety of optimiza- 49

31 vices for stable and secure operation of a power system [4,5]. AGC tion algorithm have been reported to find the optimal parameters 50

32 is one of the most important ancillary services among all. The main of PID controller such as Fuzzy logic (FL) [19–24], genetic algorithm 51

33 goal of this ancillary service is to provide a balance between gen- (GA) [25,26], bacterial foraging [27,28], particle swarm optimiza- 52

34 eration and load demand of each area and maintain the frequency tion (PSO) [29,30] etc. In addition to these, ICA may be used as an 53

35 and tie-line power flow within the specified limit. In deregulated alternate tuning method due to its global search and stable con- 54

36 power system, poolco-based transactions, bilateral transactions, vergence characteristics. ICA is inspired by the socio political idea 55

37 and a combination of poolco and bilateral i.e. mixed transactions [31]. In recent years, ICA has been successfully applied to the dif- 56

38 are used for various contracts between Discos and Gencos [6–8]. ferent control processes or optimization problems, some of them 57

are the economic dispatch problem [32], power system stabilizer 58

[33], oil flow rate [34], reactive power dispatch [35] and tuning of 59

PID controller [36–39]. 60

∗ Corresponding author.
According to the recent literature available, the controller 61
E-mail addresses: [email protected], [email protected]
designed using ICA has capability of handling parameter 62
(N. Kumar).

http://dx.doi.org/10.1016/j.asoc.2016.07.005
1568-4946/© 2016 Published by Elsevier B.V.

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
 G Model
ASOC 3686 1–9 ARTICLE IN PRESS
2 N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx

63 uncertainty, possess robustness and eliminate steady state errors and Gg and Gt represents the transfer function model of governor 110

64 efficiently. It also gives better stability [33–39]. The versatility and and turbine respectively. Pg1 , Pg2 , ......Pgk , ......Pgn repre- 111

65 efficient performance of this technique in different areas motivated sents the change in the output of Gencos in area-i. Fig. 2 shows 112

66 us to investigate its implementation to the problem of AGC under the overall block diagram representation for the AGC scheme of 113

67 deregulated environment. It is seen that ICA-PID controller works area-i 114

68 well, provides stability and ensured good dynamic performance of

69 AGC in restructured environment. 3. PID controller design using ICA 115
70 In present study, ICA algorithm is used to determine the optimal
71 parameters of a PID controller for multiarea AGC scheme in dereg- 3.1. PID controller 116
72 ulated environment. Poolco and bilateral transactions have been
73 considered between the areas. Disco participation matrix (DPM) is PID controllers are the most widely used industrial controllers 117
74 used for implementation of various contracts. Mean square error because of simple and easy implementation. The structure of a PID 118
75 (MSE) of area control error (ACE) is considered as a performance controller can be expressed as the sum of three terms, proportional, 119
76 index (objective function) for this optimization problem. The pro- integral, and derivative control. The transfer function of such a PID 120
77 posed controller is tested on 2-area thermal power system [6], and controller can be expressed as: 121
78 3-area thermal-hydro power system considering GRC [40]. The per-
79 formance of ICA-PID controller has been compared with GA-PID KI
GPID (s) = KP + + KD s (4) 122
80 controller. The convergence characteristic of ICA shows the higher s
81 convergence rate than GA. Results show that ICA-PID controllers where, KP , KI , KD are the proportional, integral and derivative gains 123
82 have reduced settling time and oscillations in dynamic response of of the controller. For the good performance of the system, these 124
83 frequency, tie-line power and change in generation as compared to gains should be determined optimally. In this study, ICA is utilized 125
84 PID controllers optimized by GA. to determine the optimal parameters of PID controller for each area. 126

It is necessary to use a proper objective function for optimal tuning 127

85 2. System modeling of controller parameters using evolutionary algorithms. The opti- 128

mal values of gains of controller are obtained by minimizing the 129

86 Any mismatch between real power generation and load considered objective function. In this paper mean square error of 130
87 demand, give rise to area control error (ACE) [7]. Coefficients that ACE is taken as the optimization function which can be formulated 131
88 distribute ACE to several Gencos are known as ACE participa- in the following manner, 132
89 tion factors (˛pf ). Power system satisfactory operation requires Minimize mean square of area control error 133
90 the removal or minimization of ACE. In a deregulated electricity
1 1
n n
91 market different transactions can take place such as poolco based
2
92 transaction, bilateral transaction and the combination of these f = [(ACEi )2 ] = J = [(Bi Fi + Ptiei error ) ] (5) 134
n n
93 two transactions [8]. Poolco based transaction does not change i=1 i=1
94 the scheduled tie-line exchange, but the bilateral transaction can
Here, f is used to determine the optimum gains of the PID controller 135
95 change the tie-line exchange which can modify the ACE.
for step load disturbance in different areas with the consideration 136
96Q5 The net tie-line power flow from an area-i can be represented
of the following constraints 137
97 as given below,
n n KP,i min ≤ KP,i ≤ KP,i max
98 Ptiei−schd = Ptiei +  Dij −  Dji (1)
j=1 j=1 KI,i min ≤ KI,i ≤ KI,i max (6) 138
i=
/ j i=
/ j
KD,i min ≤ KD,i ≤ KD,i max

99 where, Dij and Dji represent the demand of the Disco of area-j and where, KP,i min , KI,i min , KD,i min and KP,i max ,KI,i max , KD,i max are the 139

100 area-i from the Genco of area-i and area-j respectively, Ptiei shows lower bounds and upper bounds of the PID controller for area-i. The 140

101 the change in tie-line power when no bilateral transaction is con- ACE minimization problem has been solved using the ICA which is 141

102 sidered. n represents the number of areas. The error in the tie-line explained in the following section. 142

103 power flow can be represented as,

3.2. Imperialist competitive algorithm (ICA) 143
104 Ptiei error = Ptiei actual − Ptiei schd (2)

105 Therefore, the modified ACE signal can be represented as, ICA was proposed by Esmail Atasphaz- Gargari for optimiza- 144

tion [31]. It starts with an initial population known as country, 145

106 ACEi = Bi fi + Ptiei error (3)
which is further divided in colony and states to make empires. After 146

107 where, Bi is the frequency bias factor and fi is the frequency devi- forming empires, imperialistic competition takes place among all 147

108 ation in area-i. Fig. 1 represents the block diagram of the kth Genco the empires. During this competition weak empires will lose their 148

109 in area-i. The pf is the Genco participation factor, Ri is the droop, power and collapse and at the end of competition only one empire 149

will exists. The details of ICA are well explained in the references 150

[31–39]. 151

The major steps involved to determine the optimal PID param- 152

eters using ICA are given below. 153

3.2.1. Initialize the empires 154

In this step, first an array of variables (country) which is to be 155

optimized is determined as, 156

Fig. 1. Block diagram of Genco-k of area-i. Country = [p1 , p2 , p3 , .....pN ] (7) 157

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
 G Model
ASOC 3686 1–9 ARTICLE IN PRESS
N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx 3

Fig. 2. AGC block diagram for area-i.

158 where, pi is the set of variables to be optimized. Now the cost

159 of a country by evaluating the cost function at the variables
160 [p1 , p2 , p3 , .....pN ]is determined as,

161 cos t = f (country) = f (p1 , p2 , p3 , ...pN ) (8)

162 To form the initial empires, colonies are divided among imperi-
163 alist uniformly using the normalized cost of an imperialist, which
164 can be defined as

165 Cn = cn − max(ci ) (9)

i
Fig. 3. Movement of colonies toward their relevant imperialist in a randomly devi-
166 where, cn is the cost of the nth imperialist and Cn is its normalized ated direction.
167 cost. Now the normalized power (portion of colonies that should
168 be possessed by that imperialist) of each imperialist is determined 3.2.4. Convergence 194
169 as, At conversion point all empires collapse expect the most pow- 195

Cn erful one. At this point all colonies have same positions and same 196
170 Pn = | | (10)
Nimp costs. At this condition imperialistic competition ends and the algo-

197

Ci rithm stop. The steps of ICA are shown in Fig. 4. 198

i=1
4. Results and discussion 199
171 where, Nimp is the most powerful countries to form empires. Finally
172 the initial number of colonies of the nth empire can be determined The performance of the proposed ICA–PID controller has been 200
173 as, checked on 2-area [6] and 3-area [40] power systems. The simu- 201

lation for all the cases has been performed in MATLAB/Simulink 202
174 N × Cn = round(Pn × Ncol ) (11)
R2011b environment. 203

175 where, N × Cn is the initial number of colonies of the nth empire and
176 Ncol is the number of initial colonies. 4.1. Two area system 204

177 3.2.2. Move the colonies towards their relevant imperialist To check the performance of the ICA-PID controller, two area 205

178 Each imperialist’s colonies are assimilated to their respective AGC scheme based on Fig. 2 is considered. Both the areas have two 206

179 imperialist given as, Gencos and two Discos respectively. (Genco11 , Genco12 , Disco11 207

and Disco12 (area-1), and Genco21 , Genco22 , Disco21 and Disco22 208
new old old
180 xcol = xcol + ˇ × x ⊗ (ximp − xcol ) (12) (area-2)). The parameters considered in AGC scheme is given in 209

old Table 1. The optimal parameters of the controller have been deter- 210
181 where, xcol
represents the old position of colony, ˇ represents the
mined using the ICA and GA. The parameters of GA and ICA are given 211
182 assimilation factor, x is a random variable with uniformly distri-
new represent the position of imperialist and new in Tables 2 and 3 respectively. The optimal values of parameters of 212
183 bution. ximp , xcol
PID controller determined Fig. 5 compare the convergence rate of 213
184 position of colony respectively. Fig. 3 shows the movement of
ICA and GA for 2-area system. It is clear from this figure that the 214
185 colonies toward their relevant imperialist. In Fig. 3, d is the distance
186 between colony and imperialist,  is a parameter that adjusts the
187 deviation from the original direction and  is a random number Table 1
Two area power system parameters. Q6
188 added to the direction of movement.
Tgi = 0.08 s Tpi = 24 s
Governor time constant Power system time constant
189 3.2.3. Imperialist competition
Kpi = 120 Hz/pu MW Power Tti = 0.3 s
190 The imperialistic competition decreases the power of weaker system gain constant Turbine time constant
191 empires and increases the power of more powerful empires. Pow- Ri = 2.4 Bi =0.425
192 erless empires will collapse in imperialistic competition and their Speed regulation Frequency bias constant
193 colonies will be divided among other empires. Tij =0.0707, Synchronizing constant

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
G Model
ASOC 3686 1–9 ARTICLE IN PRESS
4 N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx

Table 2
GA Parameters.

Value

Population size 20
Fitness scaling rank
Elite count 2
Crossover scattered
Selection Stochastic uniform

Table 3
ICA parameters.

Parameters Value

Number of countries 30
Number of Decades 100
Number of initial imperialist 2
Assimilation Coefficient (␤) 2
Assimilation Angle Coefficient (␥) 0.5

convergence of ICA is faster than GA. ICA algorithm converges in 20 215

generations while GA converges after 30 generations . 216

4.1.1. Case study 217

In this case, 0.2 pu increase load demand has been consid- 218

ered in area-1 (0.1 pu in Disco11 and 0.1 pu in Disco12 ). To meet 219

this demand Disco11 and Disco12 have contract with Genco11 and 220

Genco12 . Each Genco has ACE participation factors, given as, ˛pf11 = 221

0.5,˛pf12 = 0.5,˛pf21 = 0.5, and ˛pf22 = 0.5. The contract of Dis- 222
223
cos with Gencos is represented by following DPM .
⎡ ⎤
0.5 0.5 0 0
⎢ 0.5 0.5 0 0 ⎥
DPM = ⎢ ⎥
⎣ 0 0 0 0⎦
224

0 0 0 0

A step increase in the demand of each Discos of area-1 has been 225

considered at t = 0. The frequency deviation in area-1 and area-2 are 226

given in Fig. 6(a). Fig. 6(b) shows deviation in the tie-line power. 227

The results given in Fig. 6(a) show that frequency deviation in each 228

area goes to zero at steady state. Since Disco11 and Disco12 in area- 229

1, have increase in load demand, the transient dip in frequency 230

of area-1 is larger than that of area-2. The actual power through 231

tie-line goes to zero as shown in Fig. 6(b). At steady state, change 232

in generation of all Gencos must match the demand of Discos. This 233
Fig. 4. Flowchart of the ICA algorithm.
desired generation of a Genco can be expressed in terms of contract 234

participation factors and the total demand of Discos as given below, 235


Pgi = cpfij PLj (13) 236

Convergence curve j
1
where, PLj is the total load demand of Disco j and cpfij are elements 237

of DPM. 238
Value of objecve funcon

0.8
Pgi = cpfi1 PL1 + cpfi2 PL2 + cpfi3 PL3 + cpfi4 PL4 (14) 239
ICA
0.6 GA For the case under consideration, 240

Pg11 = 0.5 × PL1 + 0.5 × PL2 = 0.1 pu (15) 241

0.4

0.2 Table 4
Optimum values for PID controllers.

ICA-PID
0
0 10 20 30 40 50 60
Two area KP KI KD
No. of generaons
−2 −0.0092 −2
Case-1 GA-PID
Fig. 5. Comparison of convergence characteristics of ICA with GA.
KP KI KD
−1.968 −0.014 −1.999

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
 G Model
ASOC 3686 1–9 ARTICLE IN PRESS
N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx 5

-3 -3
x 10 Frequency deviaons (Area-1) x 10 Freqeuncy deviaons (Area-2)
1
0
0

-1

Delta F2
Delta F1
-5 ICA-PID
ICA-PID GA-PID
-2
GA-PID
-10
-3

-4
0 5 10 15 0 5 Time (s) 10 15
Time (s)

(a). Frequency deviations (rad/s).

-4
x 10 Tie-line power deviaons
5

0
Delta Pe

-5
ICA-PID
-10 GA-PID

-15

-20
0 5 Time (s) 10 15

(b). Tie line power (pu).

Change in generaon (Area-1) (G11) Change in generaon (Area-1) (G12)

0.15 0.15
Delta Pg11

0.1
Delta Pg12

0.1

GA-PID GA-PID
0.05 0.05 ICA-PID
ICA-PID

0 0
0 1 2 3 4 5 0 1 2 3 4 5
Time (s) Time (s)

-3 -3
x 10 Change in generaon (Area-2) (G21) x 10 Change in generaon (Area-2) (G22)
5 4
2
0
0
Delta Pg22
Delta Pg21

-2
GA-PID -4
GA-PID
-5 ICA-PID -6
ICA-PID
-8

-10 -10
0 2 4 6 8 10 0 2 4 6 8 10
Time (s) Time (s)

(c). Generated power (pu ).

Fig. 6. (a) Frequency deviations (rad/s). (b) Tie line power (pu). (c) Generated power (pu).

242 Similarly, Pg12 = 0.1 pu, Pg21 = Pg22 = 0 pu. Fig. 6(c) shows that the tie-line power, frequency deviations, and change in gen- 247

243 that real power output of each Genco in area-1 settled to 0.1 pu erations are stabilized at steady state with lesser settling time and 248

244 value at steady state. Since Genco21 and Genco22 of area-2 are not reduced oscillation using ICA-PID than GA-PID controller. It is seen 249

245 contracted by any Disco for power transaction, their real power out- from the Tables 5 and 6, that ICA-PID controller is less oscillatory 250

246 put settled at zero at steady state, as shown in Fig. 6(c). It is observed and faster than GA-PID controller. 251

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
 G Model
ASOC 3686 1–9 ARTICLE IN PRESS
6 N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx

Table 5 Convergence curve

Comparative Study of Settling Time (Two area power system). 0.025

Value of objecve funcon

Settling Time (second)
0.02
Controller f1 f2 Ptie ICA
ICA-PID 10 12 12 GA
0.015
GA-PID 13 14 15

0.01

Table 6 0.005
Comparative Study of Peak Overshoot (Two area power system).

Case 1 Peak Overshoot (pu) 0

0 20 40 60 80 100
Controller f1 f2 Ptie No. of generaons
ICA-PID −0.0115 −0.00275 −0.001225
GA-PID −0.0139 −0.00374 −0.001655 Fig. 7. Comparison of convergence characteristics of ICA with GA.

4.2.1. Case 1 272

Table 7 Let us consider a case where each Disco has a demand of 0.01 273
Three area power system parameters. pu. Each Disco has a contract of power transaction with each Genco 274
275
Tgi = 0.08 s Tpi = 20 s
of the system as per the given DPM.
governor time constant power system time constant ⎛ ⎞
0.1 0.2 0.3 0.4
Kpi = 120 Hz/pu MW Tti = 0.3 s
⎜ 0.3 0.3 0.2 0.2 ⎟
DPM = ⎝
0.3 ⎠
power system gain constant turbine time constant 276
Ri = 2.4 Bi =0.425 0.2 0.3 0.2
speed regulation frequency bias constant 0.4 0.2 0.3 0.1
Tij = 0.545 Tri = 10 s
synchronizing constant Reheat turbine time constant ˛pf11 = ˛pf12 = 0.5, are the ACE participation factors consid- 277

ered for this case. At steady state each Genco must generate real 278

power as given in (14). Therefore, the net change in the real power 279

generation of Gencos of area-1 can be determine as,Pg 11 = (0.1 + 280

0.2 + 0.3 + 0.4) × 0.01 = 0.01 pu value.Dynamic response compar- 281

252 4.2. Three area system
ison for ICA-PID and GA-PID is shown in Fig. 8. Fig. 8(a) shows the 282

frequency deviations of area-1 to area-3, which settle to zero value 283

253 To check the performance of the ICA-PID controller, a 3-area
at steady state. The enlarged portion of frequency deviation figure 284
254 power system with reheat turbine and generation rate constraint
of area-1 show the difference between the responses obtained with 285
255 (GRC) of 3% per minute has also been considered [40]. The sys-
ICA-PID and GA-PID controller. Fig. 8(b) shows that tie-line power 286
256 tem has two Discos and two Gencos in area-1 (Genco11 , Genco12 ,
deviation between different areas settled down to their desired val- 287
257 Disco11 , and Disco12 ), one Disco and one Genco (G2 , D2 , and G3 ,
ues with lesser settling time and reduced oscillation using ICA-PID 288
258 D3 ) in area-2 and area-3 respectively. The parameters of GA and
than GA-PID controller. From Fig. 8(c), it is clearly seen that Gencos 289
259 ICA are same as given in Tables 2 and 3 respectively. The nominal
of area-1 to area-3 settled their real power output to their desired 290
260 parameters for 3-area system are given in Table 7.
values smoothly using ICA-PID in comparison to GA-PID controller. 291
261 Two cases are considered in the 3-area power system. In the
262 first case, load changes occur in each area whereas, in the second
4.2.2. Case 2: contract violation 292
263 case a contract violation occurs where Disco of area-1 violates the
A Disco may violates a contract, by demanding additional power 293
264 contract by demanding additional power than the contract. The
than the contract. This additional power may force the frequency 294
265 optimal values of parameters of PID controller determined using
to deviate from the nominal value. To keep the power system oper- 295
266 ICA and GA are given in Table 8.
ational at nominal values, Gencos of the same area must supply 296
267 Fig. 7 compare the convergence of ICA with GA, applied to deter-
this additional demand. On consider a case where Disco11 of area- 297
268 mine PID parameters for 3-area power system for the both cases. It
1 demands additional power of 0.01 pu. Therefore the total load 298
269 is clear from Fig. 7 that for 3-area system the convergence of ICA is
demand in area-1 will be the contracted load of Disco11 + additional 299
270 faster than GA. ICA converges in 45 generations while GA converges
load of Disco11 + load of Disco12 = (0.01 + 0.01) + 0.01 = 0.03 pu. This 300
271 in 65 generations.
additional load of area-1 is reflected in the generation of Genco11 301

and Genco12 of area-1. Since Discos of area-2 and area-3 demanding 302

the same load given in the contract, the power generation in area-2 303

and area-3 remains the same as the case 1, given in Section 4.2.1. 304
Table 8 The dynamic response obtained for frequency deviations, tie-line 305
Optimum Values for PID Controllers (Three area power system).
power flow and change in generation are shown in Fig. 9. 306

ICA-PID Fig. 9(a) and Fig. 9(b) show that ICA-PID gives better tracking of 307

KP KI KD
frequency deviations and tie-line power in comparison with GA- 308

PID controller. Fig. 9(c) shows that Gencos of area-1 respond to 309
case 1 2 0.4181 2
the additional demand of area-1 and settled to a new generation 310
case 2 1.5932 0.3042 0.1804
level while Gencos of area-2 and area-3 settled to the same values 311

GA-PID as given in case 1. It is also seen that the deviation in change in 312

real power generation is quite less with ICA-PID, compared to GA- 313
KP KI KD
PID controller. It is observed that the tie-line powers, frequency 314
case 1 1.115 0.289 0.03 deviations, and change in generations are stabilized and settle to 315
case 2 2 0.359 0.447
their desired values at steady state. The responses (transient and 316

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
G Model
ASOC 3686 1–9 ARTICLE IN PRESS
N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx 7

Frequency deviaons (Area-1) Frequency deviaons (Area-1)-zoomed in view

0.2

0 0

Delta F1

Delta F1
-0.2 -0.05 GA-PID
GA-PID ICA-PID
ICA-PID
-0.4 -0.1

0 50 100 150 40 50 60 70 80 90 100

Time (s) Time (s)

Frequency deviaons (Area-2) Frequency deviaons (Area-3)

0.1 0.1

0 0
Delta F2

Delta F3
-0.1 -0.1
GA-PID
-0.2 GA-PID -0.2
ICA-PID
ICA-PID -0.3
-0.3

-0.4 -0.4

-0.5 -0.5
0 50 100 150 0 50 100 150
Time (s) Time (s)

(a). Frequency deviations (rad/s).

-3 -3
x 10 Tie-line power deviaons (Area 1-2) x 10 Tie-line power deviaons (Area 2-3)
5 8

6
GA-PID
0
4 ICA-PID

Delta Pe 2-3

Delta Pe 1-2

-5 2
GA-PID
0
ICA-PID
-10
-2

-15 -4
0 50 100 150 0 50 100 150
Time (s) Time (s)

-3
x 10 Tie-line power deviaons (Area 3-1)
8

6 GA-PID
ICA-PID
Delta Pe 3-1

-2

-4
0 50 100 150
Time (s)

(b). T ie line po wer (pu).

Change in generaon (Area-1) -G11 Change in generaon (Area-1 )- G12

0.014 0.014

0.012 0.012

0.01 0.01
Delta Pg11

Delta Pg12

0.008 0.008

0.006 GA-PID 0.006

GA-PID
0.004 ICA-PID 0.004 ICA-PID
0.002 0.002

0 0
0 50 100 150 0 50 100 150
Time (s) Time (s)

Change in generaon (Area-2) Change in generaon (Area-3)

0.015 0.015

0.01 0.01
Delta Pg2

Delta Pg3

GA-PID GA-PID
ICA-PID ICA-PID
0.005 0.005

0 0
0 50 Time (s) 100 150 0 50 Time (s) 100 150

(c). Generated po wer (pu).

Fig. 8. (a) Frequency deviations (rad/s). (b) Tie line power (pu). (c) Generated power (pu).

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
G Model
ASOC 3686 1–9 ARTICLE IN PRESS
8 N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx

Frequency deviaons (Area-1)

0.2 Frequency deviaons (Area-1) -zoomed in view
0.03
0 0.02 GA-PID
ICA-PID
0.01

Delta F1
Delta F1
-0.2
GA-PID 0
-0.4 ICA-PID -0.01

-0.6 -0.02
-0.03
-0.8
0 50 100 150 50 60 70 80 90 100
Time (s) Time (s)

Frequency deviaons (Area-2) Frequency deviaons (Area-3)

0.2 0.2

0 0
Delta F2

Delta F3
-0.2 -0.2
GA-PID GA-PID
-0.4 ICA-PID -0.4 ICA-PID

-0.6 -0.6

-0.8 -0.8
0 50 Time (s) 100 150 0 50 Time (s) 100 150

(a). Frequency deviation s (rad/s).

-3
Tie-line power deviaons (Area 1-2) x 10 Tie-line power deviaons (Area 2-3)
15

0
Delta Pe 2-3
10
Delta Pe 1-2

GA-PID
ICA-PID
-0.01
GA-PID 5
ICA-PID
-0.02 0

-0.03 -5
0 50 100 150 0 50 100 150
Time (s) Time (s)

-3
x 10 Tie-line power deviaons (Area 3-1)
15

10
Delta Pe 3-1

GA-PID
ICA-PID
5

-5
0 50
Time (s) 100 150

(b). T ie line po wer (pu).

Change in generaon (Area-1 )(G11) Change in generaon (Area-1) (G12)

0.015 0.015
Delta Pg11

Delta Pg12

0.01 GA-PID 0.01

ICA-PID GA-PID
ICA-PID
0.005 0.005

0 0
0 50 0 50 Time (s) 100 150
Time (s) 100 150

Change in generaon (Area-2) Change in generaon (Area-3)

0.02 0.02

GA-PID GA-PID
0.015 ICA-PID 0.015 ICA-PID
Delta Pg2

Delta Pg3

0.01 0.01

0.005 0.005

0 0
0 50
Time (s) 100 150 0 50
Time (s)
100 150

(c). Generated power (pu).

Fig. 9. (a). Frequency deviations (rad/s). (b) Tie line power (pu). (c) Generated power (pu).

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005
 G Model
ASOC 3686 1–9 ARTICLE IN PRESS
N. Kumar et al. / Applied Soft Computing xxx (2016) xxx–xxx 9

Table 9 [14] C.T. Pan, C.M. Liaw, An adaptive controller for power system load-frequency 370
Comparative Study of Settling Time (Three area power system). control, IEEE Trans. Power Syst. 4 (1989) 122–128. 371
[15] M.A. Sheirah, M.M. Abd-EI-Fattah, Improved load frequency self-tuning 372
Controllers Settling Time (second) regulator, Int. J. Control 39 (1984) 143–158. 373

Case 1 [16] Y. Wang, R. Zhou, C. Wen, Robust load-frequency controller design for power 374

f1 f2 f3 Ptie1-2 systems, Proc. Inst. Electr. Eng. C 140 (1993) 11–16. 375
[17] H. Bevrani, Y. Mitani, K. Tsuji, Robust decentralized AGC in a restructured 376
ICA-PID 90 85 90 80
power system, Energ Convers. Manage. 45 (2004) 2297–2312. 377
GA-PID 95 95 95 130
[18] M.R. Toulabi, M. Shiroei, A.M. Ranjbar, Robust analysis and design of power 378
Case 2 system load frequency control using the Kharitonov’s theorem, Electr. Power 379

f1 f2 f3 Ptie1-2 Energy Syst. 55 (2014) 51–58. 380

ICA-PID 90 90 90 90 [19] H. Shayeghi, A. Jalili, H.A. Shayanfar, Multi-stage fuzzy PID power system 381
automatic generation controller in deregulated environments, Energy 382
GA-PID 95 95 95 105
Convers. Manage. 47 (2006) 284–2829. 383
[20] F. Valdez, P. Melin, O. Castillo, A survey on nature inspired optimization 384
algorithms with fuzzy logic for dynamic parameter adaptation, Expert Syst. 385
317 steady state) show the superior performance of ICA-PID than GA- Appl. 41 (14) (2014) 6459–6466. 386
318 PID controller, in view of reduced settling time, and oscillations. [21] F. Valdez, P. Melin, O. Castillo, O. Montiel, A new evolutionary method with a 387
hybrid approach combining particle swarm optimization and genetic 388
319 Table 9 reveals that ICA-PID controller provides better settling per-
algorithms using fuzzy logic for decision making, 2008 IEEE Congr. Evol. 389
320 formance than GA-PID. Comput. CEC 2008 (2008) 1333–1339 (art. no. 4630968). 390
[22] O. Castillo, G. Huesca, F. Valdez, Evolutionary computing for topology 391
optimization of type2 fuzzy controllers, Stud. Fuzziness Soft Comput. 208 392
321 5. Conclusion
(2007) 163–178. 393
[23] O. Castillo, G. Huesca, F. Valdez, Evolutionary computing for fuzzy system 394
322 In this paper, a PID controller is designed and applied to a mul- optimization in intelligent control, Proc. Int. Conf. Artif. Intell. ICAI 04 (1) 395

323 tiarea power system in restructured environment. 2-area thermal (2004) 98–104. 396
[24] Valdez F., Melin P., Castillo O., 2010. Fuzzy logic for parameter tuning in 397
324 power system and 3-area thermal-hydro power system with GRC evolutionary computation and bio inspired methods, Lecture Notes in 398
325 have been used to check the performance of the controller. ICA is Computer Science (including subseries Lecture Notes in Artificial Intelligence 399

326 used to determine the optimal parameters of PID controller. The and Lecture Notes in Bioinformatics), 6438 LNAI (PART 2), 465474. 400
[25] Y.L. Karnavas, D.P. Papadopoulos, AGC for autonomous power system using 401
327 convergence of the ICA is faster compared to other search algo- combined intelligent techniques, Electr. Power Syst. 62 (3) (2002) 225–239. 402
328 rithm like GA. The PID controller is designed in such a way that it [26] F. Daneshfar, H. Bevrani, Multiobjective design of load frequency control 403

329 minimizes the ACE. Different market transactions and load change using genetic algorithms, Electr. Power Energy Syst. 42 (2012) 257–263. 404
[27] E.S. Ali, S.M. Abd-Elazim, Bacteria foraging optimization algorithm based load 405
330 scenarios have been considered to show the effectiveness of the frequency controller for interconnected power system, Electr. Power Energy 406
331 controller. Results show that the frequency error got eliminated Syst. 33 (2011) 633–638. 407
332 in steady state and Gencos shared the load demand according to [28] L.C. Saikia, S. Nidul, J. Nanda, Maiden application of bacterial foraging based 408
fuzzy IDD controller in AGC of a multi-area hydrothermal system, Electr. 409
333 market transactions. ICA-PID controller has also been compared Power Energy Syst. 45 (2013) 98–106. 410
334 with GA-PID controller. The comparative results show that ICA-PID [29] S. Panda, N.P. Padhy, Comparison of particle swarm optimization and genetic 411
335 controller has better performance and improves system responses algorithm for FACTS-based controller design, Appl. Soft Comput. 8 (4) (2008) 412
1418–1427. 413
336 more effectively compared to GA-PID controller in terms of settling
[30] S. Pandaa, B. Mohanty, P.K. Hotab, Hybrid BFOA-PSO algorithm for automatic 414
337 time and oscillations. generation control oflinear and nonlinear interconnected power systems, 415
Appl. Soft Comput. 13 (2013) 4718–4730. 416
[31] E. Atashpaz-Gargari, C. Lucas, Imperialist competitive algorithm: an algorithm 417
338 References
for optimization inspired by imperialistic competition, in: IEEE 2007 418
Evolutionary Computation (CEC), IEEE Press, New York, 2007, pp. 4661–4667. 419
339 [1] O.I. Elgerd, C. Fosha, Optimum megawatt- frequency control of multi-area [32] M.J. Morshed, A. Asgharpour, Hybrid imperialist competitive-sequential 420
340 electric energy systems, IEEE Trans. Power Syst. 89 (1970) 556–563. quadratic programming (HIC-SQP) algorithm for solving economic load 421
341 [2] N. Jaleeli, D.N. Ewart, L.H. Fink, Understanding automatic generation control, dispatch with incorporating stochastic wind power: a comparative study on 422
342 IEEE Trans. Power Syst. 7 (1992) 1106–1122. heuristic optimization techniques, Energy Convers. Manage. 84 (2014) 30–40. 423
343 [3] H. Saadat, Power System Analysis, McGraw-Hill, New York, 1999. [33] H. Hosseini, T. Behrooz, A Novel method for designing PSS-AVR by imperialist 424
344 [4] J. Kumar, K.H. Ng, G. Sheble, AGC simulator for price based operation part I, competitive algorithm (ICA) for three-area AGC System, J. World Electr. Eng. 425
345 IEEE Trans. Power Syst. 12 (2 May) (1997) 527–532. Technol. 1 (2012) 43–50. 426
346 [5] J. Kumar, K.H. Ng, G. Sheble, AGC simulator for price based operation part II, [34] M.A. Ahmadia, M. Ebadi, A. Shokrollahi, S.M.J. Majidi, Evolving artificial neural 427
347 IEEE Trans. Power Syst. 12 (2 May) (1997) 527–532. network and imperialist competitive algorithm for prediction oil flow rate of 428
348 [6] V. Donde, M.A. Pai, I.A. Hiskens, Simulation and optimization in an AGC the reservoir, Appl. Soft Comput. 13 (2013) 1085–1098. 429
349 system after deregulation, IEEE Trans. Power Syst. 16 (3) (2001) 481–489. [35] M. Ghasemi, S. Ghavidel, M.M. Ghanbarian, A. Habibi, A new hybrid algorithm 430
350 [7] B. Tyagi, S.C. Srivastava, Automatic generation control scheme based on for optimal reactive power dispatch problem with discrete and continuous 431
351 dynamic participation of generators for competitive electricity markets, in: control variables, Appl. Soft Comput. 22 (2014) 126–140. 432
352 Proc. National Power Systems (NPSC), December 16–18, IIT-Bombay, 2008. [36] E. Atashpaz-Gargari, C. Lucas, Designing an Optimal PID Controller Using 433
353 [8] B. Tyagi, S.C. Srivastava, A LQG based load frequency controller in a Imperialist Competitive Algorithm, First Joint Congress on Fuzzy and 434
354 competitive electricity environment, Int. J. Emerg. Electr. Power Syst. 2 (2) Intelligent Systems Ferdowsi University of Mashhad, Iran, Aug, 2007, pp. 435
355 (2005). 29–31. 436
356 [9] R.D. Christie, A. Bose, Load frequency control issues in power system operation [37] H. Shabani, B. Vahidi, M. Ebrahimpour, A robust PID controller based on 437
357 after deregulation, IEEE Trans. Power Syst. 11 (3) (1996) 1191–1200. imperialist competitive algorithm for load-frequency control of power 438
358 [10] F. Liu, Y.H. Song, J. Ma, S. Mei, Q. Lu, Optimal Load-frequency Control in systems, ISAT 52 (2013) 88–95. 439
359 Restructured Power Systems IEE Proc. on Generation, Transmission and [38] E.A. Gargari, C. Lucas, A novel approach for PID controller design in MIMO 440
360 Distribution, IEEE Press, New York, 2003, pp. 87–95. distillation column process, Int. J. Intell. Comput. Cybern. 1 (2008) 337–355. 441
361 [11] E. Rakhshani, Intelligent linear-quadratic optimal output feedback regulator [39] M. Soheilirad, P. Farzan, M.L. Othman, K. Karami, M. Hojabri, PID controller 442
362 for a deregulated automatic generation control system, Electr. Power adjustment for MA-LFC by using imperialist competitive algorithm, in: IEEE 443
363 Compon. Syst. 40 (5) (2012) 513–533. 7th International Power Engineering and Optimization Conference (PEOCO), 444
364 [12] Z. Al-Hamouz, N. Al-Musabi, H. Al-Duwaish, S. Al-Baiyat, On the design of IEEE Press, New York, 2013, pp. 507–512. 445
365 variable structure load frequency controllers by tabu search algorithm: [40] S. Debbarma, L.C. Saikia, N. Sinha, AGC of a multi-area thermal system under 446
366 application to nonlinear interconnected models, Electr. Power Compon. Syst. deregulated environment using a non-integer controller, Electr. Power Syst. 447
367 33 (1) (2005) 1253–1267. Res. 95 (2013) 175–183. 448
368 [13] B.M. Al-Hadithi, A.J. Barragán, J.M. Andújar, A. Jiménez, Variable structure
369 control with chattering elimination and guaranteed stability for a generalized
T-S model, Appl. Soft Comput. 13 (2013) 4802–4812.

Please cite this article in press as: N. Kumar, et al., Multi area AGC scheme using imperialist competition algorithm in restructured
power system, Appl. Soft Comput. J. (2016), http://dx.doi.org/10.1016/j.asoc.2016.07.005