Professional Documents
Culture Documents
0885-8950 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
4092 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 5, SEPTEMBER 2017
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
OJAGHI AND RAHMANI: LMI-BASED ROBUST PREDICTIVE LOAD FREQUENCY CONTROL FOR POWER SYSTEMS WITH COMMUNICATION 4093
Fig. 1. Dynamic model of the ith control area in a multi-area LFC system.
where where
x̄(t) = Δf ΔPm ΔPv
x(t) = Δf ΔPm ΔPv ΔE (8)
ȳ(t) = ACE (2) ⎡ ⎤
⎡ D 1 ⎤ −
D 1
0 0
− 0 ⎢ M M ⎥
⎢ M M ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ −1 ⎥ 1 ⎢ 0 −1 1
0⎥
1 ⎢ ⎥
Ā = ⎢
⎢ 0
⎥, B̄ = 0
⎥ 0 , A= ⎢ Tch Tch ⎥,
⎢ Tch Tch ⎥ Tg ⎢ −1 −1 ⎥
⎣ −1 ⎦ ⎢ ⎥
−1 ⎢ 0 0⎥
0 ⎣ RTg Tg ⎦
RTg Tg
β 0 0 0
−1 ⎡ ⎤
F̄ = 0 0 , C̄ = β 0 0 (3) 0 0 0 0
M
⎢ ⎥ 1
⎢0 0 0 0 ⎥ B= 0 0 0
Considering no net tie-line power exchange in the one-area ⎢ ⎥ Tg
⎢ ⎥
LFC system and defining β as the frequency bias factor, the Ad = ⎢ ⎥
⎢ 0 0 0 −1 ⎥, F = −1
Also, ΔE is defined as
t
ΔE(t) = ACE(τ )dτ, (5) B. Multi-Area LFC Model
0
All generation units in each control area for a multi-area LFC
which demonstrates the integral of frequency deviation that scheme are simplified as an equivalent generation unit. Fig. 1
needs to go to zero at steady state [14]. illustrates the schematic diagram of the model of ith power
Due to the ACE delay, the linearized model of the governor area. The dynamic model of an N -area LFC system is given as
valve position is given by [35]: follows [1], [23]:
Δf (t) ΔPv (t) ΔE(t − d(t)) u(t)
ΔP˙v (t) = − − − + (6)
RTg Tg Tg Tg ẋ(t) = Ax(t) + Ad x(t − d(t)) + Bu(t) + F ΔPd (t)
(9)
where d(t) denotes the delay of the system. y(t) = Cx(t)
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
4094 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 5, SEPTEMBER 2017
where Ji (k) is the quadratic cost function at the ith area given
Consider that all variables and parameters for ith area use the as
same notations as one-area LFC system with an additional sub- ∞
script i. Ji (k) = xi (k + l|k) Qi xi (k + l|k)
The ACE signal in a multi-area LFC system is the sum of l=0
tie-line power exchange and frequency deviation weighted by a
bias factor as follows: + ui (k + l|k) Ri ui (k + l|k), (17)
n i ×n i m i ×m i
in which Qi ∈ R and Ri ∈ R are two given sym-
ACEi = βi Δfi + ΔPtie−i . (12) metric positive definite matrices, and l is the prediction horizon.
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
OJAGHI AND RAHMANI: LMI-BASED ROBUST PREDICTIVE LOAD FREQUENCY CONTROL FOR POWER SYSTEMS WITH COMMUNICATION 4095
Aii xi (k + l) + Bi ui (k + l) + Adi xi (k + l − d) + Di zi (k + l) Pi Aii xi (k + l) + Bi ui (k + l) + Adi xi (k + l − d)
d−2
k +l+1−d
+ Di zi (k + l) − xi (k + l) Pi xi (k + l) + xi (k + l − j − 1) P1i xi (k + l − j − 1) + xi (j) P1i xi (j)
j =0 j =0
k
+l−d
d
d
d−2
− xi (j) P1i xi (j) + xi (j) P1i xi (j) − xi (j) P1i xi (j) − xi (k + l − j − 2) P1i xi (k + l − j − 2)
j =0 j =k +l j =k +l−1 j =0
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
4096 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 5, SEPTEMBER 2017
where
⎡ ⎤ Algorithm 1: Proposed decentralized robust predictive load
Gi 0 0 0 frequency control strategy.
⎢ 0 Mi 0 0 ⎥
Δ11 =⎢
⎣ 0
⎥, 1: Inputs:
0 G1i 0 ⎦ The state space equations of a multi-area power
0 0 0 M1i system.
Δ12 = Δ 2: Initialize:
21 =
⎡ Initial values of states and initial condition of the
(Aii Gi + Bi Yi ) 0 0 0 system for each area. The start time is k = 1.
⎢ 3: while k < kend ∗ do
⎢ 0 Mi Mi 0
⎢ 4: for i = 1 to N do
⎢ (A G )
⎣ di 1i 0 0 0 5: Obtain the feedback gain of ith area Li by
(Di M1i ) 0 0 M1i solving the main optimization problem presented in the
⎤ (27), and then compute the control input as
(Q0.5
i Gi ) (Ri0.5 Gi ) ui (k + l|k) = Li (k)xi (k + l|k).
⎥
0 0 ⎥ 6: end for
⎥,
0 0 ⎥ 7: Apply the control input to the system at time k.
⎦ 8: k = k + 1:
0 0 9: end while
* kend is the final time of running the algorithm.
Δ22 = diag(Gi , Mi , G1i , M1i , γi I, γi I).
In summary, the optimization problem at ith area that minimizes
the cost function (17), and guarantees the stability of the area in
an interconnected multi-area LFC system can be reformulated C. Non-Predictive LMI-Based Robust LFC Approach
as the following optimization problem with the LMI constraints. The proposed design algorithm can be converted to a non-
predictive LMI-based robust optimal LFC scheme with the fol-
min γi
γ i ,G i ,G 1 i ,M i ,M 1 i ,Y i lowing cost function at ith area.
subject to (23), (26). (27)
∞
The main aim of this optimization problem is to minimize the Ji = xi (k) Qi xi (k) + ui (k) Ri ui (k). (30)
upper bound of the cost function (17) in order to regularize k =0
the states of the system including the frequency deviation with
the minimum control input. It is called non-predictive, because the prediction horizon, l, is
Finally, at each sample time k, the decentralized control input omitted in the cost function.
of ith area is obtained as a state-feedback control law given by The state-feedback control input ui (k) = Li xi (k) can be
achieved by a similar procedure as presented in Section III-
ui (k + |k) = Li (k)xi (k + |k), (28)
B, and the corresponding optimization problem would also be
in which the gain Li is computed by the optimization variables the same as (27).
obtained from solving (27) as follows Remark 3: In non-predictive LMI-based RMPC, to solve the
optimization problem (27), in the LMI constraint (23), xi (k|k)
Li (k) = Yi G−1
i (29) and xi (k − d|k) should be replaced by xi (0) and xi (−d), re-
and the procedure is repeated for time k + 1. spectively.
The proposed robust predictive load frequency control can be Remark 4: In the proposed predictive LFC algorithm, the
summarized in Algorithm 1. optimization problem is solved at all times k to obtain time-
Remark 2: In the decentralized LFC approach, the optimiza- varying Li (k) and achieve l-step ahead prediction of states and
tion problem (27) is solved locally at ith area, for i = 1, · · · , N , inputs; however, in the non-predictive approach, the optimiza-
where the complexity is related to the dimension of that area. tion problem is solved once to obtain the constant state feedback
By parallel processing of areas, the computational effort of the gain Li . As a result, the non-predictive LFC approach has lower
problem for an N -area power system is in the order of a one- computational effort but it loses the time-varying property of
area’s control problem. This can significantly reduce the com- the proposed predictive LFC strategy.
putational time of the algorithm in particular when the number
of areas increases.
D. Centralized LMI-Based Robust Predictive LFC Approach
Although this control scheme is a decentralized RMPC ap-
proach for multi-area LFC systems, it can be easily changed to The proposed decentralized algorithm can be applied as a cen-
a centralized method for a single-area power system, and also tralized approach for a single-area LFC system. To this end, by
can be applied as a (non-predictive) robust LFC approach with eliminating the interaction matrix in (13), considering one Lya-
lower computational complexity. punov function for one-area, and following the same procedure,
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
OJAGHI AND RAHMANI: LMI-BASED ROBUST PREDICTIVE LOAD FREQUENCY CONTROL FOR POWER SYSTEMS WITH COMMUNICATION 4097
Δ11 Δ12
Δ= > 0, (32)
Δ21 Δ22
where
⎡ ⎤
G 0 0
Δ11 =⎣0 M 0 ⎦,
0 0 G1
Δ12 = Δ 21 =
⎡ ⎤
(AG + BY ) 0 0 (Q0.5 G) (R0.5 G) Fig. 2. Time responses of different controllers for uncertain one-area LFC
⎣ 0 M M 0 0 ⎦, system (Case I).
(Ad G1 ) 0 0 0 0
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
4098 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 5, SEPTEMBER 2017
Fig. 4. ΔE and Δf responses of all areas in a multi-area LFC scheme with different methods (Case III).
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
OJAGHI AND RAHMANI: LMI-BASED ROBUST PREDICTIVE LOAD FREQUENCY CONTROL FOR POWER SYSTEMS WITH COMMUNICATION 4099
Remark 6: The proposed predictive LFC approach solved The system parameters of the three-area power system are
well in the range of real-time applications, e.g. the average time presented as follows [35].
to solve the optimization problem in (27) at each sample time 1) First area
for one area is less than 0.8 sec.
Tch1 = 0.3 s, Tch2 = 0.3 s, Tg 1 = 0.1 s, Tg 2 = 0.1 s
It should be noted that all simulations are performed on a
same PC with an Intel Core i7 CPU and 4 GB RAM. R1 = 0.05, R2 = 0.05, D1 = 1.5, M1 = 10
2
β1 = + D1 (36)
V. DISCUSSIONS R1
In the above studies, one-area and multi-area power systems 2) Second area (equivalent unit of four generators)
with communication delays are considered to evaluate the per-
Tch3 = 0.17 s, Tg 3 = 0.4 s, R3 = 0.05, D3 = 1.5,
formance of the proposed robust predictive LFC approach. For
a single-area LFC system, Figs. 2–3 show the better perfor- 4
M3 = 12, β3 = + D3 (37)
mance of the proposed control strategy compared with the two R3
existing LFC approaches. It should be pointed out that consider- 3) Third area (equivalent unit of three generators)
ing the communication delay with uncertainty and time-varying
parameters in the LFC system are key features different from Tch4 = 0.17 s, Tg 4 = 0.4 s, R4 = 0.05, D4 = 1.5,
some existing works. Moreover, for a multi-area LFC system, 3
Figs. 4–5 indicate that the proposed decentralized predictive M4 = 12, β4 = + D4 . (38)
R4
LFC strategy is preferable than the existing decentralized LFC
approaches. Considering decentralized control structure for a REFERENCES
multi-area LFC system with time-delays, uncertainty, and time-
[1] H. Bevrani, Robust Power System Frequency Control, 2nd ed. New York,
varying parameters is the other new feature which is out of the NY, USA: Springer, 2014.
scope of those previous works. [2] N. Chuang, “Robust H ∞ load-frequency control in interconnected power
systems,” IET Control Theory Appl., vol. 10, no. 1, pp. 67–75, 2016.
[3] S. Snmez and S. Ayasun, “Stability region in the parameter space of PI
VI. CONCLUSION controller for a single-area load frequency control system with time delay,”
IEEE Trans. Power Syst., vol. 31, no. 1, pp. 829–830, Jan. 2016.
In this paper, a new LMI-based robust predictive LFC is [4] C.-K. Zhang, L. Jiang, Q. Wu, Y. He, and M. Wu, “Delay-dependent robust
load frequency control for time delay power systems,” IEEE Trans. Power
proposed for power systems with communication delays. In Syst., vol. 28, no. 3, pp. 2192–2201, Aug. 2013.
this approach, the stability of the systems is guaranteed by the [5] A. Khodabakhshian and M. Edrisi, “A new robust PID load frequency
Lyapunov stability theorem and state-feedback control input is controller,” Control Eng. Pract., vol. 16, no. 9, pp. 1069–1080, 2008.
[6] W. Tan, “Unified tuning of PID load frequency controller for power sys-
obtained by solving a semi-definite programming with LMI con- tems via IMC,” IEEE Trans., Power Syst., vol. 25, no. 1, pp. 341–350,
straints. Although the algorithm is developed as a decentralized Feb. 2010.
control approach for multi-area LFC systems, it can be applied [7] R. Vijaya Santhi, K. Sudha, and S. Prameela Devi, “Robust load frequency
control of multi-area interconnected system including SMES units using
as a centralized method for single-area power systems. This con- type-2 fuzzy controller,” in Proc. 2013 IEEE Int. Conf. Fuzzy Syst., 2013,
trol scheme is robust with respect to uncertainties in dynamic pp. 1–7.
model and has good performance for time delays in the com- [8] C. Boonchuay, “Improving regulation service based on adaptive load fre-
quency control in LMP energy market,” IEEE Trans. Power Syst., vol. 29,
munication network. Moreover, it is shown that the predictive no. 2, pp. 988–989, Mar. 2014.
LFC (PLFC) approach is applicable to time-varying power sys- [9] S. Nag and N. Philip, “Application of neural networks to automatic load
tems caused by failures in the networks. The proposed approach frequency control,” in Proc. Int. Conf. Control, Instrum., Energy Commun.,
2014, pp. 345–350.
is also reformulated as a non-predictive LFC (NPLFC) method. [10] C. Boonchuay, “Improving regulation service based on adaptive load fre-
The NPLFC has lower computational effort, however, it loses the quency control in LMP energy market,” IEEE Trans. Power Syst., vol. 29,
good performance of the proposed predictive LFC in particular no. 2, pp. 988–989, Mar. 2014.
[11] L. Dong and Y. Zhang, “On design of a robust load frequency controller
for time-varying power systems. Simulation results for differ- for interconnected power systems,” in Proc. Amer. Control Conf., 2010,
ent cases of uncertain and time-varying LFC systems with time 2010, pp. 1731–1736.
delays, along with the decentralized scheme for a multi-area [12] H. Bevrani, Y. Mitani, and K. Tsuji, “Robust decentralised load-frequency
control using an iterative linear matrix inequalities algorithm,” IEE Proc.-
power system show the robustness and superior performance of Gener., Transm. Distrib., vol. 151, no. 3, pp. 347–354, 2004.
the proposed LFC approach. [13] A. Bensenouci and A. A. Ghany, “Mixed H ∞ /H 2 with pole-placement
design of robust lmi-based output feedback controllers for multi-area load
frequency control,” in Proc. EUROCON 2007—Int. Conf. Comput. Tool,
APPENDIX A 2007, pp. 1561–1566.
[14] M. Rahmani and N. Sadati, “Hierarchical optimal robust load-frequency
The nominal values of parameters for the one-area LFC sys- control for power systems,” IET Gener., Transm. Distrib., vol. 6, no. 4,
tem are given by [35] pp. 303–312, Apr. 2012.
[15] W. Yao, L. Jiang, Q. Wu, J. Wen, and S. Cheng, “Delay-dependent stability
analysis of the power system with a wide-area damping controller embed-
Tch = 0.17 s, Tg = 0.4 s, R = 0.05, D = 1.5, M = 12, ded,” IEEE Trans., Power Syst., vol. 26, no. 1, pp. 233–240, Feb. 2011.
[16] S. Wen, X. Yu, Z. Zeng, and J. Wang, “Event-triggering load frequency
4 control for multi-area power systems with communication delays,” IEEE
β= + D. (35) Trans. Ind. Electron., vol. 63, no. 2, pp. 1308–1317, 2016.
R
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.
4100 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 5, SEPTEMBER 2017
[17] K. Vrdoljak, I. Petrovic, and N. Peric, “Discrete-time sliding mode control [31] A. M. Ersdal, L. Imsland, and K. Uhlen, “Model predictive load-frequency
of load frequency in power systems with input delay,” in Proc. 12th Int. control,” IEEE Trans. Power Syst., vol. 31, no. 1, pp. 777–785, Jan. 2016.
Power Electron. Motion Control Conf., 2006, pp. 567–572. [32] L. Dong, Y. Zhang, and Z. Gao, “A robust decentralized load frequency
[18] M. Liu, L. Yang, D. Gan, D. Wang, F. Gao, and Y. Chen, “The stability controller for interconnected power systems,” ISA Trans., vol. 51, no. 3,
of AGC systems with commensurate delays,” Eur. Trans. Elect. Power, pp. 410–419, 2012.
vol. 17, no. 6, pp. 615–627, 2007. [33] Y. Mi, Y. Fu, C. Wang, and P. Wang, “Decentralized sliding mode load
[19] L. Yongjuan, M. Yang, Y. Yang, and W. Limin, “The study of sliding mode frequency control for multi-area power systems,” IEEE Trans. Power Syst.,
load frequency control for single area time delay power system,” in Proc. vol. 28, no. 4, pp. 4301–4309, Nov. 2013.
2015 27th Chin. Control Decis. Conf., 2015, pp. 602–607. [34] M. Yang, L. Wenlin, and J. Yuanwei, “Decentralized load frequency vari-
[20] X. Xie, Y. Xin, J. Xiao, J. Wu, and Y. Han, “WAMS applications able structure control of multi-area interconnected power system,” in Proc.
in chinese power systems,” IEEE Power Energy Mag., vol. 4, no. 1, 2011 Chin. Control Decis. Conf., May 2011, pp. 666–669.
pp. 54–63, Jan./Feb. 2006. [35] X. Yu and K. Tomsovic, “Application of linear matrix inequalities for load
[21] H. Bevrani and T. Hiyama, “Robust decentralised PI based LFC design frequency control with communication delays,” IEEE Trans. Power Syst.,
for time delay power systems,” Energy Convers. Manage., vol. 49, no. 2, vol. 19, no. 3, pp. 1508–1515, Aug. 2004.
pp. 193–204, 2008. [36] P. Kundur, N. J. Balu, and M. G. Lauby, Power System Stability and
[22] S. Sonmez and S. Ayasun, “Stability region in the parameter space of Control, vol. 7. New York, NY, USA: McGraw-Hill, 1994.
PI controller for a single-area load frequency control system with time [37] H. K. Khalil, Nonlinear Systems. Englewood Cliffs, NJ, USA: Prentice-
delay,” IEEE Trans. Power Syst., vol. 31, no. 1, pp. 829–830, Jan. 2016. Hall, 2002.
[23] L. Jiang, W. Yao, Q. Wu, J. Wen, and S. Cheng, “Delay-dependent stability [38] L. El Ghaoui and S.-I. Niculescu, Advances in Linear Matrix Inequality
for load frequency control with constant and time-varying delays,” IEEE Methods in Control. Philadelphia, PA, USA: SIAM, 2000, vol. 2.
Trans. Power Syst., vol. 27, no. 2, pp. 932–941, May 2012.
[24] S. Sonmez, S. Ayasun, and C. Nwankpa, “An exact method for computing
delay margin for stability of load frequency control systems with con-
stant communication delays,” IEEE Trans. Power Syst., vol. 31, no. 1, Pegah Ojaghi received the B.Sc. degree in electrical
pp. 370–377, Jan. 2016. engineering from Hamedan University of Technol-
[25] C. Peng and J. Zhang, “Delay-distribution-dependent load frequency con- ogy, Hamedan, Iran, in 2012 and the M.Sc. degree in
trol of power systems with probabilistic interval delays,” IEEE Trans. electrical engineering from Imam Khomeini Interna-
Power Syst., vol. 31, no. 4, pp. 3309–3317, Jul. 2016. tional University, Qazvin, Iran, in 2015.
[26] M. S. M. Cavalca, R. K. H. Galvão, and T. Yoneyama, “Robust model pre- Her current research interests include model pre-
dictive control using linear matrix inequalities for the treatment of asym- dictive control, robust control, nonlinear systems,
metric output constraints,” J. Control Sci. Eng., 2012. [Online]. Available: load frequency control, and stability analysis and con-
http://dx.doi.org/10.1155/2012/485784 trol of power systems.
[27] P. Ojaghi, N. Bigdeli, and M. Rahmani, “An LMI approach to robust
model predictive control of nonlinear systems with state-dependent un-
certainties,” J. Process Control, vol. 47, pp. 1–10, 2016. Mehdi Rahmani (S’10–M’13) received the B.Sc.,
[28] J. Zhang, X. Zhao, Y. Zuo, and R. Zhang, “Linear programming-based M.Sc., and Ph.D. degrees in electrical engineering
robust model predictive control for positive systems,” IET Control Theory from Sharif University of Technology, Tehran, Iran,
Appl., vol. 10, no. 15, pp. 1789–1797, 2016. in 2002, 2004, and 2013, respectively.
[29] X. Liu, H. Nong, K. Xi, and X. Yao, “Robust distributed model predictive In 2013, he joined the Department of Electrical En-
load frequency control of interconnected power system,” Math. Probl. gineering, Imam Khomeini International University
Eng., vol. 2013, 2013, Art. no. 468168. (IKIU), Qazvin, Iran, where he has been the Head of
[30] X. Liu, X. Kong, and K. Y. Lee, “Distributed model predictive control for the EE group since 2016. His current research inter-
load frequency control with dynamic fuzzy valve position modelling for ests include robust control, model predictive control,
hydro-thermal power system,” IET Control Theory Appl., vol. 10, no. 14, estimation, distributed systems, and power systems’
pp. 1653–1664, 2016. modeling, stabilization, optimization and control.
Authorized licensed use limited to: University of Nantes. Downloaded on January 07,2022 at 09:07:05 UTC from IEEE Xplore. Restrictions apply.