Professional Documents
Culture Documents
ISA Transactions: Zhihuan Chen, Yanbin Yuan, Xiaohui Yuan, Yuehua Huang, Xianshan Li, Wenwu Li
ISA Transactions: Zhihuan Chen, Yanbin Yuan, Xiaohui Yuan, Yuehua Huang, Xianshan Li, Wenwu Li
ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans
School of Hydropower and Information Engineering, Huazhong University of Science and Technology, 430074 Wuhan, China
School of Resource and Environmental Engineering, Wuhan University of Technology, 430070 Wuhan, China
c
College of Electrical Engineering and New Energy, China Three Gorges University, 443002 Yichang, China
b
art ic l e i nf o
a b s t r a c t
Article history:
Received 10 July 2014
Received in revised form
11 October 2014
Accepted 7 November 2014
Available online 4 December 2014
This paper was recommended for publication by Dr. Ahmad B. Rad.
A hydraulic turbine regulating system (HTRS) is one of the most important components of hydropower
plant, which plays a key role in maintaining safety, stability and economical operation of hydro-electrical
installations. At present, the conventional PID controller is widely applied in the HTRS system for its
practicability and robustness, and the primary problem with respect to this control law is how to
optimally tune the parameters, i.e. the determination of PID controller gains for satisfactory performance. In this paper, a kind of multi-objective evolutionary algorithms, named adaptive grid particle
swarm optimization (AGPSO) is applied to solve the PID gains tuning problem of the HTRS system. This
newly AGPSO optimized method, which differs from a traditional one-single objective optimization
method, is designed to take care of settling time and overshoot level simultaneously, in which a set of
non-inferior alternatives solutions (i.e. Pareto solution) is generated. Furthermore, a fuzzy-based
membership value assignment method is employed to choose the best compromise solution from the
obtained Pareto set. An illustrative example associated with the best compromise solution for parameter
tuning of the nonlinear HTRS system is introduced to verify the feasibility and the effectiveness of the
proposed AGPSO-based optimization approach, as compared with two another prominent multiobjective algorithms, i.e. Non-dominated Sorting Genetic Algorithm II (NSGAII) and Strength Pareto
Evolutionary Algorithm II (SPEAII), for the quality and diversity of obtained Pareto solutions set.
Consequently, simulation results show that this AGPSO optimized approach outperforms than compared
methods with higher efciency and better quality no matter whether the HTRS system works under
unload or load conditions.
& 2014 ISA. Published by Elsevier Ltd. All rights reserved.
Keywords:
Multi-objective
Adaptive grid particle swarm optimization
Hydraulic turbine regulating system
Non-dominated Sorting Genetic Algorithm
II (NSGAII)
Strength Pareto Evolutionary Algorithm II
(SPEAII)
1. Introduction
Despite signicant strides in the development of advanced
control schemes over the past decades, such as optimal control,
adaptive control, robust control, neuron control and intelligent
control have been developed for power system stabilization and
oscillation damping process [18], the classical ProportionalIntegral-Derivative (PID) controller and its variants is still a
preferred choice for power plants because of its structural simplicity, reliability, and favorable ratio between performance and cost.
A hydraulic turbine regulating system (HTRS) is the adjustment
component of the hydropower station, which is the biggest
contributor in all the renewable energy [9] and until now PID
http://dx.doi.org/10.1016/j.isatra.2014.11.003
0019-0578/& 2014 ISA. Published by Elsevier Ltd. All rights reserved.
174
ki
kd s
s
where kp , ki and kd are the proportional gain, integral gain, differential gain of PID controller. us, rs and xs are the Laplace transform of controller output u, given speed output r and synchronous
generator speed x.
There always exists a lower limit and upper limit for the
controller, which can signicantly reduce the mechanical wear
and tear of guide vane once there is a frequently adjustment of
turbine blades. This limitation can be simplied as a saturation
element in the system as follows:
um
u Nz
z z0 z z0
3
2z0
where um is the upper bound output of controller, z0 is the corresponding maximum input of um .
2.2. Model of servomechanism
The electro-water servomechanism is the actuator of water
turbine. It is made up by the major servomotor and auxiliary
servomotor. Generally, the parameter of auxiliary servomotor is far
less than the parameter of major servomotor, thus the model can
be conducted as a one-order system
(
Gc s T y s1 1
4
ys Gc s U us
where T y is major servo time constant and ys is the Laplace
transform of servo output y.
2.3. Model of water turbine and penstock system
175
Reservoir
Surge tank
Power Grid
Tunnel
Generator
Turbine
gate
Hydraulic
turbine
Penstock
Controller
Draft tube
q
r
+
Governor/
Controller
u Servomechannism
Guide
vane
Water
turbine
Conduit
mt
Generator
h
x
Fig. 1. Schematic of hydropower plant with its structures and components (a) general layout form and (b) block diagram form.
hs
Tws
qs
q f 2 x; y; h
eh mt =h; ex mt =x; ey mt =y
eqh q=h; eqx q=x; eqy q=y
em T w s
Gt s ey 11
eqh T w s
mt s Gt s Uys
where em eqy eh =ey eqh , mt s is the Laplace transform of turbine mechanical torque mt . The block diagram of Gt s is shown in
Fig. 2.
xs
1
mt s mg s T a s en
10
176
Fig. 3. The frame diagram of transfer functions of the hydraulic turbine regulating system.
NS=[3.7,1.2]
7 A
E
F
E
F
4
G
H
I
J
K
K
L
1
0
NS
L
M
0
0
Fig. 4. Graphical representation of a new element's insertion into the adaptive grid (lies within the current boundaries of the grid).
NS=[6.2,0.42]
7 A
B
C
E
F
177
E
F
H
I
J
K
K
L
1
0
NS
Fig. 5. Graphical representation of a new element's insertion into the adaptive grid (lies out the current boundaries of the grid).
Fig. 6. The introduction of this special mutation operator (a) pseudo-code of the operator and (b) behavior of the operator.
178
: ( x2 A : x1 ! x2
Z
J 2 ITSE
te2 tdt
12
17
f 2 x;
f k x
i 1; 2; ; p ;
j 1; 2; ; r;
p r m
13
14
15
where x x1 ; x2 ; xn ; A is the feasible solution vector or decision variables, is the available domain of solutions, Fx is a
vector of objectives, f x is the objective function and gx and hx
are the constraint functions of problem, p and q are the number of
inequality constraint and equality constraint, respectively, m is the
total number of constraint functions.
These functions f 1 x; f 2 x; ; f k x, usually in conict with each
other, are a mathematical description of performance criteria.
There are few encounters that a vector of the decision variables
that optimizes all objectives simultaneously, thus the concept of
Pareto optimality associated with the trade-off between different
objectives is adopted [19]
T
16
179
8.3
x 104
25
20
Fitness assignment
8.2
15
10
8.1
5
8
0
-5
7.9
-10
-15
7.8
-20
7.7
10
20
30
40
50
60
-25
-50
-40
-30
-20
-10
10
Iteration
Fig. 9. The approximation result of saturation element and hyperbolic tangent element.
20
30
40
50
180
those grids that contain more than one particle and it can be seen as a
form of tness sharing [30], and then we apply roulette-wheel
selection with these tness values to select the suitable grid. Once
the grid has been selected, from which we will take the particle
within it randomly. This chosen particle is the corresponding global
best position gbest as leader.
Fig. 10. The owchart of the AGPSO-based hydraulic turbine regulating system parameter optimization.
181
Table 1
Parameters in the hydraulic turbine regulating system.
Ty
Ta
Tw
ey
eqy
eh
eqh
en
0.3
5.72
0.83
1.40
1.23
0.35
0.15
0.45
Table 2
Comparison of obtained Pareto-front solutions under unload condition in trials.
Trials
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
QM (%)
SM
DM
AGPSO
NSGAII
SPEAII
AGPSO
NSGAII
SPEAII
AGPSO
NSGAII
SPEAII
75.6757
60.8696
43.4783
38.5965
42.7083
69.0476
30.7087
32.0000
37.3626
27.8261
37.2727
56.9892
57.1429
51.0000
36.1345
45.7944
39.7959
40.6250
39.1667
33.3333
0
32.6087
50.4348
37.7193
57.2917
0
43.3071
39.2000
62.6374
46.9565
47.2727
9.67742
29.5918
48.0000
35.2941
26.1682
28.5714
50.0000
38.3333
47.6190
24.3243
6.52174
6.08696
23.6842
0
30.9524
25.9843
28.8000
0
25.2174
15.4545
33.3333
13.2653
1.00000
28.5714
28.0374
31.6327
9.37500
22.5000
19.0476
0.6526
0.6501
0.7345
0.6359
0.6617
0.6796
0.8081
0.7063
0.5729
0.6569
0.4839
0.7387
0.5528
0.6081
0.6601
0.5752
0.6710
0.7355
0.6291
0.5597
0.7568
0.4224
0.4094
0.3444
0.4246
0.6979
0.5542
0.3419
0.3605
0.3805
0.3573
1.7460
0.8621
0.4321
0.2881
0.4446
0.4963
0.3498
0.3672
0.4215
1.0983
1.4194
1.2471
0.9004
1.5583
1.2893
1.1299
1.2940
1.4144
1.1965
1.1277
1.1160
1.2316
1.3552
1.0283
1.3277
0.8224
1.1257
1.2128
1.1939
4.2321
4.2037
4.2806
4.3705
4.6034
4.4517
4.2064
4.3068
4.3193
4.3128
4.2629
4.6003
4.3198
4.2056
4.5107
4.3421
4.3911
4.2577
4.2948
4.1534
1.8072
2.4402
1.9568
2.9418
2.5481
2.1953
3.0093
3.2320
3.7060
3.1772
3.4611
1.3125
2.3923
2.5894
3.3564
2.8030
2.3248
3.0276
2.3346
2.7736
3.3532
3.7644
4.4100
2.3833
3.6747
4.0695
3.5543
4.0469
4.2644
4.3626
3.8157
3.4498
3.8559
3.7058
3.4473
4.2607
3.0476
4.2271
2.3954
4.3099
182
behavior in the HTRS system. However, DFM only can works on the
control system that the linear parts and nonlinear parts are easy to
decoupled, which is not suitable for the close-coupled nonlinear
HTRS system, thus in this work, a hyperbolic approximation with
respect to tangent function is applied in the HTRS system. The
formulation of proposed hyperbolic approximation is described as
follows:
z
u an tan h
21
b
and once there are proper values setting of parameter a and b, the
output of hyperbolic function (dotted line) is nearly the same as
the output of saturation function (solid line), which is shown in
Fig. 8(b). In this study, we set um 20; z0 20, and the parameter
chosen of a and b has been computed by the PSO approach. The
bound of input signal is set as [ 50 50], and simulation step is
assigned to 0.001, the tness function is set as the error between
approximation output and reference output (i.e.f itnessi jci
zij, where ci is the approximation output, and zi is the saturation output). Experiment result shows that while the parameter
set a; b is assigned as 20:435; 16:104, the tness function gets
the smallest value. The iteration convergence process and nal
comparison of approximation output and saturation output are
shown in Fig. 9.
5.2. The dynamic analysis of the HTRS system
The block diagram of the HTRS system is illustrated in Fig. 3,
and the function of each component in Fig. 3 has been discussed in
details with Eqs. (1)(10). From Fig. 3, it can be seen that there are
two different input ports (i.e. given speed input and load torque
input) of this system, which means there should be discussed with
unload running condition and load running condition.
5.2.1. Unload running condition
In this case, there is no load turbulence interrupted the HTRS
and only given speed input acts on this system. The whole transfer
13.8
AGPSO
NSGAII
SPEAII
13.75
1.4
1.2
1
0.8
0.6
0.4
0.2
SF
Z-N
IMC
AGPSO
0
-0.2
10
12
14
16
18
20
Fig. 12. The generator speed output of the hydraulic turbine regulating system
under unload condition.
Table 4
Time domain specializations of traditional tuning methods under unload condition.
Specializations
SF
ZN
IMC
Er
Tr (s)
Mp (%)
0
0.4086
31.2821
0
0.4400
21.9150
0
0.3723
33.8184
1 em T w s
1
U
1 eqh T w s T a s en
22
13.7
x_ 1 x2 b1 y
x_ 2 a0 x1 a1 x2 b0 a1 b1 y
23
J2/ITSE
13.65
where
x1 x; a0 en =eqh T w T a ; a1 T a en eqh T w =eqh T w T a ; b0 ey =eqh
T w T a ; b1 ey em =eqh T a , and the state-equation of the servomechanism (i.e. Gc s) can be deduced as
13.6
13.55
13.5
x_ 3
13.45
13.4
13.35
50.3
50.35
50.4
50.45
50.5
50.55
50.6
J1/ISE
Fig. 11. Comparison of Pareto fronts of J1 and J2 under unload condition.
Table 3
Results of multi-objective algorithms with best compromise solution under unload
condition.
Method
kp
ki
kd
Er
Tr (s)
Mp (%)
AGPSO
NSGAII
SPEAII
18.8777
18.1032
18.6882
0.4460
0.4516
0.4529
5.8563
5.5566
5.8054
0
0
0
0.3427
0.3503
0.3449
5.9450
5.9958
6.0218
1
u x3
Ty
24
z
1
x3
an tan h
Ty
b
25
where
R
z kp ki =s kd sr x1 kp r x1 ki r x1 dt kd r_ x_ 1 , as
R
sume x4 r x1 dt, while the given speed output is a constant,
we can get
(
x_ 4 r x1
z kp r x1 ki x4 kd x2 b1 x3
26
ey mg mg eqh eey T w s
1 eqh T w sT a s en
28
183
max
f j f j
max
min
f j f j
max
30
min
where f j
and f j are the maximum and minimum values of the
jth objective function.
For each solution ith, the membership function i is calculated
as
nj 1 ij
31
n
i
m
i 1 j 1 j
max
j A f1;2;;mg
32
Table 5
Comparison of obtained Pareto-front solutions under load condition in trials.
Trials
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
QM (%)
SM
DM
AGPSO
NSGAII
SPEAII
AGPSO
NSGAII
SPEAII
AGPSO
NSGAII
SPEAII
54.5455
36.2205
38.0952
29.1339
35.0000
27.7778
42.0168
57.6087
57.5758
43.4783
29.5455
24.2188
38.4620
41.2706
37.0370
56.2500
34.8148
40.6504
36.0902
27.7311
0
40.1575
36.5079
41.7323
32.5000
46.0317
16.8067
0
0
29.5652
40.9091
45.3125
36.3640
39.6831
35.5556
0
41.4815
29.2683
30.0752
37.8151
45.4545
23.6220
25.3968
29.1339
32.5000
27.7778
41.1765
42.3913
42.4242
26.9565
29.5455
30.4688
25.1940
19.0473
27.4074
43.7500
23.7037
30.0813
33.8346
34.4538
0.6883
0.6148
0.5611
0.6356
0.6806
0.6505
0.7218
0.6730
0.6695
0.6793
0.6591
0.5645
0.6623
0.6361
0.6734
0.5831
0.6332
0.6295
0.7102
0.6776
0.3443
0.4889
0.4933
0.5862
0.4382
0.5698
0.5333
0.5483
0.3888
0.4174
0.4220
0.4102
0.4412
0.4869
0.4033
0.4706
0.3571
0.5205
0.4115
0.3919
0.3261
0.2908
0.2809
0.2788
0.2717
0.2861
0.3413
0.3586
0.3514
0.5850
0.2188
0.2697
0.4816
0.3033
0.3008
0.3496
0.2776
0.2301
0.2571
0.2417
14.6178
14.0370
13.9877
14.4205
14.7086
14.4817
14.5309
14.2259
14.7804
14.5669
14.5919
14.0495
14.5618
14.4389
14.5353
14.3495
14.6938
14.4592
14.4975
14.3953
14.7009
13.7359
13.8902
14.2093
13.2332
13.3163
14.4170
14.1911
14.4022
11.3948
12.4577
11.8552
14.1425
13.4409
13.2076
15.0797
12.2674
12.5026
13.0672
13.4058
14.0921
14.1380
14.2516
14.2602
14.2911
13.9937
14.0070
13.6536
14.1777
14.3665
14.1264
14.1261
14.3860
14.3426
14.3569
14.1030
14.1005
14.0032
14.1955
14.1359
184
1.5
116.5
AGPSO
NSGAII
SPEAII
116
115.5
J2/ITSE
115
114.5
114
113.5
113
112.5
6
18
2.
2
18
2.
2.
2
18
18
1.
1.
18
4
1.
18
2
18
1.
18
18
18
0.
112
J1/ISE
Fig. 13. Comparison of Pareto fronts of J1 and J2 under load condition.
Table 6
Results of multi-objective algorithms with best compromise solution under load
condition.
Method
kp
ki
kd
Er
Tr (s)
Mp (%)
AGPSO
NSGAII
SPEAII
10.7298
10.6731
10.6985
0.5047
0.4987
0.4924
4.7957
4.7935
4.7709
0
0
0
1.6501
1.6513
1.6548
1.0800
1.3770
1.2225
0.5
-0.5
SF
Z-N
IMC
AGPSO
10
12
14
16
18
20
Fig. 14. The generator output of the hydraulic turbine regulating system under load
condition.
Table 7
Time domain specializations of traditional tuning methods under load condition.
Specializations
SF
ZN
IMC
Er
Tr (s)
Mp (%)
0.0084
0.5286
47.5143
0
0.6016
42.4235
0
0.7727
12.9541
185
1.2
1.2
0.8
Turbine speed
Turbine speed
0.8
0.6
0.4
0.6
0.4
0.2
0
0.2
Ta=5.53
Ta=5.53
-0.2
Ta=5.73
Ta=5.93
Ta=5.73
Ta=5.93
-0.4
Ta=6.13
-0.2
10
12
14
16
18
-0.6
20
Ta=6.13
Time
10
12
14
16
18
20
Time
1.2
1.2
0.8
Turbine speed
Turbine speed
0.8
0.6
0.4
0.6
0.4
0.2
0
0.2
en=0.42
en=0.48
10
12
14
16
18
en=0.45
en=0.48
-0.4
en=0.50
en=0.50
-0.2
en=0.42
-0.2
en=0.45
-0.6
20
Time
10
12
14
16
18
20
Time
1.2
1.2
1
1
0.8
Turbine speed
Turbine speed
0.8
0.6
0.4
0.6
0.4
0.2
0
0.2
Tw=0.81
Tw=0.85
10
12
14
16
18
Tw=0.83
Tw=0.85
-0.4
Tw=0.88
Tw=0.88
-0.2
Tw=0.81
-0.2
Tw=0.83
20
-0.6
Time
10
12
14
16
18
20
Time
Fig. 15. Robustness analysis of obtained best compromise solution for AGPSO-based controller.
where di is the Euclidean distance between consecutive solutions in the obtained non-dominated set of solutions, and dmean
is the average of these distance. This metric allows us to
measure the uniformity of the spread of the points of the solution set.
(iii) Diversication Metric (DM): This metric measures the spread of
the obtained Pareto-front solution set and is dened by
D
q
i
i
N
i 1 maxxt yt
34
186
where J xit yit J is the Euclidean distance between the nondominated solution xit and the non- dominated solution yit .
7. Conclusion
In this paper, an ingenious multi-objective adaptive grid particle
swarm optimization (AGPSO) technique is applied to tune the PID
gains in a nonlinear HTRS system. Different conicting goals like
the integral of squared error (ISE) and the integral of time multiplied squared error (ITSE) are chosen as objectives and the AGPSObased approach is employed to generate the Pareto optimal set
of solutions. This proposed approach is used in the HTRS system
under unload and load condition respectively, in which the superiority of AGPSO optimized controller is shown by comparing
different multi-objective measurements with NSGAII optimized
controller and SPEAII optimized controller. Furthermore, a best
compromise solution approach, which selected a suitable solution
from the obtained Pareto front by a fuzzy-based membership value
187
[17] Pan I, Das S. Chaotic multi-objective optimization based design of fractional order
controller PI D in AVR system. Int J Electr Power Energy Syst 2012;43:393407.
[18] Coello CAC, Pulido GT, Lechuga MS. Handling multiple objectives with particle
swarm optimization. IEEE Trans Evol Comput 2004;8:25679.
[19] Chen Z, Yuan X, Ji B, Wang P, Tian H. Design of a fractional order PID controller
for hydraulic turbine regulating system using chaotic non-dominated sorting
genetic algorithm II. Energy Convers Manag 2014;84:390404.
[20] Kishor N, Saini R, Singh S. A review on hydropower plant models and control.
Renew Sustain Energy Rev 2007;11:77696.
[21] Chen Z, Yuan X, Tian H, Ji B. Improved gravitational search algorithm for
parameter identication of water turbine regulation system. Energy Convers
Manag 2014;78:30615.
[22] Kou P, Zhou J, Li C, He Y, He H. Identication of hydraulic turbine governor
system parameters based on bacterial foraging optimization algorithm. In:
Proceedings of the 2010 sixth international conference on natural computation
(ICNC), IEEE 2010. P. 3339343.
[23] Herreros A, Baeyens E, Pern JR. Design of PID-type controllers using multiobjective genetic algorithms. ISA Trans 2002;41:45772.
[24] Reynoso-Meza G, Garcia-Nieto S, Sanchis J, Blasco FX. Controller tuning by
means of multi-objective optimization algorithms: a global tuning framework.
IEEE Trans Control Syst Technol 2013;21:44558.
[25] Reynoso-Meza G, Sanchis J, Blasco X, Garca-Nieto S. Physical programming for
preference driven evolutionary multi-objective optimization. Appl Soft Comput 2014;24:34162.
[26] Eberhart RC, Kennedy J. A new optimizer using particle swarm theory. In:
Proceedings of the sixth international symposium on micro machine and
human science. New York, NY 1995. p. 3943.
[27] Yuan X, Su A, Yuan Y, Nie H, Wang L. An improved PSO for dynamic load
dispatch of generators with valve-point effects. Energy 2009;34:6774.
[28] Yuan X, Nie H, Su A, Wang L, Yuan Y. An improved binary particle swarm
optimization for unit commitment problem. Expert Syst Appl 2009;36:804955.
[29] Yuan X, Su A, Nie H, Yuan Y, Wang L. Unit commitment problem using enhanced
particle swarm optimization algorithm. Soft Comput 2011;15:13948.
[30] Tuba M, Bacanin N. Articial bee colony algorithm hybridized with rey
algorithm for cardinality constrained mean-variance portfolio selection problem. Appl Math Inf Sci 2014;8:283144.
[31] Wang H, Wang W, Wu Z. Particle swarm optimization with adaptive mutation
for multimodal optimization. Appl Math Comput 2013;221:296305.
[32] Panda S, Multi-objective PID. controller tuning for a FACTS-based damping
stabilizer using non-dominated sorting genetic algorithm-II. Int J Electric
Power Energy Syst 2011;33:1296308.
[33] Deb K, Agrawal S, Pratap A, Meyarivan T. A fast elitist non-dominated sorting
genetic algorithm for multi-objective optimization: NSGA-II. Lect Notes Comput Sci 1917;2000:84958.
[34] Das S, Das S, Pan I. Multi-objective optimization framework for networked
predictive controller design. ISA Trans 2013;52:5677.
[35] Zitzler E, Laumanns M, Thiele L. Improving the strength Pareto evolutionary
algorithm for multiobjective optimization. Evolutionary methods for design,
optimisation and control with application to industrial problems (EUROGEN
2001). International Center for Numerical Methods in Engineering (CIMNE);
2002; 95100.
[36] Shigang F, Qian A. Multi-objective reactive power optimization using SPEA2.
High Volt Eng 2007;33:1159.
[37] Garca S, Quintana D, Galvn IM, Isasi P. Portfolio optimization using SPEA2
with resampling. Intelligent data engineering and automated learning-IDEAL
2011. Lecture notes in computer science 2011;6936:12734.
[38] Tavakkoli-Moghaddam R, Azarkish M, Sadeghnejad-Barkousaraie A. A new
hybrid multi-objective Pareto archive PSO algorithm for a bi-objective job shop
scheduling problem. Expert Syst Appl 2011;38:1081221.