Professional Documents
Culture Documents
Design of PID Controller For Sun Tracker System Using QRAWCP Approach
Design of PID Controller For Sun Tracker System Using QRAWCP Approach
11 (2018) 133–145
___________________________________________________________________________________________________________
Design of PID controller for sun tracker system using QRAWCP approach
Abstract
In this paper, a direct formula is proposed for design of robust PID controller for sun tracker system
using quadratic regulator approach with compensating pole (QRAWCP). The main advantage of the pro-
posed approach is that, there is no need to use recently developed iterative soft computing techniques
which are time consuming, computationally inefficient and also there is need to know boundary of search
space. In order to show the superiority of the proposed approach, performance of the sun tracker system
is compared with the recently applied tuning approaches for sun tracker systems such as particle swarm
optimization, firefly algorithm and cuckoo search algorithm. The performance of the existing and pro-
posed approaches are verified in time domain, frequency domain and also using integral performances
indices. It is found that the performance is improved in transient, robustness, and uncertainty aspects in
comparison to recently proposed soft computing approaches.
Keywords: PID controller, QRAWCP, Servo motor, Sun tracker, Wind disturbance
cells. These panels are of different types like active, sky, partially clear sky, and cloudy sky which is im-
passive and time based solar tracker 6 . The active proved to 18%, 14% and 13% respectively. In liter-
solar trackers are of two types such as, single axis ature, sun trackers are controlled using PID, fuzzy
and dual axis 7 . In single axis, only one angle is var- 11 , adaptive neural 12 , ANFIS based proportional in-
ied and other is fixed, whereas dual axis, is like a tegral (PI) 13 and bio inspired PI controller 14 . As
two degree of freedom (DOF) system. In this, one is we know that, even today also PID controller is used
an azimuth angle (θ ) which is measured from clock- in industries 15 . This PID is designed either based
wise rotation across the base and other is tilt angle on analytically or graphical methods and also from
(α) which is measured from incoming sunlight in- past few decades stochastic optimization techniques
cidence to panel. This α is also known as eleva- 16 . Various papers exist on tuning of PID controller
tion angle measured from horizon. In literature, it 17 . The first paper on PID tuning came in the litera-
is found that an additional angle, i.e., zenith angle ture in 1942 by Ziegler Nichols (ZN) 18 . In this ap-
is also considered, which is measured from vertical proach, the direct formula of deterministic approach
direction. This is shown in Fig.1. is given for the tuning of PID parameters. The ex-
tension of this work is proposed by Cohen-Coon
method in 1952 19 , which is based on ZN technique
along with dead time consideration. Recently, num-
ber of techniques has also been proposed and used in
various applications such as process control 20 , dc-
dc converters 21 and robustness for random pertur-
bations 23,22 . Similar to these analytical approach,
graphical methods have been reported for tuning of
PID parameters such as root locus, bode plot, stabil-
ity boundary locus 25,24 etc.
programmed with the controller for future positions. other bird’s nest for hatching, but the probability is
In 10,1 , researchers analyzed efficiency in terms of that, host bird may find the cuckoo’s egg and throw it
output power between fixed panel and solar tracker out. Here, the cuckoo’s egg is the objective solution
system. This results are carried on physical fix PV for the problem. In 31 optimal control technique is
and solar tracker system which shows improvement proposed, which is simple computational for convex
in power output during different conditions like clear system. In 32 , it is shown that the optimal and robust
134
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
PID controller can be designed using pole placement tilted angle (α) which measures the inclination angle
approach. The method has been applied to inverted of sunlight. As sun tracker is a non-interacting sys-
pendulum problem. However, no direct formula is tem, controller designed for single axis will be the
given for tuning of PID controller. Therefore, in this replica for another axis also. Therefore, the analysis
paper, a new method has been proposed for obtain- has been carried out on a single axis sun tracker sys-
ing direct formula for designing PID controller us- tem as shown in Fig. 3. Mathematical model of sun
ing quadratic regulator approach with compensating tracker is determined using basic laws of physics.
pole (QRAWCP). The advantage of this method is For the major hardware parts such as DC servo mo-
that there is no need for any iterative procedure for tor, it’s speed transfer function G1 is given in (1) and
designing of PID controller. The validity of the pro- G2 is for position control in (2). A gear ratio (N)
posed approach is carried out by checking robust- and other parameters with values are listed in Ta-
ness and optimality. The results are compared with ble 1. Now considering single axis model as shown
the recently proposed standard soft computing meth- in Fig. 3. Applying Kirchhoff law to DC motor, we
ods. get velocity (ω(s)) control transfer function as
The rest of the paper is organized in the fol-
lowing manner: In section 2, parameters of the sun ω(s) Kt
G1 (s) = = (1)
tracker system and its model are described. In sec- Va (s) (Ra + La s)(Js + b) + (Kt Kb )
tion 3, QRAWCP approach to design PID controller
where, Va (s) is an armature input voltage. By inte-
is presented. In section 4, simulation results of pro-
grating (1), the position control transfer function can
posed approach is presented. Finally, in the last sec-
be written as,
tion, conclusion and future scope are discussed.
θy (s)
G2 (s) =
2. Model of sun tracker system Va (s)
(2)
Kt
=
La Js + (bLa + Ra J)s2 + (Ra b + Kt Kb ) s
3
Table 1: Parameters of sun tracker system
Moreover, the sun tracker is interfaced with some
Parameter Value Unit additional components such as error discriminator
(Ke ), amplifier gain (K), servo amplifier (Ks ) and
Error discriminator(Ke ) 0.001 V/rad gear ratio (N). If all these parameters are consid-
Amplifier gain(K) 10000 V/V ered, then open loop transfer function for this sun
Servo amplifier(Ks ) 1 V/V tracker system becomes,
Armature resistance(Ra ) 6.25 ohm
Armature inductance(La ) 0.001 H θy (s)
G(s) =
Torque constant(Kt ) 0.01125 Nm/A θr (s)
(3)
Back emf constant(Kb ) 0.0125 Nm/A Ks KKe Kt N
=
Inertia of motor rotor(J) 1 × 10−6 kgm2 /rad La Js + (bLa + Ra J)s2 + (Ra b + Kt Kb ) s
3
135
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
and y = θy , respectively. Using this, the state space (Ki ) and derivative (Kd ) (PID) controller C(s) can be
model can be written as written as
⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
ẋ1 0 1 0 x1 0 Kd s2 + sKp + Ki
⎣ ẋ2 ⎦ = ⎣ 0 C(s) = (6)
0 1 ⎦ ⎣ x2 ⎦ + ⎣ 0 ⎦ u s
b3 a4
ẋ3 − b1 − b1 − bb21
b4
x3 b1 Step 2: The closed-loop characteristic equation for
⎡ ⎤
x PID controller C(s) and plant G(s) is given as,
1
y = 1 0 0 ⎣ x2 ⎦
Δ(s) = 1 + G(s)C(s) (7)
x3
(5) Simplifying (7) and equating to zero, we get,
b2 3 b3 + a4 Kd 2
3. Proposed QRAWCP approach to design PID s + s +
4
s
b1 b1
controller for sun tracker system (8)
b4 + a4 Kp a4
+ s + Ki = 0
The quadratic regulator approach with compensa- b1 b1
tion pole (QRAWCP) approach to tune PID con- Step 3: The state space form for sun tracker sys-
troller is described as follows. tem is given in (5). Using this, design of PID by
Step 1: The transfer function of the sun tracker sys- QRAWCP approach are given in step 4 to step 7.
tem model G(s) is given in (4), and it’s parameters They are given below.
are given in Table 1. The proportional (K p ), integral Step 4: Determination of performance index in
136
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
terms of initial condition: (10), the performance index (10) can be written as,
The quadratic regulator approach is an optimal state ∞
feedback controller which can be designed to mini- Jl = xT Qx +(Kl x)T R (Kl x) dt
0 (12)
mize a specific quadratic cost function, also known ∞
as performance index (PI). The PI is designed for = xT Q + KlT RKl x dt
0
constraints like control voltage (u), output signal (y),
error (e) or unconstrained objectives of linear time Let,
invariant (LTI) system . The optimal control vector d T
u(t) is obtained by using the following equation: xT Q + KlT RKl x = − x Px
dt (13)
u(t) = −Kl x(t) (9) = −xT Pẋ − ẋT Px
On putting (11) in (13) and then substituting in (12),
Here, unconstrained optimal action is considered. we get,
Therefore, cost function of the system is defined as
Jl = −xT P (A − BKl ) + (A − BKl )T P x (14)
∞
Jl = xT Qx +uT Ru dt (10) In (14), P must be positive definite matrix. By com-
0 paring (13) with (14), we get,
where Q ∈ Rl×l and R ∈ Rm×m are symmetric pos- P (A − BKl ) + (A − BKl )T P = − Q + KlT RKl
itive definite matrices. Using (9), the LTI system (15)
equation becomes, As (A − BKl ) is a stable matrix. Therefore, solv-
ing for a positive definite matrix P which will satisfy
ẋ = (A − BKl )x (15), the cost function can be evaluated as,
(11)
ẋ = Ax
∞
Jl = xT Q + KlT Kl x dt (16)
where, 0
= (A − BKl )
A
From (13), we can write as,
If A and B are controllable, then optimal state feed- ∞
back controller can be designed. Thus A
have eigen- Jl = −xT Px 0
(17)
values on the left of the s-plane. Substituting (9) in = −x (∞)Px(∞) + xT (0)Px(0)
T
137
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
As the system (11) is stable, eigenvalues of (17) From (23), the closed-loop characteristic equation
must have negative real part. Therefore, x(∞) → 0. can be written as
Thus, we get Jl = xT (0)Px(0). It is obtained in terms
of initial condition. a4 a4 2 a4 a4
Step 5: From (15), the minimization of Jl gives Kl s + p33 + 1
3
s + p23 + 1 s+
b1 b1 b1 b1
by using feedback control law u = −Kl x. The feed-
a4 a4
back gain Kl is found by using p13 + 1 =0
b1 b1
(24)
Kl = R−1 BT P (18)
Step 9: The order of closed-loop system (8) is of
and further simplifying, we get Riccati equation as, fourth order and (24) is of third order. Therefore,
in order to compare these two equations, we need to
AT + PA − PBR−1 BT P + Q = 0 (19) add one pole. The methodology of adding pole is
explained below.
In (19), Q and R are selected in such a way that Let us consider fourth pole (s + λ4 ) on the left half
Q = diag(q11, q22, q33) is q11 > q22 > q33 > 0 and of the s-plane. Then (24) can be written as,
R = χ T χ > 0, where χ is non singular matrix.
Step 6: Using Riccati equation (19) and state model a4 a4 2 a4 a4
s + p33 + 1
3
s + p23 + 1 s
from (5), positive definite matrix P is obtained b1 b1 b1 b1
which is given below. a4 a4
+ p13 + 1 (s + λ4 ) = 0
⎡ ⎤ b1 b1
p11 p12 p13 (25)
P = ⎣ p12 p22 p23 ⎦ (20) The above (25) can also be written as
p13 p23 p33
a4 a4 3
Step 7: Using (18), state feedback control gain Kl is s + λ4 + p33 + 1
4
s +
b1 b1
obtained as
a4 a4 a4 a4
λ4 p33 + 1 p23 + 1 s2 +
b1 b1 b1 b1
Kl = [R]−1 BT [P]
⎡ ⎤ a4 a4 a4 a4
p p12 p13 λ4 p23 + 1 p13 + 1 s+
a4 ⎣ 11 (21) b1 b1 b1 b1
= [1] 0 0 p12 p22 p23 ⎦
b1 a4 a4
p13 p23 p33 λ4 p23 + 1 =0
b1 b1
(26)
a4 a4 a4 If (26) is compared with (8), λ4 is calculated as
Kl = p11 p23 p33 (22)
b1 b1 b1
a4 a4 b2
new
Step 8: Using state feedback control law for A, λ4 = − p33 + 1 − (27)
b1 b1 b1
system matrix can be written as,
138
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
139
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
case 2 and case 3 for without disturbance, with in- peak time (t p ) and steady state error (ess ). These
put disturbance and with output disturbance, respec- are tabulated in Table 3. From this table, it is found
tively. that, the proposed approach shows improved results
Case 1: Fig. 4 shows step response with controller in comparison to other PIDs. Fig. 5 shows control
for sun tracker system without disturbance. From signals using saturation limit of ±50V . It is found
this figure, it shows that performance of the pro- that overall control output of proposed QRAWCP
posed QRAWCP technique is better in comparison approach is less in comparison to other techniques.
to other PID approaches which are based on soft The performance of proposed PID is verified for dis-
computing technique. turbance di(t) and do(t), where di(t) is input distur-
bance and do(t) is output disturbance. Both these
disturbances are written mathematically in (33).
1 ⎧
⎨ 0, t < 2 or t > 4
0.8 d(t) = 0.15 × (sin 4πt + cos 2πt
Amplitude(V)
1 ⎩
+ sin πt + sin 4πt) − 0.5, 2 t 4
0.6 0.98 (33)
0.4 0.2 0.4 The output response of these equations are shown in
QRAWCP PID
CSA PID
Fig. 6. The performance analysis of the system in
0.2 PSO PID case of input disturbance and output disturbance is
FFA PID presented in case 2 and case 3, respectively.
0
0 0.5 1
Time(sec)
trollers
-0.5
50
QRAWCP PID
40 CSA PID
-1
PSO PID
2
Amplitude(V)
FFA PID
30 1 0 1 2 3 4 5
20 0 Time(sec)
0 0.5 Figure 6: Disturbance (d(t)) applied to system
10
140
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
141
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
where, S∞ 2 ensure robustness. However, DM is gin, better sensitivity is compare to other meth-
calculated using frequency domain analysis which is ods. Finally, the comparative studies of proposed
given by (37) QRACWP approach with existing approaches30 are
carried out by calculating integral performance in-
PM ◦ Π dices. They are explained below.
DM = (37)
180◦ × ωgc
The above robustness has been carried out for pro- 4.3. Integral performance indices
posed method and existing controllers. They are
The commonly used performance measures are In-
given in Table 4.
tegral Squared Error (ISE), Integral Absolute Error
(IAE) and Integral Time-weighted Absolute Error
(ITAE). The mathematical form is given as
1.5
∞
ISE = (θr − θy )2 dt
Amplitude(V)
0
1
∞
IAE = |(θr − θy )| dt (38)
0 ∞
0.5 1 QRAWCP PID
ITAE = t |(θr − θy )| dt
CSA PID 0
0.9 PSO PID
FFA PID
2 2.2 where, (θr − θy ) is the error between reference input
0
0 1 2 3 4 5 position and measured output position (θy ) at time
Time(sec) ‘t’. All integral errors are calculated for sun tracking
system which are shown in Table 5. From this table,
Figure 9: Step responses of system with output dis- it is observed that, the proposed QRAWCP method
turbance for design of PID has minimum error in comparison
to PSO PID30 , FFA PID30 and CSA PID 30 .
FFA PID
tem are not constant. They vary from minimum
to maximum value. This is primarily due to non-
0 linearity, environmental change and also due to age-
ing. The main objective of controller design is that it
5 should work even though there exist uncertainties in
0 the system. Here, for sun tracker system model, its
-5
parameters are represented in terms of uncertainties
2 3 4
-50 as
0 1 2 3 4 5
Time(sec) K e = Ke ± ΔKe , K = K ± ΔK, K s = Ks ± ΔKs ,
Figure 10: Control input responses with output dis- La = La ± ΔLa , Ra = Ra ± ΔRa , K t = Kt ± ΔKt ,
turbance J = J ± ΔJ, b = b ± Δb, K b = Kb ± ΔKb
It is observed that, proposed PID controller Here, 50% uncertainty is considered in all the pa-
shows better stability margin, reduces delay mar- rameters of the system. For this uncertainty, the up-
142
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
QRAWCP PID
∞ 2359.10 69.2549 0.0004 1.2293 0.0005
(Proposed)
PSO PID 57.8081 8.8371 69.9214 407.8013 1.2320 0.1381
FFA PID 56.8303 9.0061 69.6572 389.6705 1.2356 0.1350
CSA PID 56.5502 9.2369 68.8322 388.9042 1.2425 0.1301
per and lower bounds are shown below. Similar to section 4.1, the performance analy-
sis is carried out for both lower and upper bounds.
Ke ∈ [0.00050, 0.00150] From Fig. 11 and Fig. 12, it is observed that the per-
K ∈ [5000.00, 15000.00] formance of the proposed control is better in com-
parison to existing controllers except control input
Ks ∈ [0.50, 1.50] is slightly more at early part of the response. Fur-
La ∈ 5 × 10−4 , 1.50 × 10−3 ther, performance in case of parametric uncertainly
is carried out by determining performance indices in
Ra ∈ [3.125, 9.375]
case of without and with input disturbances. They
Kt ∈ [0.005625, 0.016875] are shown in Table 5 and 6, respectively. Both these
J ∈ 5 × 10−7 , 1.50 × 10−6 tables indicate that the proposed controller approach
is better in comparison to existing controllers.
b ∈ 0.5 × 10−6 , 1.50 × 10−6
Kb ∈ [0.00625, 0.01875]
50 50
1 QRAWCP PID QRAWCP PID
40 CSA PID 1 40 CSA PID
0.8 PSO PID PSO PID
Amplitude(V)
Amplitude(V)
Amplitude(V)
Amplitude(V)
30
FFA PID 0.8 30 2 FFA PID
1.02
0.6 1 1
0.6
20 0.98 20 0
0.4 QRAWCP PID 0.4 QRAWCP PID
CSA PID 10 2 3 4 CSA PID 10 2 3 4
0.2 PSO PID 0.2 PSO PID
FFA PID 0 FFA PID 0
0 0
0 0.5 1 0 0.5 1 0 1 2 3 4 5 0 1 2 3 4 5
Time(sec) Time(sec) Time(sec) Time(sec)
(a) Lower bound (b) Control input (a) Lower bound (b) Control input
50 50
QRAWCP PID QRAWCP PID
1 40 CSA PID 1 40 CSA PID
PSO PID PSO PID
Amplitude(V)
Amplitude(V)
Amplitude(V)
Amplitude(V)
0.8 30
FFA PID 0.8 30
FFA PID
2
0.6 0.6 1 1
20 20 0
0.4 QRAWCP PID 0.4 0.95 QRAWCP PID
CSA PID 10 CSA PID 10 2 3 4
0.2 PSO PID 0.2 2 3 4 PSO PID
FFA PID 0 FFA PID 0
0 0
0 0.5 1 0 0.5 1 0 1 2 3 4 5 0 1 2 3 4 5
Time(sec) Time(sec) Time(sec) Time(sec)
(c) Upper bound (d) Control input (c) Upper bound (d) Control input
Figure 11: Response for 50% parametric uncertain- Figure 12: Response for 50% parametric uncertain-
ties in system without disturbance ties with input disturbance in system
143
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
Table 6: Integral performance indices for 50% perturbation with input disturbance
5. Conclusion References
144
International Journal of Computational Intelligence Systems, Vol. 11 (2018) 133–145
___________________________________________________________________________________________________________
145