The feedback (closed-loop) control theory was applied
to derive a global control algorithm for an irrigation canal
with a converged channel junction. The control problem
was formulated as an optimal control problem directly
using the discrete linear quadratic regulator (LQR)
theory. The hydraulic formulation of the flow at a channel
junction was complicated because of energy losses, lateral
flow approaching angles and flow mixing. Using lateral
inflow angles to the downstream channel instead of
channel junction angles, based on the momentum flux
contribution a momentum equation at channel junctions
was derived. The performance of the model was
evaluated in terms of deviations in the depths of flow and
the upstream gate opening. The variation in the upstream
gate opening and the deviations in the depth of flow in
the canal reach were not erratic during the simulation
period. The system either approached a constant value or
came close to the equilibrium condition. This indicates
that the optimal control theory (feedback loop) is still
applicable on irrigation canals with a combining channel
junction. Although a single canal reach was considered in
this study to demonstrate in detail a procedure to derive
a control algorithm in the presence of a converging
channel junction, the same procedure can be used to
derive a feedback control algorithm to run irrigation
canals with diverging channel junctions or channel
junctions with multiple reaches.
10 trang |
Chia sẻ: huongthu9 | Lượt xem: 408 | Lượt tải: 0
Bạn đang xem nội dung tài liệu Simulation of a Feedback Control Technique Through Irrigation Canal Junctions, để tải tài liệu về máy bạn click vào nút DOWNLOAD ở trên
Introduction
In recent years, the better operation of canals has
become an increasingly important and attainable goal.
Improved operation of irrigation conveyance systems will
improve service to water users, conserve water through
increased efficiency, reduce operation and maintenance
costs, increase delivery flexibility, and provide more
responsive reactions to emergencies. Timely delivery of
the required quantity of water is necessary for improved
agricultural production. The supply-oriented operational
concept has not been able to provide the needed flexibility
in terms of water quantity and timing to achieve
improved crop yields and water-use efficiency. The
demand-oriented operational concept bases operations on
downstream conditions (Bureau of Reclamation, 1995).
Most irrigation systems should use this downstream
concept. The canal system should be operated to satisfy
downstream needs, responding to what is taken out of
the system rather than to what is put in. A significant
portion of wasted irrigation water is attributed to
inefficient canal system operation. Feedback control
systems can improve the efficiency of the operation of
Turk J Agric For
29 (2005) 391-400
' TBÜTAK
391
Simulation of a Feedback Control Technique Through
Irrigation Canal Junctions
mer Faruk DURDU
Adnan Menderes niversitesi, Ziraat Fakltesi, TarÝmsal YapÝlar ve Sulama Blm, 09100 AydÝn - TURKEY
Received: 02.11.2004
Abstract: A linear quadratic controller based algorithm was developed for simulating the dynamics of a single-reach irrigation canal
with a channel junction. Using the concepts of feedback control theory, an expression for an upstream gate opening of an irrigation
canal reach with a channel junction operated based upon a constant-level control was obtained. In the derivation, the canal reach
between 2 gates was divided into 5 nodes, and the finite difference forms of the continuity and momentum equations were written
for each node. The Taylor series was applied to linearize the equations around equilibrium conditions. At the third node of the canal
reach, a channel junction occured and the equations were derived based on the channel junction parameters. The hydraulic
description of flow at channel junctions is difficult because of flow approaching angles, energy losses and turbulence. An example
problem with a single pool was considered for evaluating the technique used to design a linear quadratic controller (LQR) for
irrigation canals with a channel junction. Considering the computational complexity and the accuracy of the results obtained, the LQR
feedback control theory was found to be adequate for irrigation canals with channel junctions.
Key Words: channel junctions, linear quadratic controller, mathematical modeling, open-channel flow
Sulama KanallarÝnÝn BirleßtiÛi Noktalarda Geri-dnßml Kontrol TekniÛinin Simlasyonu
zet: Sulama kanallarÝnÝn birleßtiÛi noktalarda suyun kontrolu iin doÛrusal karesel control tekniÛine dayalÝ bir algoritma
gelißtirilerek simlasyon yapÝlmÝßtÝr. Geribeslemeli control tekniÛi kullanÝlarak kanallarÝn birleßtiÛi noktanÝn memba kÝsmÝndaki
kapaÛÝnÝn aÝlÝp kapanmasÝ iin sabit-seviyeli control methoduna gre bir ifade gelißtirilmißtir. Bu ifadenin ÝkartÝlmasÝnda gz nne
alÝnan havuz (iki kapak arasÝnda kalan kÝsÝm) beß boÛuma blnmß ve her bir boÛum iin sonlu farklar ile ifade edilen sreklilik ve
momentum eßitlikleri ÝkartÝlmÝßtÝr. Taylor serileri de bir denge konumu gz nne alÝnarak bu eßitliklere uygulanmÝßtÝr. Havuzun
nc boÛumunda kanal bir baßka kanal ile birleßtiÛinden sreklilik ve momentum eßitlikleri de buna gre ÝkartÝlmÝßtÝr. Birleßen
kanaldan gelen akÝßÝn yaklaßÝm aÝsÝ, enerji kayÝplarÝ ve trblansÝ birleßim yerindeki akÝßÝn hidrolik tanÝmÝnÝ zorlaßtÝrmaktadÝr.
Gelißtirilen doÛrusal karesel kontrol algoritmasÝnÝ deÛerlendirmek iin bir rnek problem zlmßtr. Hesaplamalardaki zorluklar
ve simlasyondan elde edilen sonular gz nne alÝnarak, doÛrusal karesel kontrol tekniÛinin birleßen kanallar iin yeterli ve uygun
bir teknik olduÛu sonucuna varÝlmÝßtÝr.
Anahtar Szckler: kanallarda birleßmeler, doÛrusal karesel control tekniÛi, matematiksel modelleme, aÝk kanallarda akÝm.
*Correspondence to: odurdu@adu.edu.tr
irrigation canal networks and increase the dependability
of water supply at farm level. With automatic feedback
controllers, irrigation distribution networks can have
better robustness, reliability and safety, as well as
minimize water waste and reduce the cost of operation
and maintenance when compared with manual operations
(Bureau of Reclamation, 1995).
Feedback control systems for gate structures and
turnouts have been applied to irrigation canal systems in
the past. Over the years, the concepts of feedback control
theory have been applied for deriving control algorithms
for the real-time control of irrigation canals (Balogun et
al., 1988; Begovitch and Ortega, 1989; Reddy et al.,
1991; Goussard, 1993; Rodellar et al., 1993; Malaterre,
1994; Malaterre, 1998). Balogun et al. (1988), Garcia et
al. (1992), and Durdu (2003, 2004a, 2004b) simulated
feedback control concepts for deriving global control
algorithms for irrigation canals. However, those
algorithms dealt with an irrigation canal without channel
junctions and they were designed for an independent
irrigation canal. Generally, the canal network is a
relatively large system. Accordingly, even using numerical
method, simultaneous prediction of flow variables is a
challenging problem. Because of this difficulty, the above
control algorithms consider a single canal in an irrigation
network. This assumption may be used in certain
situations. However, the flow phenomenon in an
irrigation network is much more complicated than that in
a single independent canal. The hydraulic description of
the flow at the channel junction is complicated and
difficult because of the high degree of flow mixing,
separation, turbulence and energy losses (Choi, 1991).
Furthermore, the approaching angles of flow into the
main canal are important for the numerical solution of
flow equations. Therefore, the flow equations at a
channel junction node will be different than those at other
regular nodes. The objective of this paper is to present a
discrete feedback control scheme for the operation of
irrigation canals with a converging channel junction.
Model Development
The governing equations of continuity and momentum
(or energy) express the phenomena of the water flow in
the irrigation canal network. However, they furnish no
direct answers as to the values of discharge and water
depths, which are solutions of the basic flow equations,
which are functions of time and space. The theoretical
analysis is so complicated that solutions can only be
obtained by a numerical method. In the operation of
irrigation canals, in order to maintain the flow rate into
the laterals close to the desired value, decisions regarding
the opening of gates in response to random changes in
the water withdrawal rates into lateral canals are
required. This is accomplished by maintaining the depth
of flow or the volume of water in a given pool at a target
value. This problem is similar to the process-control
problem, in which the state of the system is maintained
close to the desired value by using real-time feedback
control. In this study, only a single reach of a canal with
a channel junction was considered. The reach was divided
into N-1 sub-reaches of length D x. There are 5 nodes (N
= 5) in the canal reach (Figure 1). For convenience, the
spacing between the nodes (D x) was assumed constant.
Using a finite-difference approximation, the continuity
and momentum equations are written for each node. At
the third node, there is a converging channel junction into
the main irrigation canal. Because of the approaching
angles of flow, energy loss, turbulence, and flow mixing,
the continuity and momentum equations will be different
at this particular node than those at other regular nodes.
In this study, it was assumed that the flow was subcritical
throughout the canal and the canal was horizontal and
trapezoidal.
Gradually varied unsteady channel flow is described by
2 basic partial differential equations (Saint-Venant
equations): the continuity and momentum equations. The
continuity equation considers the conservation of mass
and the momentum equation expresses the conservation
of momentum. The Saint-Venant equations, presented
below, are used to model flow in irrigation canals:
(1)
(2)
in which Q = discharge, (m3 s-1); A = cross-sectional area
of flow, (m2); ql = lateral flow, (m
2 s-1); z = water surface
elevation, (m); t = time, (s); x = longitudinal direction of
channel, (m); g = gravitational acceleration, (m2 s-1); So =
canal bottom slope (mm-1); and Sf = the friction slope,
(mm-1). In deriving Eq. (2), the effect of the net
1
A
¶ Q
¶ t
+ 1
A
¶ Q
¶ x
Q2
A
= gS0 - gSf - g
¶ x
¶ z
¶ A
¶ t
- ¶ Q
¶ x
- q1 = 0
Simulation of a Feedback Control Technique Through Irrigation Canal Junctions
392
acceleration terms stemming from removal of a fraction
of the surface stream was assumed negligible. In Eqs. (1)
and (2), the spatial derivatives were replaced by finite
difference approximations by dividing the pool into a few
segments (N number of nodes). A central difference
scheme was used for the interior nodes (1<j<N), and a
forward difference and a backward difference were
applied to the first and last nodes, respectively. To solve
those equations, the boundary conditions were expressed
in terms of the continuity equation (Reddy, 1996):
Qi-l,N = Qi,l = Qgi (3)
and the gate discharge equation
(4)
in which Qi-1,N = flow rate through the downstream gate
(or node N) of pool i-1, (m3 s-1); Qgi = flow rate through
upstream gate of pool i, (m3 s-1); Qi,1 = flow rate through
the upstream gate (or node 1) of pool i, (m3 s-1); Cdi =
discharge coefficient of gate i; bi = width of gate i, (m);
ui = opening of gate i, (m); zi-1,N = water surface elevation
at node N of pool i-1, (m); zi,1 = water surface elevation
at node 1 of pool i, (m); and i = pool index (i = 0 refers
to the upstream constant level reservoir). In Eq. (4), the
change in the bottom elevation of the canal across the
gate was assumed negligible. At the upstream and
downstream nodes of the canal reach, the momentum
equation was replaced by the gate discharge equation, Eq.
(4), which is basically expressed in terms of depth of flow
at these nodes. At each intermediate node, there are 2
unknowns (Q and z). At the upstream and downstream
ends, there are 2 unknowns per node (z and u). With N
as the number of nodes in the reach, the total number of
equations for the reach is 2(N-2) + 2. Therefore, to solve
the set of equations, the 2 boundary conditions must be
specified.
The precise hydraulic description of the flow at a
channel junction is complicated and difficult because of
the high degree of flow mixing, separation, turbulence
and energy losses (Choi, 1991). The influence of the
approaching angles (g ) is different, depending upon the
discharge ratio between the upstream and downstream
(Choi, 1991). The maximum depth difference between
the upstream and downstream occurred near the depth
ratios of 0.4 to 0.6 in the 30 and 60 degree junctions. On
the other hand, the maximum depth difference occurred
near the depth ratio of 1.0 in the 90 degree junction. It
is noted that the relative upstream flow depth is a
function of discharge distribution for various angles of
junction (Choi, 1991). In a canal junction, the process of
flow mixing and turbulence causes a loss of energy in
addition to the friction loss that occurs in the canal itself.
The other major process that effects a loss of energy at a
junction is flow separation. The separation zone is
developed downstream from the entrance of a lateral
channel. The momentum of the lateral canal ensures that
the flow detaches from the sidewall as it enters the main
canal and a separation zone of lower speed with
circulating flow is introduced (Choi, 1991). Best and Reid
(1984) conducted experiments to determine the
maximum width and length of flow separation in the main
canal at the section downstream of the junction. Their
experiment was based on 15, 45, 70 and 90 degree canal
junctions, with Froude numbers between 0.1 and 0.3.
Qgi = Cdibiui 2g(zi-l,N - zi,l)
. F. DURDU
393
upstream pool
downstream pool
z(i+1,1)
u(i+1)
lateral withdrawal
gate 2gate 1
z(i-1,N)
u(i)
Q
1 2 3 4 5
Nodes
D x
Figure 1. Schematic of a single canal reach.
The experiment showed that the separation width
increases with an increase in discharge ratios between the
lateral canal and the downstream canal (Choi, 1991).
The accurate description of the junction hydraulics is
important for realistic and reliable simulation of a
feedback controller in an irrigation canals network. In
addition to the continuity relationship, the dynamic
relationship at the channel junction can be represented by
either the energy or momentum equations. In the past,
the momentum equation was rarely used because the
terms of this equation are vector quantities, and the
momentum contribution from upstream channels,
especially from the laterally connected upstream channels,
is usually difficult to define (Choi, 1991). Even though
difficulties exist in applying the vector quantities in the
momentum equation for channel junctions, the
conservation of momentum at channel junctions may be
necessary to consider the effects of channel junction
angles and to account for other upstream channel
discharge effects. Using lateral inflow angles to the main
canal instead of channel junction angles, based on the
momentum flux contribution, a momentum equation at
channel junctions can be derived.
The hydraulic description of flow at channel junctions
can be represented by the conservation of momentum
between upstream and downstream canal segments, and
the junction continuity equation. The momentum
equation can be derived by considering the contribution
of momentum flux by the upstream and downstream
channels. The momentum flux in a downstream channel is
equal to the sum of momentum flux from the upstream
channels. As illustrated in Figure 2, g is the lateral inflow
angle into the downstream canal as opposed to the
channel junction angle a . The relationship between g and
a in channel junctions has been studied experimentally
(Best and Reid, 1984). The lateral inflow angle (g ) to the
downstream channel was shown by Best and Reid (1984)
to be dependent upon discharge ratios between lateral
inflow and downstream outflow and channel junctions
angles (a ). The momentum and continuity equations at
channel junctions can be presented as follows (Choi,
1991):
(5)
(6)
where Ax is the projection of the cross-sectional area in
the longitudinal x-direction in the one-dimensional
equation (Ax = A*cos g ), Vx is the velocity component in
the x direction, which is longitudinal direction in a
dimensional equation (Vx = V*cos g ), and Qx is the
discharge in the x direction. The turnouts were assumed
to be located immediately upstream of the last node in
the canal reach (Figure 1). Although the lateral
withdrawals were concentrated at one point, for
modeling purposes, it was assumed to be uniformly
distributed between the adjacent nodes, and was related
to ql of Eq. (1) as follows. The mathematical
representation of flow through turnout structures is
given as follows (Reddy, 1996):
(7)
q1 =
qi,n∑
n = 1
Oi,j
s
¶ Qx
¶ t
+ ¶ (VxQ)
¶ x
+ gAx
¶ z
¶ x
- S0 + Sf = 0
¶ Ax
¶ t
+ ¶ Qx
¶ x
= q1
Simulation of a Feedback Control Technique Through Irrigation Canal Junctions
394
g
a1 2 53 4
Upstream gate Downstream gate
Nodes
Qj 2
Qci
Qj+1
Main canal
Side canal
D x
Figure 2. Channel junction profile.
where s = D x in the case of a backward/forward
difference scheme; and s = 2D x in a central difference
scheme; D x = distance between 2 nodes, (m); qi,n =
withdrawal rate from outlet n of pool i, (m3 s-1); and Oi,j
= number of outlets represented around node j of pool i.
The Saint-Venant open-channel equations were
linearized about an average operating condition of the
canal to apply the linear control theory concepts to the
problem (Balogun, 1985). Linear control theory is well
developed and is easier to apply than nonlinear control
theory. To apply linear control theory, Eqs. (1, 2, 5, 6)
were linearized about an average operating point. In this
study, the derivation of equations for regular nodes was
not emphasized. For details about the derivation of the
equations for regular nodes, refer to Balogun (1985),
Reddy (1990), or Durdu (2003). Using the Taylor series
around the average operating point, and truncating terms
higher than the first-order, the deviation variables were
obtained as follows:
d Qi,j = Qi,j — Q0i,j (8)
d zi,j = zi,j — z0i,j (9)
d ui = ui — u0i (10)
d ui+1 = ui+1 — u0i+1 (11)
in which Q0, z0, and u0 represent the flow rate, water
surface elevation, and gate opening, respectively, at the
equilibrium condition; and d Q, d y, and d u represent the
deviations in flow rate, water surface elevation, and gate
opening, respectively, from the equilibrium condition. In
Eqs. (8-11), i is the index of the canal reach, and j is the
index of the node in the canal reach. Application of the
finite-difference technique, and substitution of Eqs. (8-
11) into Eqs. (1) and (2) results in a set of linear
equations (Reddy, 1990) for the regular nodes. For the
channel junction node, Eqs. (8-11) were substituted into
Eqs. (5) and (6) as follows (Choi, 1991):
Continuity equation at channel junction:
q [Q+j+1 — (d Q
+
j + d Q
+
cj*cosg )/D x] + (1 - q )[d Qj+1 —
(d Qj + d Qcj*cosg )/D x] + (1/2 D t)bj+1/2
(d z+j + d z
+
j+1 - d zj - d zj+1) =
q ql
+
j+1/2 + (1- q ) ql j+1/2 (12)
Momentum equation at channel junction:
(1/2 D t)[(d Q+j + d Q
+
cj*cosg ) - (d Qj + d Qcj*cosg ) +
d Q+j+1 - d Qj+1) + (q / D x)[d Vj+1d Q
+
j+1 — d Vcj*cosg (d Q
+
j +
d Q+cj*cos g )] + (1- q )/ D x[ d Vj+1 d Qj+1 - d Vcj*cos g ( d Q
+
j +
d Q+cj*cosg )] + ((q gAx(j+j+1)/2)/D x)(d z
+
j+1 - d z
+
j) + ((1- q )
gAx(j+j+1)/2)/D x) (d zj+1 - d zj) = ((q gAx(j+j+1)/2)/ 2)(So
+
(j+j+1)/2) +
((1-q )gAx(j+j+1)/2)/2) (S0(j+j+1)/2) - ((q gAx(j+j+1)/2)/2)[Sfj+1 +
(2Sfj+1/Qj+1) ( d Q
+
j+1 - d Qj+1) - (2Sfj+1/Kj+1)(¶ Kj+1/¶ h)(d z
+
j+1 -
d zj+1)] - ((q gAx(j+j+1)/2)/2)[Sfj + (2Sfj/(d Qj + d Qcj*cosg ))
(( d Q+j + d Q
+
cj*cos g )- ( d Qj + d Qcj*cos g )) - (2Sfj/Kj)
(¶ Kj/¶ z)(d z
+
j - d zj)] - ((1-q )gAx(j+j+1)/2)/2) Sfj+1 - ((1-
q )gAx(j+j+1)/2)/ 2) Sfj + (q /2)ql
+
(j+j+1/2) Vl
+
(j+j+1)/2 + ((1-q )/2)ql
(j+j+1)/2 Vl(j+j+1)/2 (13)
where q is a weighting coefficient, 0 £ q £ 1, which
controls the stability of the numerical results; g is flow
approaching angles at the channel junction (Figure 2);
d Q+j is discharge from time level n+1 at node j (m
3 s-1);
Vcj is velocity of flow at the channel junction (ms
-1); Qcj is
discharge at the secondary canal at time level n (m3 s-1);
Sf is friction slope; So is channel bed slope, and ql is
lateral flow, (m2 s-1); j is node index; and i is canal reach
index. After the substitution of Eqs. (8-11) into
momentum and continuity equations for both regular
nodes and channel junction node, a set of linear equations
was obtained as follows (Malaterre, 1994):
A11d Q
+
j + A12d z
+
j + A13d Q
+
j+1 + A14d z
+
j+1 = A«11d Qj +
A«12d zj + A«13d Qj+1 + A«14d zj+1 + C1 (14)
A21d Q
+
j + A22d z
+
j + A23d Q
+
j+1 + A24d z
+
j+1 = A«21d Qj +
A«22d zj + A«23d Qj+1 + A«24d zj+1 + C2 (15)
where d Q+j and d z
+
j are discharge and water-level
increments from time level n+1 at node j; d Qj and d zj are
discharge and water-level increments from time level n at
node j; and A11, A«21,É. A12, A22 are the coefficients of the
continuity and momentum equations, respectively,
computed with known values at time level n. For details
about coefficients of linear equations for channel
junctions, refer to Choi (1991). Similar equations are
derived for channel segments that contain a lateral, a gate
structure, a weir or some other type of hydraulic
structure. From the equations above, the state of system
equation at any sampling interval k can be written, in
compact form as follows:
AL d x(k+1) = ARd x(k) + Bd u(k) + Cd q(k) (16)
. F. DURDU
395
where A = l x l system feedback matrix, B = l x m control
distribution matrix, k = time increment (s); and D q =
variation in demands (or disturbances) at the turnouts
(m2 s-1). The elements of the matrices A, B, and C depend
upon the initial condition. Eq. (16) can be written in a
state-variable form along with the output equations as
follows (Reddy, 1999):
d x(k+1) = Fd x(k) + Gd u(k) + Yd q(k) (17)
d y(k) = Hd x(k) (18)
where F = (AL)
-1 *AR, G = (AL)
-1*B, and Y = (AL)
-1*C,
d x(k) = l x 1 state vector, d u(k) = m x 1 control vector,
d q(k) = p x l matrix representing external disturbances
(changes in water withdrawal rates) acting on the system,
d y(k) = r x 1 vector of output (measured variables), H =
r x l output matrix, l = number of dependent (state)
variables in the system, m = number of controls (gates)
in the canal, p = number of outlets in the canal, and r =
number of outputs. The elements of the matrices F , G ,
and Y depend upon the canal parameters, the sampling
interval, and the assumed average operating condition of
the canal. In Eq. (17), the vector of state variables is
defined as follows (Reddy, 1996):
d x = (d Qi,1, d zi,2, d Qi,2,ÉÉ d zi,N-1, d Qi,n-1, d Qi,N) (19)
Feedback Control
Feedback control is another term to describe closed-
loop control (Figure 3). A closed-loop control system
utilizes an additional measure of the actual output to
compare the actual output with a desired output
response. The measure of the output is called the
feedback signal. In irrigation canals, feedback type
control systems are used to minimize the magnitude and
durations of the mismatch between the supply and the
demand. Eq. (17) is called the discrete state equation in
this study. This equation describes the condition or
evolution of the basic internal variables. In hydraulic
engineering problems, the depth of flow, flow rate, and
velocity as a function of distance can be considered the
state or internal variables. Sometimes, the volume of
water in a given reach of a canal can also be considered a
state variable (Reddy, 1990). In this paper, the water
surface elevation and flow rate were considered the state
variables. Given the initial conditions [d x(0)], d u, and d q,
Eq. (17) can be solved for variations in flow depth and
flow rate as a function of time. If the system is really at
equilibrium [i.e. d x (0) = 0 at time t = 0] and there is no
change in the lateral withdrawal rates (disturbances), the
system would continue to be at equilibrium forever; then
there is no need for any control action. Conversely, in the
presence of disturbances (known or random), the system
would deviate from the equilibrium condition. The actual
condition of the system may be either above or below the
equilibrium condition, depending upon the sign and
magnitude of the disturbances. If the system deviates
significantly from the equilibrium condition, the discharge
rates into the laterals will be different (either more or
less) than the desired values. However, in canal
operations, the main objective is to keep these deviations
to a minimum so that a nearly constant rate of discharge
is maintained through the turnouts. The objective of
control theory is to find a control law that will bring an
initially disturbed system to the desired target value in the
presence of external disturbances acting on the system
(Reddy, 1996). During the last 2 decades, the concepts of
the optimal control theory have been applied for deriving
feedback control algorithms for the real-time control of
irrigation canals. The application of optimal control
theory for the derivation of canal control algorithms
eliminates the need for the trial and error procedure that
has been traditionally used. In order to maintain the
target water levels in the pools at specified values, in the
presence of random disturbances acting on the system,
Simulation of a Feedback Control Technique Through Irrigation Canal Junctions
396
q(z) disturbances
LQR
Controller
Canal
G(z)H(z)
U(z)
Input
Output
Figure 3. A feedback control system scheme.
the control structures (gates) in the irrigation canal are
frequently adjusted. This can be accomplished by applying
a large proportional control, in which the change in gate
opening is proportional to the changes in flow depths and
flow rates, of the following form (Reddy 1996):
d u(k) = -K d x(k) (20)
where K the m x l controller gain matrix. The selection of
values for the elements of the gain matrix is the design
problem that is posed here. In the case of a multi-input
and multi-output (MIMO) system, since the gains are not
uniquely determined, a systematic procedure for the
selection of the elements of matrix K is required. The LQR
control problem is an optimization problem in which the
cost function, J, to be minimized is given as follows:
(21)
subject to the constraint that
-d x(k+1) + Fd x(k) + Gd u(k) = 0 k = 0, ..., K
¥
(22)
where K
¥
= number of sampling intervals considered to
derive the steady state controller; Qxlxl = state cost
weighting matrix; and Rmxm = control cost weighting
matrix. The matrices Qx and R are symmetric, and to
satisfy the non-negative definite condition, they are
usually selected to be diagonal with all diagonal elements
positive or zero. The first term in Eq. (21) represents the
penalty on the deviation of the state variables from the
average operating (or target) condition, where the
second term represents the cost of control. This term is
included in an attempt to limit the magnitude of the
control signal d u(k). Unless a cost is imposed for the use
of control, the design that emerges is liable to generate
control signals that cannot be achieved by the actuator
(Reddy, 1996). In this case, the saturation of the control
signal will occur, resulting in system behavior that is
different from the closed loop system behavior that was
predicted assuming that saturation will not occur
(Tewari, 2002). Therefore, the control signal weighting
matrix elements are selected to be large enough to avoid
saturation of the control signal under normal operating
conditions. Eqs (21) and (22) constitute a constrained-
minimization problem that can be solved using the
method of Lagrange multipliers. This produces a set of
coupled difference equations that must be solved
recursively backwards in time. For the steady-state case,
the solution for d u(k) is the same form as Eq. (20),
except that K is given by
K = [R + G T SG ]-1 G T SF (23)
S is a solution of the discrete algebraic Riccati equation
(DARE):
F
T SF - F T SG [R + G TSG ]-1 G T SF + Qx = S (24)
where R = RT > 0 and Qx = QxT ‡ 0. The control law
defined by Eq. (20) brings an initially disturbed system to
an equilibrium condition in the absence of any external
disturbances acting on the system. In hydraulic
engineering problems, the depth of flow, flow rate, and
velocity as a function of distance can be considered the
state or internal variables. Sometimes, the volume of
water in a given reach of a canal can also be considered a
state variable.
Results and Discussion
To demonstrate and compare the feasibility of a linear
quadratic regulator (LQR) controller, an optimal
regulation problem for a discrete-time single pool
irrigation canal with a converging channel junction was
simulated (Figure 2). An example problem obtained from
Reddy (1990) was used in the study. The data used were
as follows: length of canal reach = 5000 m, number of
nodes = 5, number of subreaches used = 4, D x = 1250
m, channel slope = 0.0003, side slope = 1.0, bottom
width = 1.7 m, turnout demand = 2.5 m3 s-1, discharge
required at the end of the canal = 0.52 m3 s-1, upstream
reservoir elevation = 103.2 m, downstream reservoir
elevation = 101.14 m, target depth at downstream end
= 1.2 m, gate width = 1.7 m, and gate discharge
coefficient = 0.75. First, these data were used to
calculate the steady-state values, which in turn were used
to compute the initial gate openings and the elements of
the F , G , H matrices using a sampling interval of 30 s.
The values of the initial gate opening were u1 = 0.8 m and
u2 = 0.4 m. The analysis was started by evaluating the
system stability. All the eigenvalues of the feedback
matrix were positive and had values less than one. The
system was also found to be both controllable and
observable. In the derivation of the control matrix, G ,
elements, it was assumed that both the upstream and
downstream gates of each reach could be manipulated to
control the system dynamics. The downstream-end gate
position was frozen at the original steady-state value, and
only the upstream-end gate of the given reach was
controlled to maintain the system at the equilibrium
condition. The effect of variations in the opening of the
J = ∑
i = 1
K
¥
[d xT(k)Qxlxl d x(k) + d uT(k)Rmxmd u(k)]
. F. DURDU
397
downstream gate must be taken into account through
real-time feedback of the actual depths immediately
upstream and downstream of the downstream gate (node
N). In the derivation of the feedback gain matrix, R was
set equal to 100,000, whereas Qx was set equal to an
identity matrix of dimension 8 (the dimensions of the
system). In the absence of a well-defined procedure for
selecting the elements of these matrices, these values
were selected based upon trial and error.
First a linear quadratic controller problem was
simulated for a single reach canal without a channel
junction. The system response was simulated for 7500 s.
As demonstrated in Figure 4, the deviations in flow depth
at nodes 1, 2 and 5 were high, whereas at nodes 4 and
3, the deviations were close to the equilibrium condition.
The highest flow deviation occured at node 1. The
deviation in flow depth reached almost 16 cm at this
particular node. Since the turnout was located at the
downstream end of the reach, the maximum decrease in
depth of flow occurred at the last node. Due to the
increased opening of the upstream gate to compensate
for disturbances at the turnouts, the maximum increase
in depth of flow occurred at the upstream gate (Figure
5). At the beginning the upstream gate opening was 0.05
m. However, at the end of the simulation, the position of
the upstream gate is below the equilibrium value. This is
evidence that the system would eventually return to the
equilibrium condition. To bring the system to equilibrium,
the variation in gate opening reached its highest value of
approximately 0.05 m, and gradually decreased to a small
negative value of -0.016 m at 7500 s.
After simulating the system for an independent single
reach canal, the system was simulated for a single reach
canal with a channel junction at node 3. The flow in the
side channel converges into the main canal with a 12
degree approaching angle (g ) (Figure 2). The lateral
inflow angle (g ) to the downstream canal was shown by
Best and Reid (1984) to be dependent upon discharge
ratios between lateral inflow and downstream outflow,
and the channel junction angle (a ). The side channel
discharge was 3 m3 s-1. Because of flow mixing and flow
approaching angles, there was a change in backwater
surface profiles and in water depth in the main canal. As
shown in Figure 6, the flow depth at node 3 was
increasing at the beginning of the simulation but it was
coming close to the equilibrium position at 7500 s. Along
the simulation, the variations in flow depth at node 3
have positive values if compared with the first simulation.
Again the highest deviations in flow depth were at nodes
1, 2 and 5. The reason for these deviations was the
increased opening of the upstream gate and the decrease
in depth of flow at the last node. In addition, because of
the channel junction at node 3, there were backwater
surface effects at all nodes and water surface levels
increased along the main canal. The upstream gate
opening increased to 0.09 m at 1500 s and later
decreased to below the equilibrium condition at the end
of the simulation (Figure 7). In other words, the flow
Simulation of a Feedback Control Technique Through Irrigation Canal Junctions
398
0.2
0.15
0.1
0.05
0
-0.05
-0.1
-0.15
va
ri
at
io
ns
in
f
lo
w
d
ep
th
, (
m
)
node 1
node 2
node 4
node 3
node 5
0 50 100 150 200 250
duration of simulation, (s) (dt = 30 s)
Figure 4. Variations in flow depth for the main canal without channel
junction.
0.06
0.05
0.04
0.03
0.02
0.01
0
-0.01
-0.02
va
ri
at
io
ns
in
u
ps
tr
ea
m
g
at
e
op
en
in
g,
(
m
)
0 50 100 150 200 250
duration of simulation, (s) (dt = 30 s)
Figure 5. Variations in upstream gate opening for the main canal
without channel junction.
from the side channel into the main canal starts to show
its response to backwater surface profile at 1500 s.
When water surface profile increased in the canal reach,
the variation in the upstream gate was decreasing steeply.
At 2700 s, variations in gate opening approached a
constant value, indicating that a new equilibrium
condition was established. At the end of the simulation
the deviation in gate opening was decreasing to a small
negative value of -0.004 m at 7500 s. In the absence of
changes in the withdrawal rates, the gate will return to
its equilibrium position at steady state.
Conclusions
The feedback (closed-loop) control theory was applied
to derive a global control algorithm for an irrigation canal
with a converged channel junction. The control problem
was formulated as an optimal control problem directly
using the discrete linear quadratic regulator (LQR)
theory. The hydraulic formulation of the flow at a channel
junction was complicated because of energy losses, lateral
flow approaching angles and flow mixing. Using lateral
inflow angles to the downstream channel instead of
channel junction angles, based on the momentum flux
contribution a momentum equation at channel junctions
was derived. The performance of the model was
evaluated in terms of deviations in the depths of flow and
the upstream gate opening. The variation in the upstream
gate opening and the deviations in the depth of flow in
the canal reach were not erratic during the simulation
period. The system either approached a constant value or
came close to the equilibrium condition. This indicates
that the optimal control theory (feedback loop) is still
applicable on irrigation canals with a combining channel
junction. Although a single canal reach was considered in
this study to demonstrate in detail a procedure to derive
a control algorithm in the presence of a converging
channel junction, the same procedure can be used to
derive a feedback control algorithm to run irrigation
canals with diverging channel junctions or channel
junctions with multiple reaches.
. F. DURDU
399
Figure 7. Variations in upstream gate opening for the main canal with
channel junction.
0.2
0.15
0.1
0.05
0
-0.05
-0.1
-0.15
va
ri
at
io
ns
in
f
lo
w
d
ep
th
, (
m
)
0 50 100 150 200 250
duration of simulation, (s) (dt = 30 s)
node 1
node 2
node 3
node 4
node 5
0.1
0.08
0.06
0.04
0.02
0
-0.02
va
ri
at
io
ns
in
u
ps
tr
ea
m
g
at
e
op
en
in
g,
(
m
)
0 50 100 150 200 250
duration of simulation, (s) (dt = 30 s)
Figure 6. Variations in depth of flow for the main canal with channel
junction.
References
Balogun, O.S. 1985. Design of real-time feedback control for canal
systems using linear quadratic regulator theory, Ph.D. thesis,
Department of Mechanical Engineering, University of California at
Davis, 230 p.
Balogun, O.S., M. Hubbard and J.J. DeVries. 1988. Automatic control
of canal flow using linear quadratic regulator theory, J. of
Irrigation and Drainage Eng., 114: 75-101.
Simulation of a Feedback Control Technique Through Irrigation Canal Junctions
400
Begovitch, O. and R. Ortega. 1989. Adaptive head control of a hydraulic
open-channel model, Automatica, 25: 103-107.
Best, J.L. and I. Reid. 1984. Separation zone at open-channel junctions.
Journal of Hydraulic Engineering, 110: 1588-1594.
Bureau of Reclamation Technical Publication. 1995. Canal system
automation manual, United States Department of Interior, Bureau
of Reclamation Water Resources Services.
Choi, G.W. 1991. Hydrodynamic network simulation through channel
junctions, PhD Dissertation, Colorado State University, Fort
Collins CO.
Durdu, O.F. 2003. Robust control of irrigation canals, PhD Dissertation,
Colorado State University, Fort Collins CO.
Durdu, O.F. 2004a. Regulation of irrigation canals using a two-stage
linear quadratic reliable control, Turkish J. Eng. Environ. Sci, 28:
111-120.
Durdu, O.F. 2004b. Estimation of state variables for controlled
irrigation canals via a singular value based Kalman filter, Fresenius
Environmental Bulletin, 13:1139-1150
Garcia, A., M. Hubbard M. and J.J. DeVries. 1992. Open channel
transient flow control by discrete time LQR methods, Automatica,
28: 255-264.
Goussard, J. 1993. Automation of canal irrigation systems,
International Commission on Irrigation and Drainage, New Delhi.
Malaterre, P.O. 1994. Modelisation, Analysis and LQR Optimal Control
of an Irrigation
Canal, Ph.D. dissertation, LAAS-CNRS-ENGREF-Cemagref, France.
Malaterre, P.O. 1998. PILOTE: Linear quadratic optimal controller for
irrigation canals, J. of Irrig. and Drain. Engrg., 124: 187-194.
Reddy, M.J. 1990. Local optimal control of irrigation canals, Journal of
Irrigation and Drainage Engineering, 116: 616-631.
Reddy, J.M., A. Dia A and A. Oussou. 1991. Design of control algorithm
for operation of irrigation canals, J. of Irrig. and Drain. Engrg.,
118: 852-867.
Reddy, J.M. 1996. Design of global control algorithm for irrigation
canals, Journal of Hydraulic Engineering, 122: 503-511.
Reddy, J.M. 1999. Simulation of Feedback Controlled Irrigation Canals,
Proceedings USCID Workshop, Modernization of Irrigation Water
Delivery Systems: 605-617.
Rodellar, J., M. Gomez and L. Boner. 1993. Control method for an on-
demand operation of open-channel flow, J. of Irrig. and Drain.
Engrg., ASCE, 119: 225-241.
Tewari, A. 2002. Modern Control Design with Matlab and Simulink,
Wiley, New York.
Các file đính kèm theo tài liệu này:
- simulation_of_a_feedback_control_technique_through_irrigatio.pdf