Proceedings of the

The International 20th World

Federation of Congress
Automatic Control
The International
Proceedings
Toulouse, Federation
of the
France, 20th9-14,
July Worldof Congress
Automatic Control
2017
Toulouse,
The France,Federation
International July 9-14, 2017 Available online at www.sciencedirect.com
of Automatic Control
Toulouse, France, July 9-14, 2017
ScienceDirect
IFAC PapersOnLine 50-1 (2017) 6564–6569
Distributed Model Predictive Control of
Distributed Model Predictive Control of
Irrigation Systems
Distributed Model Predictive using Cooperative
Control of
Irrigation Systems using Cooperative
Irrigation Systems Controllers using Cooperative
Controllers
Controllers 
Le-Duy-Lai NGUYEN ∗,∗∗ Ionela PRODAN ∗ ∗,∗∗
LaurentLe-Duy-Lai
LEFEVRE NGUYEN ∗,∗∗
Denis Ionela PRODAN ∗ ∗
GENON-CATALOT
∗,∗∗
LaurentLe-Duy-Lai
LEFEVRE NGUYEN ∗,∗∗ Ionela PRODAN ∗ ∗
Denis GENON-CATALOT
∗,∗∗ ∗
∗ Laurent LEFEVRE Denis GENON-CATALOT
∗
Univ. Grenoble Alpes, Grenoble INP 1 , LCIS, F-26000 Valence,
Univ.FranceGrenoble 1
Alpes, Grenoble INP lcis.grenoble-inp.fr).
, LCIS, F-26000 Valence,
∗
({firstname.lastname}@ 1
Univ. Grenoble
∗∗France
CARE, Alpes, Grenoble
({firstname.lastname}@
Asian Centre INP lcis.grenoble-inp.fr).
for Water ,Research,
LCIS, F-26000 Vietnam. Valence,
∗∗France
CARE,({firstname.lastname}@
Asian Centre for Waterlcis.grenoble-inp.fr).
Research, Vietnam.
∗∗
CARE, Asian Centre for Water Research, Vietnam.
Abstract: This paper investigates the regulatory control of large-scale irrigation systems under
Abstract:
a Distributed This paper investigates
Cooperative the regulatory
Model Predictive control
Control of large-scale
(DCMPC) irrigation
framework. Wesystems under
first address
Abstract:
atheDistributed
challenging This paper
Cooperative
issue of investigates
providing the
Model Predictive regulatory
a mathematical control
Control of large-scale
(DCMPC)
model irrigation
for anframework.
irrigation We systems
systemfirst withunder
address an
a
the Distributed
challenging Cooperative
issue of Model
providing Predictive
a Control
mathematical (DCMPC)
model for
inherent complex dynamics using the Lattice Boltzmann (LB) method. Next, we discuss different anframework.
irrigation We
systemfirst address
with an
the
inherent
controlchallenging
complexissue
implementation dynamics ofstrategies
providing
using the aLattice
(i.e., mathematical
Boltzmann
decentralized model
(LB)
and for an irrigation
method.
distributed Next, we system
strategies) discuss with the
and different
how an
inherent
control
cooperation complex
implementation
among dynamics usingcontrollers
localstrategies
MPC the Lattice
(i.e., Boltzmann
decentralized
can (LB)
and
be included tomethod.
improveNext,
distributed we discuss
strategies)
the performance and different
how of the
the
control
cooperation implementation
overall system. amongFor local strategies a (i.e.,
MPC controllers
simulations, decentralized
particular canbenchmark and of
be included distributed
toanimprove strategies)
irrigationthe canal isand
performance how of the
considered.
cooperation
overall system. amongFor local MPC
simulations, controllers
a particular can be included
benchmark of
Comparisons through simulations among the proposed control approaches validate the benefitstoanimprove
irrigationthe performance
canal is of the
considered.
overall
Comparisons system. For
through simulations,
simulations
of the cooperative distributed control approach. a particular
among the benchmark
proposed of
control an irrigation
approaches canal
validate is considered.
the benefits
Comparisons
of the cooperative through simulations
distributed among
control the proposed control approaches validate the benefits
approach.
© 2017,
of IFAC (International
the cooperative distributed Federation
control of Automatic
approach.Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords: Irrigation systems, Lattice Boltzmann (LB) method, Model Predictive Control
Keywords:
(MPC), cooperativeIrrigation control,
systems,distributed
Lattice Boltzmann (LB) method,
and decentralized control.Model Predictive Control
Keywords:
(MPC), cooperative Irrigation control,
systems,distributed
Lattice Boltzmann (LB) method,
and decentralized control.Model Predictive Control
(MPC), cooperative control, distributed and decentralized control.
1. INTRODUCTION sensors to the centralized coordinator due to communica-
1. INTRODUCTION sensors
tion issues.to theThe centralized
decentralized coordinator
approach duewillto communica-
scale better
1. INTRODUCTION
Irrigation has always been a central concern for human for the large system. The objective of each to
sensors
tion to
issues. the
The centralized
decentralized coordinator
approach due will communica-
scale better
controller is to
Irrigation has always been a central
societies. Currently, the performance of irrigation canals concern for human tion
for
locallyissues.
the large The
system.
determine decentralized
the The objective
actions approach
that of each
optimizewill scale
controller
the better
is to
behavior
Irrigation
societies. has always
manually Currently,
operated the
can been a central
noperformance
longer meetconcern
of new for
theirrigation canals for
human
challenges the subsystem
locally
of the large system.
determine the The
(i.e., byobjective
actions that of
minimizing eachthe controller
optimize specified is
the behavior to
cost
societies.
manually Currently,
operated the
can noperformance
longer meet of
theirrigation
new canals
challenges locally
of the determine
subsystem
of irrigated areas in both water conservation and oper- function). The drawback of decentralized control is that the actions
(i.e., by that
minimizing optimize
the the behavior
specified cost
manually
of irrigated
ating operated
aspects areas canboth
in
(as discussed no longer
water meet
in Malaterre the new
conservation oper- of
et al.challenges
and(2007)). thelocal
function).
these subsystem
The
actions (i.e.,
drawback byofminimizing
are computed decentralized thecontrol
regardless specified cost
of the iseffects
that
of
Theirrigated
ating aspectsareas
automation inengineering
both water
(asofdiscussed conservation
in Malaterre
works al.and
is aetpossible way function).
oper-
(2007)). these
on thelocal The
overall drawback
actionssystem of decentralized
are computed
performance. The control
regardless of the iseffects
controllers that
do
ating
The aspects the
automation
of improving (asofmanagement
discussed
engineering in Malaterre
works
of is aetpossible
irrigation al. (2007)).
canals. Au- these
way on
not the local
consider actions
overallthesystem are computed
performance.
cooperation, regardless
negotiation of the effects
Theorcontrollers
coordination do
The
of automation
improving the of engineering
management works
of is
irrigation a possible
canals. way
Au- on
not the overall
consider
tomatic control systems can be designed to achieve the among them. As a consequence, the global performance the system performance.
cooperation, negotiation The orcontrollers
coordination do
of improving
tomatic control the management
systems can be of irrigation
designed
desired objectives of water resource management (e.g., to canals.
achieve Au-
the not
among
may consider
be them. the
As
degraded cooperation,
a consequence,
due to the negotiation
the
unexpected globalor coordination
performance
and unantici-
tomatic
desired control
equitableobjectives
distribution systems
of towatercan resource
suit be demands
all designed to achieve
management
and the among
(e.g.,
save water) may
patedbe them.
interactionsAs aamong
degraded consequence,
due tothe the global
thesubsystems.
unexpected The performance
and unantici-
distributed
desired
equitable objectives
while reducingdistribution of to
investment water
suit
andresource
all operating
demands management
and
costs.save (e.g., may
water)
Different pated
approachbe degraded
interactions
combinesamong due tothethesubsystems.
the optimality unexpected The
of centralizedanddistributed
unantici-
approach
equitable
while distribution
reducing investment to suit
and all demands
operating and
costs.save water)
Different pated
approach interactions
combines
approaches to the control of irrigation systems have been and the less complexity of decentralized approach to pro- among
the the
optimality subsystems.
of The
centralized distributed
approach
while reducing
approaches
developed to the
and investment
control
applied to ofand
real operating
irrigation
canals costs.the
systems
all over been approach
Different
have world. and
vide the
a good combines
less complexity
compromise. the optimality
of decentralized of centralized
approach approach
to pro-
approaches
developed
There are and to the
several control
applied
survey to of
realirrigation
works canals
whichall systems
over the
highlight have im- and
been
world.
the vide the
a goodless compromise.
complexity of decentralized approach to pro-
developed
There areofand
several applied
survey toreliable
real canals
works which allhighlight
over thethe im- vide
world. Motivated
a goodbycompromise.
the above mentioned issues, the overall goal
portance establishing irrigation systems. For Motivated by the
There
portance are of several survey
establishing works
reliable which
irrigationhighlight
systems. the im-
For of this paper is toabove mentioned
implement and issues,
compare therealistic
overall goal
and
example, Malaterre et al. (2007) review, structure and Motivated of this paper by the toabove
isschemes mentioned
implement and issues,
compare therealistic
overall goal
and
portance
example, of establishing
Malaterre et al.reliable
(2007) irrigation
review, systems.ofand For flexible control where:
analyze the main concepts involved in the structure
control ir- of this paper
flexible controlisschemes
to implement
where: and compare realistic and
example,
analyze
rigation the Malaterre
canals main using eta al.
concepts (2007)
involvedreview,
benchmark, in the
the structure
control
authors ofand
provideir- flexible
• costs, constraints,
control schemes profiles
where: are taken into account in
analyze
rigation the
canals
the comparison main concepts
using
results involved
a obtained
benchmark, inmanual
withthe the control
authors of ir-
provide
operations • costs, constraints,
decentralized profiles are constrained
and distributed taken into account optimiza- in
rigation
the canals
and comparison
classical using
results
control a obtained
designbenchmark,
(usingwith PID,the authors
manual
LQG, provide
loperations
1 ).
• costs, constraints,
decentralized
tion problems and profiles
(i.e.,distributed
via a MPC are constrained
taken
design); into account
optimiza- in
the comparison
and classical control results obtained
design (usingwith PID,manual
LQG, loperations
1 ). decentralized
tion problems and
(i.e.,distributed
via a MPC
• the cooperation among local MPC controllers can beconstrained
design); optimiza-
One of the project
and classical control demonstrators
design (using PID, (focusing
LQG, on l1 ). the au- • tion problems (i.e.,
One of the project demonstrators (focusing on the au-
the cooperation
included in order tovia
among alocal
improve MPC MPCdesign);
the controllers can
performance of thebe
tomation of “Canal de la Bourne” (CB) irrigation network • included
the cooperation
in orderamong to improvelocal MPC controllers can
the performance of thebe
One of the
tomation project de
of- “Canal demonstrators
la Bourne” (focusing oncontrol
the au- overall system.
in Drôme France) introduces the(CB) irrigation
distributed network of included in order to improve the performance of the
overall system.
tomation
in Drôme of- “Canal
a large-scale France)
irrigationde lasystem
Bourne”
introduces the(CB)
as caseirrigation
a distributed
study controlnetwork
(for more of overall
2. MODELsystem.PREDICTIVE CONTROL OF
in
a Drôme
details, see- Nguyen
large-scale France)
irrigationetintroduces
al.system
(2014)). the
as a distributed
case study
Controlling suchcontrol
(forsystem
more of 2. MODEL PREDICTIVE CONTROL OF
IRRIGATION SYSTEMS
a
in large-scale
details, see Nguyen
a centralized irrigation
way al.system
et would (2014)).
giveas Controlling
a case
rise to study such
a large (for more
system
compu- 2. MODEL PREDICTIVE
IRRIGATION CONTROL OF
SYSTEMS
details,
in
tational see
a centralized Nguyen
charge way
and etset
al.up
would(2014)).
agive
singleControlling
risepoint
to aoflargesuch
failure. system
compu-
The There are many IRRIGATION
works whichSYSTEMS employ different control
in a centralized
tational
important charge
delaysand way
mayset would
up aingive
occur singlerisepoint
information to agathering
oflarge
failure.compu-
The There
from strategiesarewithin
manyanworks MPC whichframework employ different
for the controlcontrol
of irri-
tational
important charge and
 The work delays mayset up ain single
occur point gathering
information of failure. from The Therestrategies
gation arewithin
canals.manyTo anworks
MPC
mention which
framework
just a employ
few for
recent different
the control
ones, control
of irri-
Ocampo-
important is partially supported and funded by the Artemis Ar-
 The work delaysis
maysupported
partially
occur in and information
funded by
gathering
the Artemis
from
Ar- strategies
gation
Martinez within
canals.
and To an
Negenborn MPC
mention framework
just
(2015) a few
present for
recent
athe control
ones, of irri-
Ocampo-
methodology for
rowHead European project under grant agreement number 332987.

1
rowHead
The European
work
Institute project
ofisEngineering
partially underGrenoble
supported
Univ. grant
and agreement
Alpes.by number
funded the Artemis 332987.
Ar- gation
Martinez canals.
the optimal To mention(2015)
andmanagement
Negenborn just a present
few recent
of combined ones,
awater Ocampo-
methodology
supply for
and
1 Institute of Engineering Univ. Grenoble Alpes.
rowHead European project under grant agreement number 332987. Martinez
the optimal andmanagement
Negenborn (2015) present awater
of combined methodology
supply and for
1 Institute of Engineering Univ. Grenoble Alpes. the optimal management of combined water supply and
Copyright © 2017 IFAC 6758
2405-8963 ©
Copyright © 2017, IFAC (International Federation of Automatic Control)
2017 IFAC 6758Hosting by Elsevier Ltd. All rights reserved.
Peer review©under
Copyright 2017 responsibility
IFAC of International Federation of Automatic
6758Control.
10.1016/j.ifacol.2017.08.612
 Proceedings of the 20th IFAC World Congress
Toulouse, France, July 9-14, 2017 Le-Duy-Lai Nguyen et al. / IFAC PapersOnLine 50-1 (2017) 6564–6569 6565
navigability/sustainability in river systems based on MPC.
Sutrisno et al. (2012) propose a scheme to determine the
weighting values that are proportional to the control load
for each subsystem. This weighting scheme is then applied
to control the irrigation canal using Feasible-Cooperation
MPC and Nash-bargaining MPC. Igreja et al. (2011) pro-
pose a distributed MPC algorithm for a water delivery
canal using linearized and discretized Saint-Venant model.
Negenborn et al. (2009) focus on distributed MPC strategy
in which a discrete-time linear integral delay model is used
for prediction. Their distributed control algorithm is based
on augmented Lagrangian duality method. A methodol-
ogy for the optimal management of a combined irrigation
and water supply system based on MPC is proposed in
Puig et al. (2012). Van Overloop et al. (2015) present a
decentralized MPC scheme for large-scale systems whose
components can exchange information through a network.
2.1 Prediction model
The water height and flow dynamics in such canal are usu-
ally modeled by the one-dimension Shallow Water equa-
tions, also known as Saint-Venant equations Malaterre
et al. (2007). The Shallow Water equations for fluid dy-
namics in one dimension x are formulated as follows:
∂t (h) + ∂x (hu) = 0
(1)
1
∂t (hu) + ∂x ( gh2 + hu2 ) = F
2 downstream h around the gate. That is:
where h is the water height, u is (depth-average) horizontal
velocity of the flow, and g is the acceleration of gravity.
The flow rate Q at a particular location in a rectangular
canal of width B is deduced by: Q = Bhu. The force term
F = gh(I − J) is calculated from the bed slope I (where
I = ∂x hb with hb is the bed profile) and the friction J at
the bottom of the reach. J is deduced by the Manning-
Strickler equations (see Pham et al. (2010)).
Fig. 1. The “Canal de la Bourne” (CB) irrigation canal
modeled by Lattice Boltzmann method.
By coupling with the equations (1), the LB method
is demonstrated in several works Pham et al. (2010)
that it is an efficient and powerful numerical tool (in
terms of accuracy, numerical stability and computational
6759
time) to simulate the hydrodynamic systems. Regarding
the CB system illustrated in Fig. 1, the one-dimension
three-velocities (D1Q3) LB method is used to model the
irrigation canal in one dimension l (lattice spacing ∆l)
and with three velocities [v1 v2 v3 ] = [0 v (−v)] where
v = ∆l/∆t (∆t is time step). Modeling one-dimensional
fluid flows with D1Q3 LB method in the presence of an
external force F , the following relations result:
1 ∆t
fi (l + v∆t, t + ∆t) = fi (l, t) + (fie − fi ) + ωi 2 vi F
τ cs
(2)
where i ∈ {1, 2, 3}, τ is constant relaxation time, ωi and cs
are the parameters that are determined by the geometry
of the lattice and chosen to obtain the isotropy for the
model. The fie are the equilibrium distribution functions
calculated by Pham et al. (2010). The macroscopic vari-
ables h and u are determined by:
3
h= f i = f1 + f 2 + f 3
i=1
3
(3)

q = hu = vi fi = v(f2 − f3 )
i=1
Furthermore, we consider the coupling of two reaches
modeled by D1Q3 LB method with a gate in a submerged
regime. The flow rate through gate Qg depends on the
difference between the water heights of upstream h and

Qg = Bg αθ 2g(h − h ) (4)
where Bg is the gate width, α is the gate coefficient and θ
is the gate opening. In addition, the outflow Q of upstream
reach is equal to the inflow Q of downstream reach, that
is: Q = Q = Qg . The manipulated variable θ is deduced
from following gate equations:
vB(f2 − f3 ) = vB  (f2 − f3 ) = Bg αθ 2g(h − h ) (5)
For further use, we consider the linearization of the LB
dynamics (2) around an equilibrium height, h = he , and
an equilibrium velocity, u = ue (we deduce the equilibrium
flow rate: Qe = Bhe ue ). A small derivations i , i =
{1, 2, 3} around the equilibrium points are expressed by:
i = fi − fie (he , ue ). We consider here the downstream
configuration of our case study (as shown in Fig. 2). In
this configuration, we assume that some measurements are
available such as the water height hus at upstream, the
water height hds at downstream and the flow rate Q at the
downstream of each reach. The control action is applied on
the gate to adjust gate opening θ (i.e., manipulated input).
The controlled variable is assumed as the downstream flow
rate Q of each reach. We also consider a lateral discharge
Qp at point lp of each reach as the perturbation. As a
result, the general linearized discrete-time model for reach
i of length L (corresponding to N discretized points lj ,
j = 1, . . . , N ) used in MPC scheme is presented as follows:
x i (k + 1) =A i x i (k) + B i u i (k)
+ Bp i p i (k) + Bd i d i (k)
y i (k) =Cy i x i (k) + Dy i u i (k) + Dyd i d i (k) (6)
z i (k) =Cz i x i (k) + Dz i u i (k) + Dzd i d i (k)
q i (k) =Cq i x i (k) + Dq i u i (k) + Dqd i d i (k)
Proceedings of the 20th IFAC World Congress
6566
Toulouse, France, July 9-14, 2017 Le-Duy-Lai Nguyen et al. / IFAC PapersOnLine 50-1 (2017) 6564–6569

where we define discrete-time local states of each reach as: 2.3 Cost function
x i (k) = [1i (l1 , k) 3i (l1 , k) . . . 2i (lj , k) 1i (lj , k) 3i (lj , k)
. . . 2i (lN , k) 1i (lN , k)]T ∈ R(N −2)×(N −2) , inputs as: The regulatory control of irrigation canals expects an auto-
u i (k) = 2i (l1 , k) ∈ R, perturbation as: p i (k) ∈ R, matic adjustment of controlled structures with reliability
and durability. For instance, a gate will move up and down
interaction variables between subsystems as: d i (k) = permanently to regulate flow rate at the end of a reach
3i (lN , k) ∈ R, controlled variable as: z i (k) = Q̃ i (lN , k) ∈ close to reference value (set-point). In other words, a local
R, (where Q̃ i (k) = Q i (k) − Qe ), measured output as: controller seeks to achieve the objectives as follows (see
y i (k) = [h̃ i (l1 , k) h̃ i (lN , k)], (where h̃ i (k) = h i (k) − Negenborn et al. (2009); Puig et al. (2012)):
he ), estimated flow rate through controlled gate as: • Minimize the deviation of controlled variable from
q i (k) = Q̃ i (l1 , k) ∈ R, and corresponding matrices as: the reference value (perform the regularization) (first
A i ∈ R(N −2)×(N −2) , B i ∈ R(N −2)×1 , Bpi ∈ R(N −2)×1 , term of (9))
Bdi ∈ R(N −2)×1 , Cyi ∈ R2×(N −2) , Dyi ∈ R2×1 , Dyd
i
∈ • Minimize the change in deviation of controlled vari-
i
able (e.g., in order to encourage smooth change of
R2×1 , Czi ∈ R1×(N −2) , Dzi ∈ R1×1 , Dzd ∈ R1×1 , Cqi ∈ flow rate) (second term of (9))
i
R1×(N −2) , Dqi ∈ R1×1 , Dqd ∈ R1×1 . • Minimize the change in sequence of computed actions
(avoid the oscillation) (third term of (9))
Therefore, we present the cost function for each controller:
2.2 Structural and operational constraints
(Np −1)

i
J (k) = (||z i (k + n|k) − r i (k)||2Z1
The operational management of irrigation canals is subject
n=0 (9)
to constraints on the canal structure (e.g., limits of gate
opening, reliability and robustness of equipment, limits of + ||∆z i (k + n|k)||2Z2
reservoir capacity) as well as on canal operations (e.g., + ||∆u i (k + n|k)||2U1 )
overflow risks, bank stability, minimum off-take levels).
The control solutions for the irrigation canal illustrated where z i is the controlled variable over the prediction
in Fig. 2 and modeled by the discrete-time systems in (6) horizon Np , r i (k) is the reference value, u is system input
must satisfy constraints on water height, gate opening, and over prediction horizon Np , and Z1 , Z2 , U1 are weighting
on variation of gate opening as follows: matrices of appropriate dimensions.
hmin ≤ h i ≤ hmax
2.4 Decentralized control algorithm
θmin ≤ θ i ≤ θmax (7)
∆θmin ≤ ∆θ i ≤ ∆θmax The details on the decentralized construction can be
synthesized in Algorithm 1.
where the water height is given by: h i (k) = h̃ i (k) + he
with h̃ i (k) = 1i (k) + 2i (k) + 3i (k). The gate opening is Algorithm 1 Decentralized MPC scheme
deduced from gate equations (5), that is: 1: Inputs: Initial state x i (1), initial input u i (1), pre-
i i
vB(2 (k) − 3 (k) + f2e − f3e ) dicted perturbation p i (k) and reference trajectory
θ i (k) =  (8)
i-1 i-1 r i (k)
α 2g(hds (k) − hds (k)) 2: for k = 1 : kmax do
i-1
3: measure the flow rate Q i at downstream of each
i
where hds (k), hds (k) are given by water level sensors or reach
computed from measured outputs y i-1 (k) and y i (k). 4: estimate the interaction variable d i (k) from gate
equations (5)
5: obtain Uki ∗ (x i (k)) by solving optimization prob-
lems (9) for prediction horizon N p under the con-
straints (7)
6: apply the first element uki ∗ of Uki ∗ to the subsystem
7: determine the state x i (k + 1) and outputs
z i (k), y i (k) at time instant k from (6)
8: go to the next step (k + 1)
9: end for

3. DISTRIBUTED MODEL PREDICTIVE CONTROL

USING COOPERATIVE CONTROLLERS

In this section, we consider the canal comprising a cascade

Fig. 2. Schematic of the flow rate control in downstream of n reaches of varying length Li , i = {1, 2, . . . , n} inter-
configuration. connected by submerged gates (as shown in Fig. 2). To

6760
 Proceedings of the 20th IFAC World Congress
Toulouse, France, July 9-14, 2017 Le-Duy-Lai Nguyen et al. / IFAC PapersOnLine 50-1 (2017) 6564–6569 6567
account for the interactions between subsystems, informa-
tion exchange among controllers is required (see more de-
tails in Nguyen et al. (2016a)). Moreover, the cooperation
involved in this strategy requires that all local objectives of
the controllers have to be coordinated towards optimizing
global objective (e.g., all controlled variables rapidly reach
their set-points).
3.1 Definition and management of shared information
We consider now the downstream configuration of our
case study as shown in Fig. 2. When we couple two
reaches i-1 and i with a gate controlled by controller
C i , the upstream point lus of the gate also is the
downstream point of the reach i-1 , and the downstream
point lds of the gate is the upstream point of the
reach i . The flow rate Qg through the gate is com-
puted from the gate equations (5). In order to deter-
mine interaction variable d i (k) and to compute the gate
opening θ i , we may identify from the linearized LB
model (6) and the gate equations (5) that the neces-
sary information exchanged among the controllers con-
i-1 i-1 i+1
tains: {1 (lLi−1 , k), 2 (lLi−1 , k), 3 (lus , k)}. Appropri-
ately, sharing information from controller C i to its neigh-
bors can be chosen as: [3i (lus , τ ) 1i (lLi , τ ) 2i (lLi , τ )]T .
In addition, the gate equations (5) permit the computation
of these interaction variables, once the neighbor controllers
share the flow rate Qg through the gate at their position.
Timing for sharing information depends on synchroniza-
tion mechanism and also the coordination method for each
group of neighboring controllers (see more detail in Nguyen
et al. (2016b)).
3.2 Design of cooperative controllers
For improving the global performance of our control ap-
plication, we minimize the time that the controlled vari-
ables of all controllers reach the percentage per i of their
set-points. Thus, for cooperation, the controllers have to
exchange their objectives (e.g., set-point r i ). In the down-
stream configuration (as shown in Fig. 2) and from the gate
equations (5), we observe that the flow rate through the
controlled gate, estimated by a controller may be involved
in neighbor objectives. As a consequence, each controller
has to optimize local objectives and also minimize devi-
ations of the estimated flow rate through gate Qg and
the upstream neighbor set-point r i-1 . The cost function of
cooperative controller C i can be reformulated as follows:
(Np −1)
i
Jcoop (k) = (||z i (k + n|k) − r i (k)||2Z1
n=0
+ ||∆z i (k + n|k)||2Z2
i
+ ||∆u (k + n|k)||2U1
+ ||q i (k + n|k) − r i-1 (k)||2Q1 )
where r i-1 (k) are the reference values of the reach i-1 ,
q i (k) is the estimated flow rate through the controlled
gate over the prediction horizon Np , and Q1 are weighting
matrices of appropriate dimensions. Therefore, the reced-
ing horizon control algorithm can be implemented in a dis-
tributed (and cooperative) control fashion as summarized
in Algorithm 2.
Algorithm 2 Distributed MPC scheme for N controllers
1: for i = 1, . . . , N do
2: Inputs: Initial state x i (1), initial input u i (1), pre-
dicted perturbation p i (k) and reference trajectory
r i (k)
3: for k = 1 : kmax do
i+1
4: receive the flow rate Qg through downstream
i+1
gate from downstream controller C and the
reference value r of upstream controller C i-1
i-1
5: determine the interaction variable d i (k) from
gate equations (5)
6: obtain Uki ∗ (x i (k)) by solving optimization
problems (9) for prediction horizon N p under
constraints (7)
7: apply the first element uki ∗ of Uki ∗ to the sub-
system
8: determine the state x i (k + 1) and outputs
z i (k), y i (k) at time instant k from (6)
9: send the estimated flow rate through gate
Q i (k) = q i (k)+Qe and reference value r i (k+1)
10: go to the next step (k + 1)
11: end for
12: end for
4. SIMULATION RESULTS
This section presents some simulation results using various
control schemes over the irrigation canal presented in Fig.
2 with n = 3 reaches. First, we employ a centralized control
scheme (using a single controller), then a decentralized
control scheme (using several separate local controllers as
presented in Section 2.4), the distributed control scheme
(using several controllers exchanging the flow rate through
controlled gate as presented in Section 3.1) and finally
the distributed cooperative control scheme (using several
cooperative controllers as detailed in Section 3.2). The
simulation results of different approaches are eventually
compared according to performance criteria such as re-
sponse time, overshoot limits, and computation delay. The
response time for global performance is defined as the
time needed for all reaches to attain per i = 95%, i =
{1, 2, 3} of their reference values. The overshoot limits
are defined as the maximum one of three reaches. For all
the control approaches, we have used the same param-
eters (as shown in Tables 1, 2) and constraints (Table
3). The simulations were done in Matlab R2016b on a
TM
(10) computer IntelCore
R i5-4310U CPU 2.0GHz. The
control objective for the reach i , (denoted by the subscript
ri, i = {1, 2, 3}) is to keep the downstream flow rate Qri
(with Qri (k) = zri (k) + Qe ) to the reference value Qrefri
(with Qref
ri (k) = rri (k) + Q e ) by acting on the opening
θri of the upstream gate. In addition, we also tested the
ability of MPC to reject perturbations, for example, when
there is a lateral withdrawing from the location Lp = 4m
6761
Proceedings of the 20th IFAC World Congress
6568
Toulouse, France, July 9-14, 2017 Le-Duy-Lai Nguyen et al. / IFAC PapersOnLine 50-1 (2017) 6564–6569

Table 1. Structural parameters of irrigation −3 DTCC: Outputs and reference values

x 10
canals 3

Length of the reach (Li ) 7m

Length until the pumping station (Lpi ) 4m
2.5
Width of the reach (Bi ) 0.1m
Bed slope (Ii ) 2.6e−3
Manning coefficient (ni ) 1/97

zri = Qri − Qe (m3/s)

Performance coefficient of the gate (αi ) 0.8 2

zr1
Table 2. Simulation parameters of irrigation 1.5
canals rr1
zr2
LB parameters
Spacing step (∆x) 1m 1 rr2
Time step (∆t) 0.05s zr3
Relaxation time (τi ) 0.8
rr3
Linearization around equilibrium points 0.5
Water height at equilibrium (he ) 0.16m Qp
Boundary conditions
Upstream height of first reach (Hus ) 0.20m 5 10 15 20 25 30 35 40 45
Time (s)
Downstream height of final reach (Hds ) 0.10m
Simulation parameters
Simulation time (T ) 45s Fig. 4. Distributed cooperative control with perturbation
Prediction horizon (Npred ) 0.5s Qp - Variation of downstream flow rates (zri = Qri −
Weighting matrices Z i = Qi = R i = I Qe ) with regards to equilibrium points Qe of three
Reference value (r)(m3 /s) rr1 = 2.6e − 3
reaches (ri = {r1, r2, r3}).
rr2 = 1.8e − 3
rr3 = 1.0e − 3

−3 DTC: Outputs and reference values

Table 3. Constraints of irrigation canals x 10
3
Gate opening 0m ≤ θ ≤ 0.5m
Variation of gate opening −0.02m ≤ ∆θ ≤ 0.02m
Constraints on water heights 0.10m ≤ h ≤ 0.25m 2.5
Constraints on state variables −0.5 ≤ x ≤ 0.5
−0.5 ≤ ∆x ≤ 0.5
zri = Qri − Qe (m3/s)

Constraints on the inputs −0.002 ≤ u ≤ 0.005 2

−0.004 ≤ ∆u ≤ 0.004
zr1
1.5
Table 4. Simulation results rr1

Perf. criterion Resp. time Overshoot Comput. time zr2

(s) (m3 /s) (s) 1 rr2
Centralized 14.30 0.2565e-04 61.974222 zr3
Decentralized 23.20 1.8156e-04 39.317066
Distributed 15.65 2.3128e-04 47.479176
rr3
0.5
Cooperative 11.45 1.0217e-04 53.156295 Qp

5 10 15 20 25 30 35 40 45
Time (s)

Fig. 5. Distributed control with perturbation Qp - Vari-

ation of downstream flow rates (zri = Qri − Qe )
with regards to equilibrium points Qe of three reaches
(ri = {r1, r2, r3}).

of each reach with the discharge p(k) = Qp = 0.8e−3m3 /s

Fig. 3. Performance comparison of four MPC settings:
centralized, decentralized, distributed and distributed
cooperative control.
beginning at time instant k = 2.5s.
As shown in Figures 4 to 7, the controlled variables zri ,
starting from 0 (Qri = Qe ), have reached the reference
values rri through the use of all four control settings.
The performance of each control approach given by the
simulation results have been summarized in Table 4 and
Fig. 3.

6762
 Proceedings of the 20th IFAC World Congress
Toulouse, France, July 9-14, 2017 Le-Duy-Lai Nguyen et al. / IFAC PapersOnLine 50-1 (2017) 6564–6569 6569

−3 DCC: Outputs and reference values that: (1) it respects the constraints on the controlled and
x 10 manipulated variables; (2) it avoids excessive variations
3
on the manipulated variables allowing better use of ac-
tuators; and (3) it rejects external perturbations. Future
work will concentrate on more comparisons with other lin-
2.5
ear/nonlinear optimization-based control approaches over
different benchmarks available in the literature. Also, dif-
zri = Qri − Qe (m3/s)

ferent cost criteria and economic MPC implementations

2 will be thoroughly investigated.
zr1
rr1 REFERENCES
1.5
zr2 Igreja, J.M., Cadete, F.M., and Lemos, J.M. (2011). Ap-
rr2 plication of distributed model predictive control to a
zr3
water delivery canal. 19th Mediterranean Conference
1 on Control and Automation (MED), 682–687.
rr3 Malaterre, P., Navarro, G., Playan, E., and Ed, F.B.
Qp (2007). Control of irrigation canals : why and how?
0.5 International Journal for Numerical Modeling of Hydro-
5 10 15 20 25 30 35 40 45
Time (s) dynamics for Water Resources, 271–292.
Negenborn, R.R., Overloop, P.J.V., Keviczky, T., and
De Schutter, B. (2009). Distributed model predictive
Fig. 6. Decentralized control with perturbation Qp - Vari- control of irrigation canals. Network and Heterogeneous
ation of downstream flow rates (zri = Qri − Qe ) Media, 4(2), 359–380.
with regards to equilibrium points Qe of three reaches Nguyen, L.D.L., Lefevre, L., and Genon-Catalot, D.
(ri = {r1, r2, r3}). (2016a). A composite metric for dynamic routing in
networked control systems. In IEEE 14th Int. Conf. on
−3 CC: Outputs and reference values
x 10 Industrial Informatics (INDIN). Poitiers, France.
3
Nguyen, L.D.L., Lefevre, L., Genon-Catalot, D., and Lami,
Y. (2016b). Asynchronous information consensus in
2.5 distributed control of irrigation canals. In 21st IEEE
Conference on Emerging Technologies and Factory Au-
tomation (ETFA), 12–15. Berlin, Germany.
zri = Qri − Qe (m3/s)

2 Nguyen, L.D.L., Lefevre, L., Genon-Catalot, D., Pham,

1.5
1 rr2
0.5
5 10 15 20 25
Time (s)
Fig. 7. Centralized control with perturbation Qp - Vari-
ation of downstream flow rates (zri = Qri − Qe )
with regards to equilibrium points Qe of three reaches
(ri = {r1, r2, r3}).
5. CONCLUSIONS
This paper investigated the control of irrigation systems
using a distributed cooperative MPC framework. Using the
LB method, we provided a discrete-time nonlinear model
of an irrigation system which captures its dynamics and
the interactions among the subsystems. Different control
implementation strategies were discussed and compared
(i.e., decentralized and distributed strategies). Several sim-
ulations over a particular irrigation canal example val-
idated the proposed approaches. The MPC framework
used within the four control strategies has demonstrated
V., and Raı̈evsky, C. (2014). Optimal reactive control
zr1 of hybrid architectures : A case study on complex water
transportation systems. In 19th IEEE International
rr1
Conference on Emerging Technologies and Factory Au-
zr2 tomation (ETFA), volume 1, 1–8. Barcelona, Spain.
Ocampo-Martinez, C. and Negenborn, R.R. (2015). Model
zr3 predictive control for combined water supply and navi-
gability / sustainability in river systems. In Transport of
rr3
Water versus Transport over Water, chapter 2, 13–32.
Qp Springer International Publishing, 1st edition.
30 35 40 45 Pham, V.T., Chopard, B., Lefèvre, L., Anda Ondo, D., and
Mendes, E. (2010). Study of the 1D lattice Boltzmann
shallow water equation and its coupling to build a canal
network. Journal of Computational Physics, 229(19),
7373–7400.
Puig, V., Ocampo-Martinez, C., Romera, J., Quevedo,
J., Negenborn, R., Rodriguez, P., and de Campos, S.
(2012). Model predictive control of combined irrigation
and water supply systems: Application to the Guadiana
river. Proceedings of 2012 9th IEEE International
Conference on Networking, Sensing and Control, 85–90.
Sutrisno, Salmah, and Wijayanti, I.E. (2012). Distributed
model predictive control and application to irrigation
canal. Control, Systems & Industrial Informatics (ICC-
SII), 1(978), 126–130.
Van Overloop, P.J., Maestre, J.M., Sadowska, A.D., Ca-
macho, E.F., and De Schutter, B. (2015). Human-in-the-
Loop Model Predictive Control of an Irrigation Canal.
IEEE Control Systems, 35(4), 19–29.

6763