Get documents about "An Intelligent Decision Support System for Irrigation System
Document Sample
An Intelligent Decision Support System for Irrigation System Management
R. M. Faye*+, F. Mora-Camino++ S. Sawadogo*, A. Niang*
+L.A.A.S du CNRS 7, Avenue du Colonel *Universitt? Cheikh Anta Diop
Roche 31077 Toulouse-France Ecole SupCrieure Polytechnique B.P 10
+E.N.A.C 7, Avenue Edouard Belin 31055 Thiks-S6nCgal
Toulouse-France
developments provided human societies with new means
ABSTRACT of better controlling water resources, so a lot of effort is
made in this direction.
In this communication is considered the design of a In canal control significant progresses are obtained and
decision support system for the short term water resource General Predictive Control has been considered to
management of an irrigation system. The operations of achieve successfully this task [7] [8]. However, for short
similar systems are often impaired by different stochastic term water resource management, since canal operation
events like device failure, heavy rains or dry periods and improvement requires good information on the system
new long term goals. To be effective, such a decision status and good knowledge of the system behavior,
support system which is based on knowledge techniques empirical or hierarchical solutions have been developed
(state identification) and adaptive optimization (short PI.
term plans), requires the development of an information Today, irrigation systems performance have increasingly
system based on water resource demand and supply. This hindered by the evolution of new demands of water and
information system gathers data from different fields adverse environmental issues. In this context, ne:w
(hydrology, meteorology and agriculture) so that accurate approaches are needed for more insight into ways of
predictions about available reserves and demand levels achieving greater efficiency at decision-taking stages
can be performed. involved in water resource management, in order to
So, this communication presents the structure of the optimize the available water resources and to help
decision support system and focuses on tactical decision making for canal management.
management information needs. So this study presents a global approach of an intelligent
The case study considered deals with a three-reach decision support system for the short term water resource
irrigation system. management of an irrigation system.
Keywords: Decision Support Systems, Information 2. THE BASIC IRRIGATION SYSTEM
Systems, Irrigation Systems.
The global objective for irrigation systems is to meet,
1. INTRODUCTION regardless of uncertainties, water demand for agricultural,
industrial and domestic uses at each discharge point while
Agriculture has, throughout History, played a major role maintaining an acceptable level of water along the
in human societies endeavours to be self-sufficient in reaches and in the reservoirs during any given period [61.
food. However, irregular floods and droughts cycles have To ensure effective water resource management, a basic
seriously impeded the attainment of such an objective. irrigation system is considered for illustration in this
This is why, for Mankind, agricultural land irrigation has study. It consists of the following elements (figure 1): an
increasingly become a challenge and water resource upstream reservoir with control gates, a sequence of
control a priority. interconnected reaches with downstream control gates
During the last century, decisive civil engineering and off-take discharge devices, a final exit section with $1
technique improvements as well as digital control flow metering device.
SO
Reach i
-j
\ I--
e(t
4s X I-
Reservoir Qi+ 1(t)
Pi(t)
Fig&l: The Basic Irrigation System
0-7803-4778-l /98 $10.00 0 1998 IEEE 3908
It appears that to cope with short run water resource
management, the operations of an irrigation system must 3. STRUCTURE OF THE DECISION
be described in two ways: SUPPORT SYSTEM
I) In terms of continuous transfer relations relating
inflows to outflows in each reach and following non- The above hybrid model of irrigation systems operatrons
linear dynamics such as: leads to the definition of a finite set of discrete
operational situations or states, to which can be attached
different short-term goals. It appears that the operations
'i(t)=f(zi(t),Q,(7),Qi+l(t),Pi(t),Si(t)) of such systems are impaired by different stochastic
ret i=l toN (1) events such as device failures, heavy rains, dry periods
and new long term goals. So the approach proposed here
where N is the number of reaches, is to do first an on-line detection state transitions, then to
Zi(t) is the downstream water level in reach i at time t, identify the current situation, and finally to reformulate,
Qi(7) is the upstream inflow to reach i at time 7, following an adaptive philosophy, an optimization
Qi+i(t) is the downstream outflow to reach i at time t, problem whose goals and constraints are in accordance
Si(t) is the spilled outflow at time t, with the current situation [4].
Pi(t) is the downstream pumped flow at reach i. So different problems arise here to make effective this
These equations can be discretized and linearized with a approach:
good approximation leading to relations such as: - the definition of a set of discrete operational situations,
- the design of a Knowledge Based System sub-
component,
- the formulation of short term optimization problems.
where h, ~ are the transfer coefficients associated to the
State Identification
linearized model and cti is a reference area for each reach The definition of such a system must follow some basic
i [7]. In this case, the upstream water reserve evolves considerations:
following relation: - only significant events with respect to the management
of the water resource must be taken into consideration,
V,,, = V, + (e, - QI, - %,).At (3) - the combinatorial multiplication of cases generated by
the different operations states of each subsystems must be
where e, is the water input rate to the reservoir and Qii the contained,
upstream inflow of reach I at time t. - every operational situation must be covered by the ser of
2) In terms of qualitative or logical terms related with discrete situations.
the degree of saturation of water levels, the intensity of Relevant discrete events for the operations of this kind of
perturbations (rains or dry periods) and the operational systems are: saturation events, failure events, discrete
state of downstream control gates, pumps and off-take decisions events.
discharge devices. Therefore, these states can be considered to be compo:jed
This description is concerned with : of three complementary components:
0 physical constraints such as: - a supply component related with the distribution of the
resource along the irrigation system and involving
mainly water levels in reaches and reservoir,
zy < zi(t)s zp
- a system component related with the operational state of
OsQi(t)<Qi(t)<QP” its devices ( sensors and actuators ),
- a demand component related with past deficits and
olP,(t)Gi(t)lpimM
‘ i=ltoN short term predictions ( meteorology ).
oIs,(t)<Sy(Z,(t)) So, the different states can be characterized by a triplet
v;,, 5 v(t) 5 v;, (p, q, r) with p E 0, q E S, r E D, where 0 is the
discrete set of sub-states related with the supply
osso, <s;“(v(t)) component, S is the set of sub-states related with rhe
system component and D is the set of sub-states relai.ed
where Qy”and emu are nominal flow capacities, with the demand component.
A qualitative description represented in figure 2, shows
Qi(t) and c(t) are actual capacities. For instance, when the Knowledge Based System analysis of the situation.
The pair (i , j) determines the operational situation whic:h
the pumping devices of reach i are down, Fi(t) =O.
can be “normal”, “critical”, “disastrous”. Making a
0 qualitative evaluations of actual water demands in
decision consists in determining which pair (i , j) among
view of past deliveries and current meteorology. Here
the possible pairs must be associated tomthestate
fuzzy techniques are of great interest to qualify and
operation [I].
compose these evaluations [3].
The purpose of this global modelling is that irrigation
system is viewed as hybrid dynamical system subject to
continuous operations broken by discrete events [4].
3909
SUPPlY
Zi , Qi , V
u
0
1
\ JJ
Operation i 0 Demand j
Figure 3: State Identification
Here the State Knowledge Based System is devoted to V,+,= V, + te, - Q,, - So,P (4
two different tasks:
- identification of present state and consequently under the restrictions:
detection of state transitions,
- diagnose of the current situation. (4 >
The identification task can be achieved for each
(3 1
component of the states. In relations with the supply
component, the identification function may be realised OS& Isp”(z,(t))
using crisp or smooth definitions of the boundaries of the
discrete state and IF-THEN rules can be used to zy” I zi (t) s zy
determine the effective membership of the supply VAi, 5 v(t) I VA,
component.
For a given set of states, the human operator interference 05sot ssy(V(t))
is necessary to define tactics to be followed. For other
situations, tactics to be followed can be deduced directly where Di is the predicted or assigned demand rate and
from the current state transition. The states transitions
that should be submitted to the human operator must be P,’ is the delivery rate for period (t, t+At) at reach i, {)-it,
defined beforehand by expert analysis. With respect to i = I to N, t E [ to, t,,+T]} is a set of deficit weightings for
the demand component, short term predictions of the objective function.
demand, based on past statistical data and current deficits rl and r2 are flow capacity restrictions, r4 and rs are state
are corrected according to external perturbations such as restrictions.
heavy or sustained rains, or such as breakdowns in the At the end of the optimization the final time constraints
distribution network. So, if the subsequent discretization can be such as:
leads to identify the current discrete state in relation to
water demand, this function provides also another
valuable information for the management of the resource:
an updated short term prediction of demand to be used in
the optimization process.
The diagnose system operates as an alert system for the Note that the optimization objective can be written
human operational manager and must be able to submit to equivalently as:
him intricate tactical choices.
to+T N
Short terms problems max 1 C&(P:.At) (02)
According to the chosen tactics, a set of relevant t=toi=l
objectives and effective constraints is selected to define
the current short term optimization problem which where predicted or assigned demand rates are no molie
defines on line reference values for the control system. present.
An acceptable formulation [2] of the standard tactical
optimization problem is of the linear form: To solve this optimization problem, a program named
DYPLEX (from “dynamic simplex”) has been developed
[5]. DYPLEX is composed of four ingredients:
min”iT;&(D: - P:).At (01) - the revised simplex method,
P;.Q; f=to 1=1
- an augmented version of the original problem,
- a compacted representation of the sparse vectors and
with ~tt+l)=~(t)+[~.Q(7))-Q+,tt)-~tt)-$tt)~t~~ matrices,
TQ
- an improved selection process for the pivot element.
i=ltoN h)
3910
Typically, the horizon of optimization for this problem is which evaluates uncertain users behavior, a decision
a week and the time is discretized on an hourly basis. module to face device failure and an optimization module
It becomes clear that to the supply component substates which determines references values and thresholds.
transitions are attached variations in the transfer In fig. 4 are displayed water levels in normal operating
coef?icients of the state equations (s,) and to the conditions. Here, a nominal water level is assumed in
maximum values of downstream outflows , restriction (r,) each reach with a random demand.
and spilled outflows, restrictions (rj) and (r6). Fig. 5 displays water level variation with failed pump in
To the system component substate transitions, are reach 2. The strategy adopted in this case consists of
attached variations to the maximum values of stocking water in the second reach and postponing
downstream outflows (equation r,) and to maximum current demand until the failed device is restored. This
values of downstream pumped flows (equation r2). results in an increase of water level in reach 2.
determining p,(t) within the interval [O,min{ P,““,D; }]. Fig. 6 takes up again the previous case but here, therl: is
saturation in the upper water level in reach 2.
Also, either or not the satisfaction of the current demand Fig. 7 shows references values and thresholds from the
is postponed until the failed devices are restored, the optimization module. These values were perceived to be
demand rates appearing in restriction rz must be the evaluation of water deliveries and inflows.
modified.
To the demand component substate transitions, are
5. CONCLUSION
attached adaptations of the criterion weightings.
The above approach of the Decision Support System is In this communication, a decision support system for
represented in figure 3.
irrigation system has been considered. It appears that to
be effective, such a decision support system is strongly
4. APPLICATION
rely with knowledge techniques and adaptive
optimization. The paper brings out the organization to
A three-reach canal with pumping station is considered help managers in fulfilling the control task and the
for the simulation. Two cases are studied for validation evaluation of water deliveries references and thresholds
and structuring the approach: for optimal operations. So, the main advantage of this
- water resource management evaluation with nominal idea is that it is a global combined approach permitting to
conditions including a strategic resource allocation and evaluate dynamically inflows and deliveries. The
demand evaluation, proposed approach has been validated trough a
- evaluation of the strategy to face device failures. simulation study involving optimization in presence of
Thus four modules have been implemented: a physical failed devices.
module which describes the system, a demand modules
.._.... -_-._ - _....-........___...
_
_....___..
___....______.._.__ ._.._ __.------
_..-..
term planning
Strategic Management/Long
--I
_...........
Long Term Constraints
Resource Supervision System & Choices
______
I____________
_____________ _
____
_.___________ __.__-
._._.____.____. _______.__. _-.__.__._ _.___ _.-.-?
.-.__.______ _,-.--.. ---.--.-..
.._..___._.._._ ____
I I
v
Calendar Human ___) Formulation and
Time ! ) State Solution
Supervisor
Identification of Short term Water
(K.B.S) Resource
Optimization
b Problem Reference
Meteorqlogy State Values &
I Diagnosis Tresholds
A
I____
________.___. _..______ _ ...___-.-
__________._. .
__.__._.. ._.-.._..
_.___--_-.-_ _._ - .-.--.--...._.-.-.--.. -..--.-_.-_
.._. ._.._.._
__ ..-.... _ _.._ II
.---.--..-.--.-
__” - __.__. -._.. _-
__.._ .,,..,..... _ -.- -.-.-.- ..-...... _..--
Measurements Measurement & Control
Figure 3: Structure of the Decision Support System
3911
2.995
T!llElnhaun Tmlc n hours Tnrwlhrs
Fig. 4: Water level in normal operating conditions
3.02
29 25 2%
0 40 60 80 0 20 40 64 80
Tune !n houn Tmxltllmm
Fig. 5: Water level in presence of failed device in reach 2 without saturation
2.5
40 60 0 20 40 60 80
Tmernk 80 0 20 10 60 8”
Tmcmhovn
Fig. 6: <ter level in presence of failed dev~~?~ach 2 with saturation
[5] R.M. Faye & F. Mora-Camino & A.K. Achaibou &
6. REFERENCES A.L. Pereira, “DYPLEX: A Large Scale Dynamical
Linear Programming Method,” LA.A.S Report No.
[I] E. Boutleux, & B. Dubuisson, “A Decision System to 98047. Feb. 1998.
Detect a State Evolution of a Complex System,”
Proceedings of the 34’ Conference on Decision &
h [6] C.M. Shafi & Z. Habib, “Sheduling of Water
Control. New Orleans, L.A.-December 1995, pp 742-747. Deliveries in the Irrigation System of the Indus Basin,”
IIMI Newsletter, Vol. 3, No. I, Jan. 1997, pp. 18-l 9.
[2] P. Carpentier & G. Cohen, “Applied Mathematics in
Water Supply Network Management,” Automatica, Vol. [7] S. Sawadogo, “Modelisation, Commande Predictive
29, N” 5, pp. 1215-1250, 1993. et Supervision d’un Systeme d’ hrigation,” These de
Doctorat U.P.S Toulouse Avril 1992, No I 161.
[3] R.M. Faye & F. Mora-Camino & A.K. Achaibou,
“The Contribution of Intelligent Systems to Water [8] S. Sawadogo & P.O. Malaterre & A. Niang & R.M.
Resource Management and Control.,” Joumtes Hispano- Faye “Multivariable Generalized Predictive Control wil:h
Francaises, Systemes Intelligents et ContrBle Avance, feedforward for on-demand operation of irrigation
Barcelone 12-l 3 Nov.96. canals,” International Workshop on Regulation of
Irrigation Canals: State of the art of research and
[4] R.M. Faye & F. Mora-Camino & A.K. Achaibou, Applications (RIC’ 97), pp. 249-257, Marrakech-Morocco
“Adaptive Optimization Approach for the Supervision of April 22-24, 1997.
an Irrigation System,” Conference on Management and
Control ofProduction and Logistics (MCPL ‘97), Volume [9] Y. H. Yacov, & D. Macko, “Hierarchical Structures
1, pp.l75-181, Campinas-SP-Brazil August 3 l- in Water Systems Management,” IEEE Trans. on
September 3, 1997. Systems, Man, and Cybernetics, July 1973, pp. 396-402.
3912
Upstream inflow of reach 1 1
0’ 1 hours
5O 5 10 20 25 30
1 - Downstream outflow of reach I
5 hours
0 15 20 25 30
4
Water level at reach I
3-
I hours
15 20 25 30
Water delrvery at reach I
0 hours
0 5 10 15 20 25 30
hours
2
Dow “stream outflow of reach 3
1 -
0 -
1 - hO”CS
0 5 10 1 5 20 25 30
hours
0 5 10 15 20 25 JO
1
w .%,cr deltrcry 8, reach 3
5 -
0 hours
0 5 10 15 20 25 30
Figure 7: Optimal reference values of a three-reach canal
3913
Related docs
