The problem of irrigation canal regulation under demand delivery opera
tion was formulated as an optimal control problem. To apply the linear
optimal control theory, the Saint-Venant equations of open-channel fl
ow were linearized using the Taylor series after using a finite-differ
ence approximation on the original nonlinear, partial differential equ
ations. A proportional-plus-integral (PI) controller was developed usi
ng the concepts of linear optimal control theory. Since the order of t
he controller gain matrix was large, an optimal observer (Kalman filte
r) was designed to estimate values for the variables that were not mea
sured. An example irrigation canal with five pools was considered. Wit
h the finite-difference technique used, there was a total of 45 state
variables and five control variables (gates) in the problem. With two
measurements per pool, values for 35 state variables were estimated us
ing the observer. By subjecting the canal to random disturbances of up
to 40% of the initial inflow rate into the canal, the simulated perfo
rmance of the global feedback control algorithm along with the Kalman
filter was found to be acceptable in terms of achieving either a const
ant-volume control or a constant-level control in the canal pools in t
he presence of random disturbances in lateral flow rates.