Bayesian Inference for Feedback Control. II: Surface Irrigation Example

Abstract

Bayesian inference is applied to the real‐time feedback control of a basin irrigation system. Estimates of the Kostiakov k (infiltration parameter) and Manning n (Roughness parameter) are obtained during water advance so that the optimum cutoff time can be determined. Bayesian inference is used to determine these parameter estimates from: (1) Estimates from observation of advance time and distance and solution of the zero‐inertia border irrigation model; and (2) either historical estimates of parameters or subjective estimates made by the irrigator. Bayesian likelihoods are used to characterize the error in parameter estimates made from observation. These likelihoods are developed from observation of prior irrigations and represent statistically learned patterns for that particular field. The method is demonstrated on a 32‐ha field with eight level basins. It is shown that the Bayesian inference has some potential for improving real‐time control of surface. irrigation systems.