On the Approximate Suboptimal Control by Neural Network- Rainfall Observer

JULIÁN A. PUCHETA; GUSTAVO JUAREZ; CRISTIAN M. RODRÍGUEZ RIVERO; SERGIO O. LABORET; H. DANIEL PATIÑO; VICTOR H. SAUCHELLI

Buenos Aires

Congreso; IEEE ARGENCON 2016; 2016

Universidad Tecnologica Nacional - FR Buenos Aires

This paper presents an approach forapproximate suboptimal control of nonlinear systems withconstraints by neural network based rainfall observer forguiding crop growth in extensive agriculture. We propose aneural-network rainfall observer approximation by means ofhistorical rainfall information. The goal is to obtain a closeloopoperation with rainfall information, whose design is basedon optimal control theory. Thus, the neurocontroller designproposed helps to drive the growth development of thecultivation as cost function and final state errors areminimized by physical constraints on the process variables.Therefore, it is possible to establish the control scheme andpolicy according to the criterion that generates the highestprofit margin in the process. The contribution shows anoptimal policy to guide the crop from an initial to a desiredstate. The estimates are consistent in a weak sense, and thequestion whether they are pointwise consistent is still open.Nevertheless, in order to assess the performance and practicaltractability of the neurocontroller, real data andcomputational results are shown for soybean crop at SantaFrancisca, Cordoba, Argentina.