Real-time neuro-adaptive PI control of soil moisture by using a hybrid model

In the agriculture developed in the mountain valleys of Argentina, the efficient use of water for irrigation is essential for the development and sustainabilily of agricultural enterprises. In order to address this challenge, it is proposed to develop a hybrid model to represent as faithfully as possible the dynamics of water content in an irrigated soil, including water extraction by a crop. For this purpose, a mathematical model of the process is formulated based on the general How equation, which has been solved by means of finite differences. A radial-based neural network is incorporated into this structure to compensate off-line the model output at a point on the ground, identifying the output error. In addition, this study incorporates the design of an adaptive irrigation controller for unknown dynamics. The design is based on sliding surfaces in combination with PI and neural networks, with the goal of control objective is to maintain the soil water content at reference values setting. © 2023 Universitat Politecnica de Valencia. All rights reserved.