Model predictive control of agro?hydrological systems based on a two?layer neural network modeling framework

Water scarcity is an urgent issue to be resolved and improving irrigation water-use efficiency through closed-loop control essential. The complex agro-hydrological system dynamics, however, often pose challenges in applications. In this work, we propose a two-layer neural network (NN) framework approximate the dynamics of system. To minimize prediction error, linear bias correction added proposed model. model employed by predictive controller with zone tracking (ZMPC), which aims keep root soil moisture target while minimizing total amount irrigation. performance approximation shown better compared benchmark long-short-term-memory for both open-loop Significant computational cost reduction ZMPC achieved framework. handle offset caused plant-model-mismatch NN framework, shrinking ZMPC. Different hyper-parameters presence noise weather disturbances are investigated, time-invariant zone.