Adaptive model reduction and state estimation of agro-hydrological systems

TL;DR

This paper introduces an adaptive model reduction and state estimation approach for large agro-hydrological systems, improving accuracy and computational efficiency in precision irrigation management.

Contribution

It develops a structure-preserving adaptive model reduction method combined with an adaptive MHE algorithm for large-scale agricultural fields.

Findings

Adaptive MHE outperforms original MHE in simulations.

The approach effectively estimates soil moisture in large fields.

Simulations demonstrate computational efficiency and accuracy improvements.

Abstract

Closed-loop irrigation can deliver a promising solution for precision irrigation. The accurate soil moisture (state) estimation is critical in implementing the closed-loop irrigation of agrohydrological systems. In general, the agricultural fields are high dimensional systems. Due to the high dimensionality for a large field, it is very challenging to solve an optimizationbased advanced state estimator like moving horizon estimation (MHE). This work addresses the aforementioned challenge and proposes a systematic approach for state estimation of large agricultural fields. We use a non-linear state-space model based on discretization of the cylindrical coordinate version of Richards equation to describe the agro-hydrological systems equipped with a central pivot irrigation system. We propose a structure-preserving adaptive model reduction method using trajectory-based unsupervised machine learning techniques. Furthermore, the adaptive MHE algorithm is developed based on an adaptive reduced model. The proposed algorithms are applied to a small simulated field to compare the performance of adaptive MHE over original MHE. Finally, the proposed approach is applied to a large-scale real agricultural field to test the effectiveness and superiority to address the current challenges. Extensive simulations are carried out to show the efficiency of the proposed approach.