Learning-Based Practical Nonlinear Predictive Controller for Solar Thermal Collector Fields

Abstract

Translate Article

This study presents a learning practical nonlinear model predictive controller (LPNMPC) designed to effectively control solar collector fields (SCFs), considering particular challenges due to nonlinear dynamics coupled with varying time delays, parameter uncertainties, and disturbances. The LPNMPC introduces an adaptive Oracle function computed via recursive least squares with exponential resetting (ER + RLSs) to estimate and correct model errors. Motivated by real-world plant settings, simulations using a validated SCF model from the CIESOL research center at the University of Almería, Spain, demonstrate the controller’s ability to overcome significant model parameters and delay uncertainties under challenging scenarios, with up to 27% fewer errors compared to the stat-of-the-art PNMPC methods. Experimental tests further validate the LPNMPC’s effectiveness, showing smooth control, accurate reference tracking, and reliable temperature regulation even under cloudy conditions, confirming its applicability in real-world SCF settings.