Robust Solar Tracking with Neural Network Predictive Modeling and Sliding Mode Control

Abstract

This research presents a hybrid control technique for high-precision dual-axis solar tracking in photovoltaic systems, combining sliding mode control (SMC) with artificial neural networks (ANNs). An ANN model predicts the sun's altitude and azimuth angles based on time and date information. These predicted angles serve as reference inputs to an SMC algorithm that governs the rotational speeds of two DC motors adjusting the solar tracker's altitude and azimuth orientations. The SMC ensures robust tracking by calculating control signals that drive the DC motors to accurately follow the sun's trajectory, while leveraging the ANN's predictive capabilities. The proposed ANN-SMC approach mitigates uncertainties, rejects disturbances, and accounts for system nonlinearities, enabling optimal solar energy harvesting. Simulation results demonstrate the strategy's effectiveness, achieving highly accurate sun tracking with a mean absolute error below 0.09° for altitude and 0.27° for azimuth angles. This integration of neural networks and sliding mode control yields an efficient solar tracking system that maximizes energy yield.