Optimizing Photovoltaic Grid-Connected Power Systems Through Artificial Intelligence and Robust Nonlinear Control

Abstract

Photovoltaic (PV) systems are highly sensitive to stochastic environmental variations, particularly irradiance and temperature, which complicate the task of consistently operating at the maximum power point (MPP). This paper presents a novel dual-layer control strategy to optimize the performance of a single-phase grid-connected PV system. The proposed system integrates an Artificial Neural Network (ANN) model for real-time estimation of MPP voltage with a Lyapunov-stable nonlinear backstepping controller to regulate a DC-DC boost converter. To ensure seamless grid integration, a Phase-Locked Loop (PLL)-based Proportional Resonant (PR) controller and a PI regulator are implemented to maintain DC-link stability and minimize harmonic distortion. Unlike conventional ANN-MPPT approaches, the proposed method decouples MPP prediction from control regulation, allowing robust and fast dynamic response under varying climatic and load conditions. Extensive simulations in MATLAB/Simulink validate the effectiveness of the approach, demonstrating fast convergence to the MPP (within 30 ms), precise DC-link regulation, and excellent grid synchronization. Notably, the system achieves a low Total Harmonic Distortion (THD) of 0.2148%, outperforming existing benchmarks. Comparative analysis against P&O and sliding mode controllers confirms the superior efficiency, robustness, and power quality of the proposed architecture, highlighting its potential for scalable deployment in real-world grid-connected PV applications.