A PSC-aware ANN-based global MPPT strategy using power–voltage curve probing for photovoltaic systems

The presence of partial shading in photovoltaic (PV) generates unequal irradiance across modules, leading to a multi-peak power–voltage profile with several local operating points, significantly degrading the performance of conventional gradient-based maximum power point tracking (MPPT) methods. This paper proposes a Partial Shading Conditions PSC -aware artificial neural network ANN-based global MPPT strategy that reformulates global maximum power point GMPP tracking as a direct voltage estimation problem. The method employs a P–V curve probing mechanism based on five normalized power samples measured at predefined voltage ratios, enabling shading pattern identification without irradiance or temperature sensors. A feed-forward MLP trained offline using 3000 operating cases—including uniform irradiance, static PSC, severe multi-peak PSC, and dynamic shading scenarios—predicts the optimal operating voltage in real time. Simulation results demonstrate improved tracking accuracy, faster convergence, and reduced steady-state oscillations compared with P&O, Incremental Conductance, and PSO algorithms. Under severe PSC, the proposed approach achieves 93.93% MPPT efficiency, while providing up to 12% higher energy extraction under dynamic shading conditions. The proposed framework shows strong suitability for real-time embedded PV control applications.