Hybrid Neural Network-Assisted P&O MPPT for Boost Converter-Based PV Systems Under Variable Irradiance Conditions

Resumen

The Photovoltaic energy systems depend on effective maximum power point tracking (MPPT) methods to optimise energy extraction under varying environmental conditions. Although numerous MPPT techniques have been documented in the literature, enhancements are still needed to improve the dynamic response and minimise steady-state oscillations, particularly in real-time applications where rapid tracking is essential. This paper presents a neural-assisted adaptive MPPT approach for a boost-interfaced photovoltaic system operating under dynamic irradiance and load variations. The photovoltaic module is modelled using a single-diode equivalent circuit to capture its nonlinear behaviour, whereas a boost DC–DC converter regulates the photovoltaic array's operating point. The proposed controller integrates the traditional perturb-and-observe algorithm with a lightweight neural network that adaptively adjusts the converter duty cycle to enhance tracking precision. MATLAB simulations were conducted under rapidly changing irradiance levels and load disturbances to evaluate the performance of the controller. The results demonstrate that the proposed method achieves an MPPT efficiency exceeding 98%, reduces steady-state oscillations by approximately 35%, and decreases convergence time by nearly 25% compared to conventional perturb-and-observe and incremental conductance techniques. These advancements underscore the robustness and efficacy of the proposed strategy for fast and reliable photovoltaic power extraction in practical energy conversion systems.