BRIDGING CONVENTIONAL AND INTELLIGENT MPPT TECHNIQUES: A HYBRID ANN-P&O APPROACH FOR SMART PHOTOVOLTAIC SYSTEMS
TL;DR
The demonstrated result is that the hybrid controller preserves the rapid nominal response of the ANN and substantially reduces ripple relative to conventional P&O, which demonstrates a robustness advantage of the hybrid controller over standalone ANN.
Abstract
This study investigates maximum power point tracking (MPPT) for a 250.27 W photovoltaic module connected to a buck converter. The proposed strategy combines an artificial neural network (ANN), which estimates the maximum-power-point voltage from irradiance and temperature, with a local perturb-and-observe (P&O) correction. The ANN rapidly repositions the voltage reference after an environmental change, while the hybrid P&O stage refines that reference using a small, bounded correction. Discrete simulations at 20 kHz are performed under a stepped irradiance profile of 1000, 600, 800, 400, and 1000 W/m2 at 25 C. Under the nominal PV model, conventional P&O exhibits power ripple of up to 44.23 W and response times exceeding 100 ms during some transitions, whereas the ANN and hybrid controllers respond within a few microseconds. The demonstrated result is that the hybrid controller preserves the rapid nominal response of the ANN and substantially reduces ripple relative to conventional P&O. However, because the ANN was trained using the same nominal PV model, the hybrid and standalone ANN responses nearly overlap; consequently, these nominal simulations do not demonstrate a robustness advantage of the hybrid controller over standalone ANN. An out-of-model robustness test, based on parameter mismatch without ANN retraining, is therefore identified as future work to quantify the specific contribution of P&O refinement.