Dynamic PV modeling using fuzzy logic and genetic algorithm for enhanced mono crystalline array performance

Abstract

This paper proposes a novel photovoltaic (PV) array model using an equivalent single-diode electrical circuit with three time-varying parameters: the diode quality factor, series resistance, and shunt resistance. The goal is to determine an optimal fuzzy inference mechanism to accurately predict the evolution of each parameter under varying climatic conditions, such as solar irradiance, cell temperature, and PV voltage. A dataset of 500 experimental samples, collected from three series-connected mono crystalline PV panels, was used-300 samples for model training and 200 for validation. The proposed design normalizes input data to [0,1] and feeds it into an initial fuzzy mechanism with predefined fuzzification rules. The mechanism generates normalized outputs, which are denormalized to compute predicted PV currents. These are compared to actual measurements to calculate modeling errors, aggregated as mean squared errors (MSEs). A genetic algorithm (GA) minimizes the MSEs by optimizing the fuzzification rules, retaining only the most effective ones. This process iterates until a final fuzzy mechanism with reduced fuzzification rules is achieved, capable of supervising the evolution of each adjustable parameter under varying climatic conditions. Experimental validation confirms the accuracy and robustness of the proposed adjustable-parameter PV model, outperforming conventional fixed-parameter models. This approach provides a reliable framework for PV system modeling, significantly improving predictive accuracy and adaptability in real-world conditions, with potential applications in advanced MPPT controller synthesis and renewable energy system optimization.