Hybrid M5‐Pruned Decision Tree and Fuzzy Logic MPPT Strategy for High‐Efficiency of PV System Tracking Under Dynamic Conditions

ABSTRACT

Maximising the power extraction of photovoltaic (PV) systems under rapidly changing atmospheric conditions remains a key challenge due to nonlinear panel characteristics, partial shading, and temperature‐induced fluctuations. Although maximum power point tracking (MPPT) algorithms such as perturb and observe (P&O) and standalone fuzzy logic controllers (FLC) are widely implemented, they suffer major drawbacks, including slow dynamic response, oscillations around the maximum power point (MPP), high computational effort, and sensitivity to expert‐defined fuzzy rules. To address these limitations, this study proposes a novel hybrid intelligent MPPT strategy that integrates the predictive modelling capability of the M5‐pruned (M5P) decision tree with the adaptive reasoning of an FLC. The proposed architecture utilises the error (E), change in error (ΔE), and duty cycle (D) from the FLC as the training dataset for the M5P model, thereby reducing reliance on heuristic rules while enabling rapid and accurate predictive control. This hybridisation enhances controller adaptability, mitigates the excessive tuning burden of conventional fuzzy systems, and improves real‐time decision accuracy under dynamic irradiance and temperature variations. A comprehensive comparative evaluation demonstrates that the proposed hybrid M5P–FLC MPPT achieves superior tracking performance relative to the standalone FLC method. The hybrid approach increases tracking efficiency from 74% to 92%, minimises steady‐state oscillations, improves voltage and current stability, and accelerates convergence to the MPP. Experimental validation using MATLAB/Simulink, WEKA, and dSPACE 1103 confirms real‐time practicality and hardware scalability. The results highlight that the proposed hybrid MPPT framework is not only computationally efficient and structurally interpretable but also robust and highly responsive, making it a strong candidate for next‐generation intelligent PV control systems operating under uncertain and highly variable environmental conditions.