Intelligent Power Control in Smart Photovoltaic Systems Using M5‐Pruned Decision Tree for Enhanced Grid Voltage Modulation

Abstract

This paper presents an advanced and smart enhancement to the direct power control (DPC) strategy using grid voltage modulation for three‐phase photovoltaic (PV) inverters. It introduces and evaluates three DC‐link voltage control techniques: the proportional‐integral (PI) controller, the fuzzy logic controller (FLC), and a novel M5‐Pruned (M5P) decision tree–based algorithm. While PI‐based DPC remains widely used, it is often constrained by its sensitivity to gain tuning, limited adaptability, suboptimal dynamic response, and not ideal decoupling of active and reactive powers. FLC offers greater flexibility and can handle nonlinearities more effectively, yet it still lacks precise control and structured scalability. To address these limitations, this study proposes the M5P‐based control approach, a data‐driven, self‐adaptive strategy that combines model transparency with the ability to handle complex system behaviour efficiently. Simulation results show that the proposed M5P method significantly reduces total harmonic distortion to 0.20%, outperforming both PI (0.57%) and FLC (0.53%) controllers. Furthermore, it achieves complete decoupling of power components, enhances dynamic stability, and eliminates the need for manual gain tuning. The methodology is validated through extensive simulations in MATLAB/Simulink, highlighting its effectiveness under both steady‐state and transient conditions. These results establish the M5P‐based controller as a promising candidate for next‐generation intelligent PV grid integration systems.