Application of machine learning decision tree based power switching control for a standalone PV system with grid support.

Ensuring reliable power delivery in standalone photovoltaic (PV) systems is challenging due to the intermittent nature of solar irradiance and dynamic load variations. Conventional switching control schemes often fail to maintain optimal performance under such varying conditions, leading to reduced efficiency and delayed source transition. Although several studies have explored PV converter control and MPPT algorithms, the integration of machine learning techniques for real-time switching control in standalone PV systems with grid backup remains an emerging area. This research work aims to address the identified gap by developing an ML- Decision Tree based switching strategy that enhances power management and system efficiency. This paper presents applying switching rules to choose the correct power source and train a machine learning model to predict future switching for a standalone PV system that operates autonomously and sources power from the utility grid only when solar energy becomes insufficient. The proposed control framework employs real-time monitoring of PV array voltage, current, irradiance, and load demand as input features to the trained DT model, which predicts the optimal switching decision between PV and grid supply. The Simulation work carried out in this paper using MATLAB/Simulink demonstrates application of machine learning - Decision Tree based switching to automate switching operation with reduced switching losses and improved voltage regulation compared to rule based conventional switching in a PV system.