Effective source selection of solar-grid powered electric vehicle charger circuit using machine learning algorithm

The rapid adoption of electric vehicles (EVs) has increased the demand for charging systems integrated with renewable energy sources for efficient operation. Solar-grid integrated systems offer a promising solution; however, photovoltaic (PV) fluctuations and varying grid conditions reduce power efficiency and increase battery stress. To address these challenges, a machine learning-based source selection method for a Quasi-Z-Source Network Converter (QZNC) used in EV charging is proposed. Different machine learning algorithms, including Fine Tree, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Weighted KNN, and Random Forest, are implemented and analyzed. Simulations are carried out in MATLAB/Simulink under different load and power conditions to evaluate system performance. Under the considered MATLAB/Simulink simulation scenarios, the Random Forest classifier achieved the best performance with a classification accuracy of 99.67% and an AUC value of 1.0 on the generated test dataset. The proposed system attains a maximum efficiency of nearly 90%, maintains voltage fluctuations below 2.5 V, and reduces charging cost by about 13% compared to conventional methods. In addition, the use of a fast-recovery diode minimizes switching noise and enhances converter performance while supporting bidirectional vehicle-to-grid (V2G) operation.