Machine-Learning Based Fault Monitoring System for Electric Vehicle Onboard Chargers

Improving the electrical systems' resilience is essen-tial for a faster and smoother migration to electric vehicles (EVs), One of the essential subsystems in EVs is the power electronic converter (PEC), which provides power to the electric motors and is reconfigured for grid charging of the battery bank. This paper proposes a multiple fault classification/monitoring system for a type of PEC widely used in EVs. Hence, this study can be used by a supervisory system that can perform corrective actions, ensuring a continuous operation of the charging system. The main objective of this paper is to determine different types of faults that can cause a malfunction during the operation of the onboard converter. The detection procedure is based on the machine learning technique named Random Forest Classifier (RFC); it focuses on the semiconductor components and the grid status when powering the system for battery charging. The proposed system performance is numerically compared with a Support Vector Machine (SVM) system, can be generalized to include other faults, and shows superior performance on training and execution times and accuracy.