Machine Learning-Based Dynamic Hosting Capacity for Bidirectional EVs

The active participation of Electric Vehicle (EV) users in distribution system management applications is increasing with the developing charging station technology. Bidirectional energy flow with Vehicle to Grid and Grid to Vehicle modes facilitates electrical energy planning. Especially with Hosting Capacity (HC) analysis, the quality energy capacity of the system can be increased and a flexible grid model can be provided. In this study, various EV integrations to the European low voltage test system are examined. The uncertainties of EV consumers are determined by Monte Carlo (MC) simulation. OpenDSS program, an open-source distribution system simulator, is used to analyze performance indices such as minimum voltage, maximum voltage, and overload. The maximum HC for EVs is determined based on power flow analysis results. In addition, Machine Learning (ML) applications such as classification and regression are performed using the obtained dataset. Operational suitability and integrated EV power are estimated for various scenarios. The results of the used ML methods are compared. The aim of the study is to propose a model for the distribution system operator to make the right planning according to the energy quality indexes. The proposed model aims to determine the HC value for any scenario and to accelerate the alternative planning.