Prediction of Electric Vehicle's Annual Accessibility to Chargers for Providing Ancillary Services Using an Efficient Random Forest Method

The use of electric vehicles (EVs) in the power system has grown phenomenally, and when combined with smart grids, a wealth of raw data is accessible. It is challenging to plan and schedule for EVs due to the randomness of their driver behavior and their uncertainties. To cope with these uncertainties, a supervised machine-learning framework (Random Forest) is developed using an open-source application (emobpy) that simulates EVs to help EV aggregators and drivers predict annual charger accessibility. Since ML models are complex black boxes to decipher, a game theory method SHAP (SHapley Additive exPlanations), is employed to indicate the impact of each feature on the model outcome. EV aggregators can plan their market participation using this model. A simulation of the proposed framework in the frequency-controlled normal operation reserve market grew EV aggregators' revenue, indicating its effectiveness.