Modeling of Electric Vehicle Charging Energy Consumption using Machine Learning

Several electric vehicle charging station infrastructures (EVCS) have been built in Indonesia, both private and public. A centralized management system is required to manage several charging stations. The central system uses collected historical charging transaction data to analyze patterns of EVCS energy consumption. A supervised machine learning model is applied to forecast the charging station (CS) energy consumption for the next day based on the daily transaction of the previous days. Weather condition data are also applied as additional parameters to determine their effect on charging energy consumption. Four machine learning algorithms have been evaluated, including Random Forest (RF), Support Vector Regression (SVR), XGBoost, and Multilayer Perceptron (MLP). MLP obtained the best accuracy value with an R² score of 0.85 and RMSE of 115.17 for the model without the weather variables. Meanwhile, SVR outperforms all models using a dataset with weather variables. Overall, MLP and SVR can perform well on both types of datasets. This result is expected to be used as a reference method for energy consumption prediction in the electric vehicle charging infrastructure in Indonesia.