Published Paper Details:

As electric vehicles (EVs) continue to be embraced by users, a more efficient power management system at EV charging stations has become more necessary, hence the importance of demand forecasting. The present work deals with the prediction of the energy demand at EV charging stations by making use of machine learning and time series models. Also, models such as SARIMA, Random Forest, XGBoost, and H2O.ai's AutoML have been evaluated in forecasting energy consumption in relation to fleet size, the time of the day, respective weather conditions, and the grid load. Our research attempts to illustrate the advantages and disadvantages of each model in question on the market while noting that SARIMA model was the most precise, as it produced best scores based on explained models such as Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and log-likelihood values. The research elaborates on how the non-linear data relationships, the high computation requirements and seasonality present challenges and also emphasizes the need for the models to be interpretable and scalable due to the rise of the EV market. This paper also addresses the issue of how these models can be utilized within charging infrastructure in a smarter way and ways of integrating such charging infrastructure as involved predictive analytics into a smart grid system, aiming at efficiency enhancement, especially with regards to adoption of renewables. Findings lead to the conclusion that these efficiency-enhancing measures are also beneficial in enhancing the quality of energy predictive procurement processes.