Enhancing EV charging predictions: a comprehensive analysis using K-nearest neighbours and ensemble stack generalization

Ensemble Stacking Generalization has emerged as a viable method for forecasting Electric Vehicle (EV) charging behaviour. This method uses a variety of machine learning methods, such as Decision Trees (DT), Random Forests (RF), and k-nearest Neighbours (KNN), to improve predictions about charging behaviour, focusing on stay duration and energy consumption. These forecasts are based on previous charge data, and the methodology considerably improves predicted accuracy while reducing model variation, overcoming the drawbacks of single-regressor models. A thorough investigation of ACN data was used to properly collect the dataset relevant to Electric Vehicle (EV) energy usage and session length. The crucial details of EV charging behaviour were painstakingly documented in this dataset, including session length and kWh provided. A wide range of statistical evaluation measures were utilized to assess the suggested approaches' effectiveness thoroughly. The outcomes of our efforts to anticipate Electric Vehicles (EVs) energy consumption and session length highlight the superiority of Ensemble Stacking Generalization. This method regularly outperformed competing models by producing results that met the standards established by chosen evaluation metrics. The significance of this is that it emphasizes how the concepts of stacking approaches may be used to increase the accuracy of EV energy usage predictions considerably. It's also critical to note that forecasts significantly improve over earlier studies that used the same dataset. The ability of Ensemble Stacking Generalization to tackle the complexities of EV charging behaviour prediction from an original aspect is highlighted by this as being both inventive and robust.