A comprehensive data analysis for aggregate capacity forecasting in Vehicle-to-Grid applications

The integration of electric vehicles (EVs) into the Vehicle-to-Grid (V2G) framework necessitates accurate forecasting of the aggregated available capacity (AAC). This paper introduces a method for extracting AAC values from generic mobility data and other readily available data sources obtaining a comprehensive dataset that encompasses geographical information through GPS coordinates, meteorological data, and the consideration of national holidays and weekends. Different predictive models based on machine learning, such as Neural Networks, Long Short-Term Memory networks, Tree Ensembles, and Gaussian Process Regressors were implemented. The application of such models on different multidimensional datasets seeks to enhance the precision of capacity predictions, explaining and modeling the complex dependencies within the input data and the target available capacity, ultimately improving the efficiency and reliability of V2G systems. As the prevalence of EVs continues to rise, the proposed methodology offers a valuable tool for grid operators, energy planners, and stakeholders to optimize the integration of EVs into the broader energy landscape and facilitates the strategic planning of aggregator hub locations.