ML based Prediction for Grid support in a Solar Photovoltaic Electric Vehicle Charging Station

In the wake of environmental protection and all other inter-continental resolutions, the transportation is slowly shifting towards less polluting alternatives. Electric vehicle (EV) transformation in the transportation sector would be a better choice to reduce huge amounts of direct greenhouse gas emissions (GHG), though there are many factors involved such as source of power used for charging, pollution due to manufacturing process and disposal of hazardous materials that adds up to polluting factors. To maximize the reduction of GHG emissions, charging of EVs with a renewable energy source and reuse of EV parts in a sustainable manner is a crucial step. Globally, the governments have started supporting Electric Vehicle transportation with many schemes to encourage public interest for buying [1]. Accommodating electricity for conveyance would soar the power requirement for charging and hence increasing the burden on the grid. Under these conditions existing grid has to be supported with renewable power sources, to reduce the stress on the grid and environmental pollution. The solar photovoltaic (PV) systems are considered to be the best suitable choice for off-grid and grid-tied EV charging stations. As the solar power is intermittent in nature, it cannot provide a stand-alone and reliable solution. But if a smart controller for predicting the optimized utilization of grid power, this solar powered EV charging station can act as a sustainable power source. The proposed work deals with grid power forecasting based on solar PV availability and EV charging demand requirement in a Solar PV based EV charging station. A machine learning oriented model with a set of algorithms applied to datasets and correlation analysis using a python program was also carried out to identify the most dependent inputs and least dependent inputs for more sophisticated analysis. A linear SVM model with least RMSE of 0.07282 is found as the best technique for this implementation.