Smart EV Charging Architecture using Predictive Machine Learning and Hybrid Renewable Energy Integration
S. Lakshmi, Sriram Srinivasan, R. Tamilamuthan, S. Sreedevi, B. Pandyselvi, A. Antonycharles
Abstract
Enhanced development in electric vehicles (EVs) has led to the pressure on infrastructural roll-outs to be smart and able to be sustainable and have minimal grid stress, lower operational-costs as well as maximization in harnessing renewable energy. This paper contributes a framework cobbling together solar and wind energy, a hybrid energy storage system (HESS) and an adaptive scheduling algorithm with a predictive machine learning model. In contrast to the previous systems that implement either source coordination or load shifting, the proposed system combines the hybrid renewable energy, Vehicle-to-Grid (V2G) / Grid-to-Vehicle (G2V) two-way interaction and real-time tariff optimization schemes in a single scalable system. The predictive module predicts short term charging demand by using past load, tariff and weather information using a Long Short-Term Memory (LSTM) network. It was tested using a year of charging logs with the model giving an RMSE (Root Mean Square Error) of 0.15 kWh and a MAPE (Mean Absolute Percentage Error) of 4.6 % which gave a credible forecasting. Then, the adaptive controller picks the best power source (solar, wind, battery or grid) given the anticipated demand, availability and prices. Simulation outcomes indicate that 85 % of renewable energy can be used, that there is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4 4 \%}$</tex> decrease in the average charging cost and a 35% increase in the efficiency of the load distribution compared to conventional grid-only systems. Surveys also show that satisfaction levels have increased by 22 percent and this is more in the area of cost transparency and speed in charging. The results prove that the architecture is technically sound, costeffective and easy to use, which has a possibility of application in smart cities and fleet management. Scalability, cyber security and employment of advanced deep learning models to achieve real-time optimization is the direction of future work.