Design and Development of a Federated Learning Scheme for EV Charging with Demand Based Priority

Precise forecasting of recharging requirements is necessary to meet the rising need for electric cars (EVs) and optimize the power system, encompassing the distribution of electricity and congestion regulation. To predict the necessity of charging electric vehicles in fifteen-minute intervals this study presents a novel deep learning framework based on Federated Learning (FedL). This paper incorporates AES-128 encryption for secure user communication and station quantization to improve security. By improving demand prediction accuracy by 13. 01 percent and lowering the error of by 9.2 percent of suggested model outperforms the traditional FedLThe need for EV charging is forecasted for the next twelve months, along with an examination of Mean Absolute Percentage Error (MAPE) as well as Mean Percentage Error (MPE) during the preceding five years. Using MATLAB this study demonstrates the value of home smart charging systems especially for overnight charging since they promote user acceptance by providing increased flexibility better coordination and higher predictability. Furthermore, it use sophisticated deep learning techniques within the FedL framework to forecast charging session and power consumption by utilizing prerecorded plugin data weather traffic patterns and event-based influences.