Optimized Power Control in Grid-Connected Electric Vehicles Using Feed-Forward Neural Networks

In this paper, a feedforward neural network (FFNN)-based controller is designed for bidirectional voltage source inverters (VSI) in grid-connected electric vehicles (EVs) and presented. The suggested method addresses the inherent nonlinearities in EV charging systems while efficiently managing active and reactive power. Eliminating the dependency on system parameters and intricate mathematical modelling simplifies the control system and facilitates its implementation. High adaptability to nonlinearities and changing operating conditions is made possible by the FFNN-based controller's capacity to learn directly from input-output data. It addresses the issues of the tuning process and separate tuning for different sizes of batteries for conventional controllers that usually use the PI controller to generate reference voltage for pulse width modulation for inverters. The proposed controller is simulated and verified in various operating conditions using a 100 kW battery energy storage system, and the results show that the FFNN controller can track the active and reactive power properly while maintaining the voltage and frequency stable and reaching the steady state value in 0.4 seconds. The steady-state error for power is less than 0.05%, showing that the proposed controller performs precisely.