Artificial neural network-based energy management for fuel cell–battery–supercapacitor hybrid electric vehicles
Abstract
This paper presents an artificial neural network (ANN)-assisted energy management system (EMS) for a fuel cell–battery–supercapacitor powered electric vehicle. The proposed strategy addresses the coordinated operation of energy sources with different dynamic characteristics while maintaining stable DC-bus operation and reliable traction performance under dynamic load conditions. The fuel cell is operated as the primary smooth power source, the battery provides medium-duration energy support, and the supercapacitor compensates fast transient power variations. The ANN receives load power, battery state of charge, and supercapacitor state of charge as inputs and generates the fuel cell reference power for supervisory control. Unlike a purely black-box approach, the proposed EMS combines ANN-based reference generation with physically interpretable residual power allocation and DC-bus feedback stabilization. A complete MATLAB/Simulink model including source dynamics, converter interfaces, DC bus, and permanent magnet synchronous motor drive is developed for performance evaluation. The ANN achieves training, validation, and testing mean squared errors of 0.001 822, 0.001 936, and 0.001 994, respectively, with an R2 value of 0.942 687. Simulation results show smooth fuel cell operation, controlled battery support, rapid supercapacitor transient compensation, and regulated DC-bus performance. The findings demonstrate that the ANN-assisted EMS provides a practical and computationally simple supervisory control framework for coordinated power sharing in hybrid electric vehicle power systems.