Neural Network Controller to Manage the Power Flow of a Hybrid Source for Electric Vehicles

This paper proposes a neural network controller for a DC/DC converter used to manage the power flow of an active hybrid energy source for Electric Vehicles. The energy source is composed of a battery, an ultracapacitor and a Dual Active Bridge DC/DC converter used to optimize the power distribution between the energy sources involved. The neural network control method applied to this bidirectional converter has many advantages: reduction of the computational time, decrease in the amount of data stored and improvement in the transient response. This method is simulated in an Electric Vehicle application using a normalized driving cycle. The simulation results will show the improvements made by this control method compared to the conventional PI controller, in terms of improving the power management, reducing stress on the sources and increasing robustness due to reference changes.