Real-time nonlinear model predictive energy management system for a fuel-cell hybrid vehicle

Abstract

Proton exchange membrane fuel cell is considered as one of the most efficient sources of renewable energy. Time-varying dynamic and nonlinear equations are two major factors that make control and power optimization of fuel-cell vehicles challenging. In this paper, by using a comprehensive PEMFC vehicle model and nonlinear model predictive controller, a novel energy management method is represented. By considering both of the regenerating brake power and fuel-cell output power as the controller inputs, the steady-state equations have been derived in a nonlinear form. Also, the NMPC contains a constrained quadratic cost function which has been minimized to make the controller inputs completely optimized. Moreover, the two controller references using in this article have been achieved based on experimental FTP drive-cycle data. Finally, the optimized cost-function weighting matrices will be calculated by genetic algorithm, and then, the real-time controller inputs are applied to the vehicle model to get better results on vehicle range during an online procedure, and the online results are compared with the experimental drive-cycle data.

Model predictive control using machine learning for voltage control of a PEM fuel cell stack

“…Here, the NMPC is designed using the LSTM model from section III. The cost function for this system is derived in a quadratic form which ends up to a convex and smooth solution based on the system constraints [13]. The generated quadratic cost function which is applied to the NMPC can be expressed as Eq.…”