Rapid control prototyping for PEMFC electrical vehicle powertrain based on fractional order SMC and RL-TD3 agent hybrid control with beluga whale optimization algorithm

The global transition toward sustainable energy systems demands advanced technologies capable of achieving substantial reductions in transportation-related emissions. Proton Exchange Membrane Fuel Cell (PEMFC) powered electric vehicles represent a compelling pathway; however, their operational performance is highly contingent upon maintaining a stable DC bus voltage, which is inherently challenged by the nonlinear electrochemical behavior of the fuel cell stack and by rapid, unpredictable load fluctuations. In this context, the article proposes a new control architecture for the PEMFC powered DC-DC boost converter, with emphasis on enhancing robustness and efficiency under dynamic operating conditions. The mathematical models and Simulink implementations of the PEMFC stack, the DC-DC converter, and the complete propulsion chain, including an electric vehicle benchmark, are presented. Several control paradigms for maintaining a constant voltage in the intermediate DC bus are comparatively analyzed, including Proportional Integral (PI), Fractional-Order (FO) PI, Sliding Mode Control (SMC), and FOSMC. It also demonstrates the stability and robustness of the FOSMC controller, which provides the best results among the analyzed controllers. Furthermore, an advanced combination of the baseline FOSMC controller with a Machine Learning component is introduced, designed, trained, tested, and implemented as a Reinforcement Learning Twin-Delayed Deep Deterministic (RL-TD3) agent that provides real-time adaptive corrections. The tuning parameters of the controllers are optimized using the Beluga Whale Optimization (BWO) metaheuristic algorithm. The Software-In-the-Loop (SIL) results demonstrate the superiority of the FOSMC controller enhanced with the RL-TD3 agent, offering robustness to both parametric variations and load disturbances. In the Hardware-In-the-Loop (HIL) stage, starting from a Matlab/Simulink benchmark, the propulsion chain of a PEMFC powered electric vehicle supplied through a DC bus is further validated, including the implementation of an automatic speed-regulation control at a prescribed reference. This configuration enables Rapid Control Prototyping (RCP) using an OPAL-RT OP4512 real-time platform.