Improved Control of DC-DC Converter for Fuel Cell and Battery Hybrid System Based on Ant Colony Optimization and RL-TD3 Agent

Based on the increasing use of a Proton Exchange Membrane-Fuel Cell (PEM-FC) stack, and starting from a benchmark that has such a power generation structure including a battery, this article presents the global architecture of the proposed system and its main components, where the general objective is to keep the VDC voltage of the DC type circuit constant. Due to the long response time of the PEM-FC stack for internal thermodynamic reasons, in the proposed system, we present the improvement of the performance for the Proportional Integral (PI) controller of the DC-DC converter by using a Computational Intelligence-Ant Colony (CI-ACO) algorithm for obtaining the optimal values of the tuning parameters for the PI controller, but also by using a Reinforcement Learning -Twin-Delayed Deep Deterministic Policy Gradient (RL-TD3 agent), which through the correction signals provided contributes to obtaining superior control performance. The comparative verification of the performance for the proposed control systems was performed in the Matlab/Simulink programming environment.