Intelligent Control for Type I Partial Power Converters in EV Charging Systems: Twin-Delayed Deep Deterministic Policy Gradient Approach

In recent years, the electric vehicle (EV) industry has experienced significant advancements, simultaneously driving substantial progress in battery technology. The evolution of battery systems necessitates enhancements in charging infrastructure to attain elevated power levels during the charging process, thereby minimizing charging time. Various algorithms have been developed for driving battery charging; however, these algorithms necessitate the creation of diverse controllers to generate precise trigger signals for the semiconductors within the various power converters utilized in charging stations. This work presents the design of an innovative model-free control system for Type I impedance network Partial Power Converter (PPC) in which a Deep Reinforcement Learning (DRL) agent generates control signals during the different charging stages. Particularly, a Twin-Delayed Deep Deterministic Policy Gradient (TD3) algorithm is used to substitute the inner control loop of traditional control systems. To this end, different agents were designed, trained, and tested inside a built simulation environment. It is worth noting that TD3-based control allows for the optimal functionality of a type I impedance network PPC within the context of EV battery charging applications, according to the specified CC-CV charging algorithm. Empirical results revealed that the battery system reached an 80% state of charge in under 8 minutes starting from an initial 20%.