Electric vehicles are a critical component of environmentally sustainable mobility solutions. In addition to the extensive provision of charging infrastructure, ensuring reliable charging processes is crucial to achieving widespread public acceptance. This is challenging, because the underlying processes are more complex than those involved in refueling a conventional car. A variety of charging technologies and infrastructures are available around the world. However, implementing all systems at a single charging station is not economically viable. Moreover, deviations from standards and implementation errors can result in incompatibilities between vehicles and charging infrastructure. This work builds upon an existing Matlab/Simulink based model of electric charging, in order to simulate communication and energy transfer. Our model incorporates realistic fault scenarios, such as message timeouts, which can occur randomly. Secondly, we employ deep reinforcement learning (DRL) agents to adapt adjustable calibration parameters, thereby achieving fault correction. The simulation results of both DRL approaches, one model-free and one model-based, prove the concept of their ability in fault recovery. Our future efforts will concentrate on improving the DRL algorithms, and expanding the approach to optimize the electric charging duration while maintaining the health of the involved hardware, particularly the battery. Our long-term objective is to deploy this technology in a real world setting as part of an automated remote repair application.