A Study on a Charge and Discharge Control Method of Multiple Storage Batteries Utilizing Machine Learning for Voltage Control in Distribution Grids

Due to the secondary use of battery storages installed in electric vehicles, it is expected that the installation of battery storages on consumer sides will increase significantly in the future. When storage batteries and photovoltaic power generation are used together, reverse power flows to distribution grids may affect voltage rises of distribution grids. In this study, we will report on the charge and discharge control method that utilizes Q-learning to suppress the influence on the distribution grids as much as possible, assuming that the consumers will autonomously and decentrally control the charge and discharge of storage batteries. We evaluated the applicability to multiple storage battery charge and discharge control when PV output changes are assumed in the power distribution system. We confirmed that adding a function to maintain the output of the storage batteries to avoid their repeated recharging operation is more effective in controlling the voltage of the storage batteries within an appropriate range. We also evaluated that the system is applicable to the case where the PV output pattern is varied.