A method for estimating the state of charge (SOC) of a lithium-ion battery system based on artificial intelligence (AI) is provided. In the method, the relationship between the charging data segments and the SOC of the battery system is established through deep learning, and the SOC at any stage of the charging process can be corrected. SOC in a discharging process is estimated through ampere-hour integration. The estimation method is adaptively updated with a change in the working state of the battery system.