Currently the widespread application of electrochemical energy storage technology raises prominent concern on the safety operations of battery energy storage station. The traditional battery management system (BMS) can only support measurement of battery cell’s voltage, current and temperature. And due to limited hardware processing power, data bandwidth and network time delay, it is a key technical problem to master the healthy and safety of battery energy storage system with a large number of battery cells. The application of machine learning for lithium battery state prediction enables better safety management of battery energy storage system. For lithium-ion battery safety management requirements, this paper first provides an overview of related research on the mechanism of lithium battery abuse and thermal runaway. Then, the architecture of lithium-ion battery management system and its application characteristics are discussed and summarized, the usage of machine learning methods in the health and safety state analysis of lithium batteries are introduced in detail. Based on the above review, this paper provides a prospect of the safety management of lithium batteries in energy storage power stations. This work is supported by the National Natural Science Foundation of China (No. U1966214).