Abstract: During the charging and discharging operation of lithium-ion batteries, its core temperature directly reflects the battery state, which is an important safety performance indicator. However, the core temperature of the battery cannot be measured directly, so an accurate core temperature estimation method must be developed. The core temperature prediction model of Bi-directional long short-term memory(Bi-LSTM)neural network for the lithium-ion battery is established by taking the measurable variables(current, voltage, ambient temperature and surface temperature) which have a high correlation with core temperature as the input. The particle swarm optimization(PSO) algorithm is introduced to optimize Bi-LSTM’s parameters and improve its prediction accuracy. The experimental results show that under different charge and discharge conditions, compared with the estimation methods such as decision tree method and random forest method, this method can accurately predict the core temperature of lithium-ion battery under wide ambient temperature, and the estimation accuracy of core temperature is the highest.

Key words: lithium-ion battery, core temperature estimation, bi-directional long short-term memory neural network, particle swarm optimization algorithm