Thermal runaway is one of the critical causes of safety issues in lithium battery. To accurately estimate the thermal state of lithium battery

this paper proposes an algorithm consisted of adaptive extended Kalman filter (AEKF) and kernel extreme learning machine (KELM) to estimate the internal temperature of lithium battery. The equivalent circuit of lithium battery temperature estimation based on Bernardi heat generation model is established

its parameters are identified by genetic algorithm (GA)

and the internal temperature of lithium battery is then estimated by AEKF algorithm. With the terminal voltage

working current

surface temperature estimated value and internal temperature estimated value of lithium battery as input and internal temperature estimated error as output

KELM is used to establish temperature estimation error compensation model to compensate the temperature estimated result. In order to verify the effectiveness of the method

experiments in constant current charge-discharge and dynamic stress test (DST) conditions are carried out at different ambient temperatures. The experimental results show that the estimation errors produced by this method are all less than 0.34 ℃ in each test condition

and estimation accuracy and robustness are significantly improved compared with other methods.