Abstract: The increasing maturity of lithium-ion battery technology provides important support for the development of new energy power generation and electric vehicles. Lithium-ion battery adopts organic electrolyte, which is easy to trigger exothermic side reaction of battery material after failure, leading to thermal runaway of battery. And then it is likely to evolve into serious accidents such as combustion and explosion. State of health(SOH) is an important parameter for fault diagnosis and safety warning of lithium battery energy storage system. Accurate estimation of SOH is an effective way to improve system safety. Therefore, a temperature differential curve(DT curve) based lithium-ion battery health status assessment algorithm is proposed to fully extract the temperature information that is highly correlated with the battery health status on the surface of lithium-ion battery. The maximum point of DT curve and the voltage difference between the two extreme values in the battery charging process are taken as the characteristic quantity of SOH estimation. The SOH estimation model based on back propagation(BP) neural network is built. The test results of experiments and simulations finally show that the proposed method can effectively improve the SOH estimation accuracy of lithium-ion batteries.