The estimation method based on the electrochemical impedance spectrum can comprehensively describe the reaction process of Li-ion battery without causing damage to the battery, and is suitable for estimating the state of health of the battery. In this paper, the equivalent circuit model of Li-ion battery is established by analyzing the ac impedance spectrum of Li-ion battery and the internal reaction process of the battery, and corresponding each component of the model and the composite link with the curve segments in the impedance spectrum, and the model parameters are identified offline with the help of The Nyquist diagram of impedance spectrum. It is found that the charge transfer internal resistance and low-frequency impedance amplitude in the equivalent circuit increase monotonically with the aging of the battery. Then, the temperature compensation coefficient was introduced to solve the influence of temperature factors on the SOH estimation of lithium batteries, And the root mean squared error of SOH estimation before and after temperature compensation was reduced from 31.7 to 3.0, and the sum of squares for Error was reduced from 13054.7 to 19.8, which significantly improved the SOH estimation of batteries with different temperatures; and a model based on BP neural network was constructed, And the correlation coefficient of model prediction was 0.994. Solving the influence of production process differences on SOH estimation of lithium batteries, and the advantages and disadvantages of the two estimation methods are summarized.