Accurate estimation of the lithium batteries′ state of health (SoH) and state of charge (SoC) is an important guarantee for the safe operation of new energy vehicles. Aiming at the low accuracy and poor robustness problems of joint SoH-SoC estimation, a joint SoH-SoC estimation method based on BP neural network with variable learning rate and adaptive fading extended H∞ filter is proposed. Firstly, a novel SoH feature parameter based on time interval of unit charging voltage difference is proposed. Secondly, the traditional BP neural network is improved by using a novel BP neural network with variable learning rate to improve the error convergence speed and shorten the weights optimization search time. Finally, by designing a new type of adaptive fading factor to weight the error covariance matrix of traditional extended H infinity filter, an adaptive fading extended H infinity filter algorithm is established to reduce the influence of stale measurement on the estimation results and correspondingly improve the estimation accuracy and robustness of filter. The experimental results show that the SoH and SoC estimation errors of the proposed algorithm are smaller than 0. 35% and 0. 5% , respectively, demonstrating the high estimation accuracy and robustness.