In the field of lithium battery research, there is a need for a fast and accurate estimate of their state of health. In order to solve the existing problems of data processing and model parameter adjustment, this paper proposes a method to optimize the BP neural network based on partial charging data and use dung beetle optimization algorithm to estimate the health state of lithium batteries. The study used the University of Maryland’s lithium battery public dataset for testing and verification, and only needed to use the data of the first half hour of charging to estimate the health status of lithium batteries. The results show that the model constructed by this method has high accuracy in estimating the state of health of lithium batteries, the coefficient of determination of the model is as high as 0.977896, the mean square error is as low as 0.000763, which is 46.31% lower than that of the initial model, and the detection time is also shortened from three hours to half an hour. Compared with the traditional method, this method greatly reduces the requirements for lithium battery charge and discharge data, and improves the accuracy and speed of lithium battery health detection.