The invention provides a lead-acid power battery system fault diagnosis method. The method involves an off-line part and an on-line part. The method includes the specific steps that in the off-line state, data are collected through a simulation model, the data are preprocessed by using a normalization method, a data classification training set and a testing set of a power battery system of a support vector machine are obtained, parameter adaptive optimization is conducted through a GA algorithm, a one-to-one method is used for training to obtain a diagnostic model of the support vector machine, and SVM decision classification is conducted; in the on-line state, a fault generating device is used for simulating fault signals, the signals are collected through a collection module, the data are preprocessed by using the normalization method, the data are further input into an SVM module in off-line training, and fault online classification based on an SVM algorithm is conducted. According to the lead-acid power battery system fault diagnosis method, intelligent off-line and on-line diagnosis of faults of the battery system can be achieved, and meanwhile the fault diagnosis recognition rate is increased.