Aiming at the low accuracy and poor scalability problem of SOC estimation for lithium battery management system presently, designed the lithium battery management system based on Blackfin digital signal. This system implements the function of real-time monitoring of lithium battery, remaining power estimation, extending groups of lithium battery by the CAN bus communication, the danger alarm and automatic protection of the lithium battery, etc. With respect to the remaining power estimation algorithm, this paper proposes a BP neural network method which is optimized by genetic algorithm and the ant algorithm (abbreviated to GAAA algorithm). Experiment result shows that the algorithm has higher SOC accuracy estimated precision and faster operating speed than the BP neural network based on the genetic algorithm.