| Electric vehicles (EVs) are a kind of environmentally friendly cars powered by batteries. The performance of batteries plays an important role on the whole performance of EV, and influences directly on EV's performance, i.e., the range and accelerated ability of EV and EV's climbing capacity. Therefore a battery management system (BMS) is demanded for electric vehicles.The paper firstly gave a brief introduction the concept of EVs, the current state of BMS, as well as some conventional SOC anticipation methods. This paper took Ni-MH battery as a case study. The electrochemistry characteristic of Ni-MH battery, as well as the relation between the residual capacity of battery and the terminal voltage, current and temperature, respectively, were analyzed in-depth. Furthermore a new SOC prediction method, which combine Radial Basis Neural Network and Kalman filter methods, was proposed. This paper studied on the BMS for an in-house developed and diamond-liked new concept car. The new prediction method was applied to a newly developed BMS based on Controller Area Network (CAN).The simulation data of SOC based on this newly proposed method were identical with the experimental data.The newly developed BMS can safely separate the measuring (low voltage) from the operating circuit (high voltage) of battery. It can not only can predict the EV's driving range, but also can effectively monitor the state of battery. It will give alarming signals when abnormal occurs and remind a driver to replace the deteriorated battery. Each Electric Control Unit (ECU) of BMS uses high precision sensor to collect data. Additionally, a microcontroller with CAN port was taken as the CPU in the new battery management system. The ECU served as an intelligent data collecting node. Each node was connected with a CAN bus to share the data with each other. This kind of linkage is not only useful for reducing the wiring harness, but also can provide flexible layout of batteries and furthermore improvement security and reliability. |