In order to deal with three major problems of electric vehicle (EV): the short driving range, the short life of batteries, and the poor ability of start-up, a hybrid power system was designed and applied to the EV. It was composed of an ultracapacitor with high-specific power and long life, four lead-acid batteries, and a bi-directional DC/DC converter. To improve the stability and reliability of the hybrid-power EV, based on establishing the mathematical models of driving and regenerative-braking processes, a novel neural network sliding mode controller (NNSMC) was researched and designed for the EV. The controller comprises a back propagation neural network (BPNN), a radial basis function neural network (RBFNN), and a sliding mode controller (SMC). The BPNN is used to adaptively adjust the switching gain of the SMC on-line so as to avoid the whippings. The RBFNN is used to perform system identification and parameter prediction. The experimental results show that the NNSMC is superior to PID controller at response speed, steady-state tracking error and resisting perturbation whenever driving or braking. Additionally, the hybrid-power EV with NNSMC can improve the ability of start-up, recover more energy, lengthen the life of batteries, and increase the driving range than the EV using batteries as its single power source by about 40%, and than the hybrid-power EV with PID controller by about 4%.