The energysaving issue attracts more and more attentions recently with the increasing installations of the elevators in the buildings. And super capacitors (SUPCAPs) are the preferred energyconserving devices equipped in the elevators to store the energy regenerated by the tracking motor. A new back propagation neural network based (BPNNbased) predictive control strategy is employed to predict the energy required by the elevator at every trip. According to the information of the current stop floor, the destination floor, and the weight of the passengers, the BPNN model is setup and trained by the samples. Then the trained BPNN is applied to predict the energy required by the elevator at the beginning of each trip. In this way, the energy provided by the SUPCAPs is determined and the balance voltage (BV) can be regulated to fully compensate the peak power. Moreover, the power capacity of the gridside pulsewidth modulation (PWM) rectifier can be reduced because the SUPCAPs provide as much energy as possible in every trip. Not only the peak power which emerges when the elevator moves in heavy load or full load eliminated, but also the rated power can be partly counteracted by the SUPCAPs. Finally, a simulation based on MATLAB/Simulink as well as the corresponding experimental prototype is setup to verify the proposed method, and results from the simulation and experiments prove the effectiveness of the proposed method.