In order to detect the illegal use of electricity, electrical enterprises generally adopt traditional manual examination ways. However, both of the accuracy and efficiency of the approaches are far from satisfaction. In this paper, an analysis method based on the Extreme Learning Machine (ELM) algorithm is proposed, which is used to predict the behavior of customers' illegal electric use. Firstly, it collects the historical electric usage data and preprocesses the data to make it suitable for analysis by the algorithms. Then, it applys an algorithm based on neural network model, which is called ELM, to build the model to describe the abnormal power utilization behavior of the customers. Finally, experiments on the real electrical consumption data are conducted to evaluate the proposed method. The experimental results demonstrate that the proposed method is accurate and efficient.