The traditional methods of non-intrusive load monitoring have the disadvantages of complicated implementation, difficulty of real-time computing, and repeated training requirement of the dataset threshold. In view of this, a fast event detection algorithm based on hypothesis testing, named compound hypothesis testing algorithm, was proposed in this paper. In the algorithm, the total power signal collected by intelligent electric meter was detected through Chi-square goodness-of-fit test to find out the time point at which events may occur. Then the Z test was used for the two time windows respectively, with the event inspection being conducted for suspects only. Based on the test results, the occurrence of events can be determined accordingly. The calculation was simple and fast. On the BLUED data set, the event detection performance of compound hypothesis testing algorithm and the standard Chi-square goodness-of-fit test method were simulated and compared. The experimental results showed that, the stability and the robustness of the compound hypothesis test algorithm for different base loads were verified. It is concluded that the method proposed in the paper can not only accurately identify the switch events, but also ensure the characteristics of fast and concise operation speed, thereby improving the recognition accuracy of event detection. Therefore, the proposed method has certain application value and reference significance.