Nowadays, with the rapid development of social economy, the scale of power grid is increasing, the capacity of power grid is increasing, and the problem of power loss is becoming more and more serious. In order to make rational and efficient use of power, the research on abnormal power consumption is very important. Aiming at the problem of abnormal power consumption, this paper preprocesses the power consumption data of a regional power grid logically, and uses heuristic algorithm to build two targeted heuristic models. The data collected from the detection instruments deployed in the power grid are used for experimental verification. The relative anomaly index in the case of abnormal power consumption is 6~10 times larger than that in the normal case. Combined with the heuristic algorithm, it can better identify and detect the abnormal power consumption phenomenon of the power grid, and verify the effectiveness of the proposed method.