China’s economy has entered a new normal, supply side reform and the battle for environmental protection have continued to advance, international trade unilateralism has risen,traditional industries are facing many challenges, and the risk of electricity tariff recovery for enterprise customers is increasing.Therefore, predicting the overdue risk of power consumption of enterprises and minimizing the impact on the company's operations are becoming more and more urgent. At present, overdue risk management and control methods can only rely on the historical arrears of enterprises to conduct risk judgments, and lack of deep exploration of user power consumption trends and payment historyleads to the lack of accuracy and timeliness of enterprise overdue risk judgment. To solve the above problems, from the single perspective of enterprise electricity data, based on DBSCAN clustering algorithm,constructing the typical electricity consumption mode of the industry, and adopting the idea of k nearest neighbor, defining the pattern distance measurement function, introducing the concept of risk level, and proposing a power pattern based electricity overdue risk prediction model for enterprises to realize the self evolving electrical deviation based on overdue risk level with degree 4 and risk prediction. Theoretical analysis and experimental results verify the effectiveness of the proposed method.