Presently mining the rich hidden information contained in power big data is one of the important research directions. Aiming at the management problems such as leakage, out-ofcontrol, and out-of-control of key populations in urban governance,it is an important research direction to apply the concept of power big data fusion-driven and mining the rich hidden information contained in power big data.Firstly, the analysis and application framework based on multidimensional power big data is constructed according to the needs of municipal governance；Secondly, an anomaly analysis model based on“electricity behavior-in door behavior”is innovatively proposed, and form a dynamically weighted cosine similarity analysis model through the entropy weight method to dynamically distribute the power feature vector, so as to form a judgment threshold based on the similarity law of historical samples to realize the situation awareness and abnormal judgment of in-door electricity behavior of key populations. Thirdly, the engineering application architecture of“platform flow- technology flow- business flow”is designed for multi-source systems and multi-service application subjects. Finally, it introduces the application effect of the public security digital city project in Jiangbei, Nanjing, and verifies the important role of power big data deep mining in promoting the strengthening of the police through science and technology.