Abstract: With the rapid development of modern power system, the structure and operation mode of power system become more and more complex, and the real-time and accuracy of state estimation are also put forward. Therefore, this paper proposes a fast state estimation for power system based on deep neural network. This method selects the input measurement set of the state estimation model through correlation analysis, and then establishes the state estimation model based on deep neural network by using massive historical data. When the real-time measurement of the power system is updated, the strong correlation measurement is input into the established state estimation model to obtain the state estimation result quickly. The simulation results of an IEEE standard system and a practical provincial power grid shows that the estimation accuracy and robustness of this method are better than traditional Weighted Least Square (WLS) estimation and Weighted Least Absolute Value (WLAV) estimation. Moreover, system scale less affects the online computing time of this method. In the simulation of a practical provincial power grid, the calculation efficiency of this method is 1.43 and 27.2 times higher than that of traditional WLS estimation and WLAV estimation respectively.