With the rapid development of new energy, the scale and complexity of power grids lead to more and more serious incidents of electricity theft, which seriously affects the normal operation and production efficiency of regional power grid. In order to solve this problem, a new identification method of electricity theft driven by data-physical coupling is proposed in this paper. Based on the physical-driven probabilistic power flow calculation, the identification characteristics of electricity theft are constructed, and the data-driven feedforward neural network (FNN) is used to identify the electricity theft behavior. The experimental results show that the identification accuracy of this method can reach 92%, which can effectively identify the electricity theft situation of regional power grid in the new energy scenario.