State Estimation of Distribution Network with the Improved Deep Residual Neural Network

With the continuous advancement of the sustainable goal, a large number of renewable energy and non-traditional loads, such as electric vehicles and etc., bring many challenges to the power stability and reliability of distribution network. Distribution network state estimation can make the staff receive the state of the power grid in real time and provide the data basis for the subsequent optimization scheduling. It is of great significance to carry out research on distribution network state estimation. Therefore, aiming at the characteristics of poor observability and large fluctuation of distribution network, this paper puts forward a state estimation method for distribution network based on improved residual network (ResNet). In the offline training stage, the neural network is trained by collecting historical data, and the mapping between input and output is constructed. In the online application stage, the real-time measurement data is injected into the model obtained from offline training to complete the real-time state estimation of the distribution network. Simulation results on IEEE 33 standard system show that the proposed method, which expected to play a role in real-time state estimation of distribution network next, has better estimation accuracy and robustness than other methods.