Voltage Sag State Estimation based on Hybrid Neural Network using One Dimensional CNN and GRU

Voltage Sag State Estimation (VSSE) based on voltage information collected by monitoring devices at limited nodes in the distribution network can identify weak links in the power grid, evaluate economic costs, verify corresponding responsible parties, and take corresponding solutions for voltage sag suppression. Therefore, in this paper, a combined deep learning method using one-dimensional Convolutional Neural Networks and Gated Recurrent Unit is adopted for VSSE solution. Here, the high-dimensional features in the voltage sag monitoring information are extracted via one-dimensional Convolutional Neural Networks, and then the temporal features is further learned through Gated Recurrent Unit. Finally, the voltage sag state, that is, the voltage sag amplitude range of all other nodes without monitoring devices installed, is obtained. Based on this method, it is not necessary for physical model or any other prior conditions, such as system parameters, operating mode, and so on. The effectiveness of the proposed method is also verified via IEEE 33 system, and the VSSE accuracy can reach 98.32 % under the premise of three monitoring points.