Pseudo-metric modelling of distribution network state estimation based on CNN-BiLSTM network and customized HGGA algorithm

Abstract

To address the insufficient configuration of real-time measurement devices in distribution networks, a state estimation method for distribution networks is proposed. First, both historical data and real-time data are used as the input of the convolutional neural networks-bi-directional long short-term memory (CNN-BiLSTM) network to obtain the pseudo measurements of the branch power and load node injected power with high accuracy at the current moment. Second, the pseudo measurement set of the input state estimation is determined using the customized hybrid greedy genetic algorithm (HGGA) algorithm. Finally, the real-time measurement, pseudo measurement, virtual measurement, and weight set are input into the state estimation program to complete the real-time state estimation of the distribution network based on the weighted least squares method. Additionally, the proposed method is applied to the data fusion of different measurement systems. Theoretical analysis and example verification confirm that the proposed method can effectively improve the accuracy of pseudo measurement modelling and state estimation results.