Research on Graph Clustering Based Line Loss Prediction in Low-Voltage Distribution Network

With many distributed power sources connected to the power system, the structure of the distribution network is intricate and complex, and the operation of the network is more variable, which also puts forward higher requirements for the analysis and management of the line loss of the distribution network. In order to effectively reduce the line loss of the distribution system and improve the economical operation of the grid, this paper proposes a neural network based on the fusion graph clustering algorithm for calculating the line loss of low-voltage distribution networks. By analyzing the mechanism of power lines, grid structure, distributed power supply access capacity and other factors affecting line loss, a system of electrical characteristics indexes reflecting the distribution network under the new power system is established. And according to the low-voltage distribution network characteristic indexes, the graph algorithm is used to classify the samples, and the BP neural network model is used to train the clustered samples by class. Finally, the relationship between the sample line loss rate and the characteristic index is obtained by fitting the BP neural network, and the effectiveness of the model is verified by simulating and calculating with the actual measured data of low-voltage distribution network in a certain area.