Prediction of Distribution Network Line Loss Based on Grey Relation Analysis and XGboost

Line loss plays a vital role in evaluating the economic operation of the power system. However, in the 10KV distribution network, due to the backward configuration of measurement and testing equipment, it is difficult to collect data and nodes with too much operating information, resulting in line loss. Forecasting becomes more and more cumbersome and difficult. Therefore, a method of line loss prediction for distribution network based on Grey Relation Analysis and XGboost is proposed in this paper. Firstly, determine the relationship between line loss and electrical indicators through Grey Relation Analysis, determine the weight of each feature indicator by using the entropy weight method to achieve dimensionality reduction of feature data, and select the best feature as the input of the prediction model. The XGboost algorithm is used to increase the penalty function regular term to optimize the loss function and establish the prediction model. Finally, 350 10kV lines are predicted, and the theoretical line loss calculation value and the BP neural network algorithm prediction value are compared to verify the practicability of the proposed method. Rationality and effectiveness.