Prediction of Power Grid Line Faults Under Cold Wave Weather Based on Hybrid Model of XGBoost and LSTM

In order to fully explore the temporal feature relationship between cold wave weather and power grid transmission line faults and improve the fault prediction accuracy, a power gird line fault prediction method based on the hybrid model of Extreme gradient boosting (XGBoost) and long-short term memory (LSTM) network is proposed in this paper for cold wave weather. Firstly, data enhancement is carried out on the unbalanced samples of grid transmission line faults under the cold wave weather, and the training sample set of line faults is constructed. Then, a single model for multi-feature prediction of power gird line faults is built based on XGBoost, and a single model for time series prediction of power gird line faults is built based on LSTM. Finally, the established power gird line fault prediction model was united based on the linear model, by training the linear model parameters to construct a hybrid XGBoost-LSTM prediction model for power gird line faults. Through the operation of power grid equipment and data validation in a region, the proposed transmission line fault prediction model has high prediction accuracy and generalization ability under cold wave weather compared with other algorithms.