Prediction of Transmission Line Icing Thickness based on Long Short-Term Memory Network

The frequent occurrence of extreme weather has led to the increasingly serious problem of ice covering of transmission lines, one of the natural disasters affected by the power grid. It has brought great challenges to the safety, reliability and stable operation of the power grid. The prediction of the ice thickness of transmission lines can effectively improve the accuracy and timeliness of grid ice prevention and disaster reduction. It has become an important research content of power grid reliability. Firstly, the transmission line icing related data are collected from the aspects of power grid equipment information and environmental meteorological information. The data are preprocessed by default value, characteristic correlation analysis, normalization, etc. A database stored in a unified format is established. Then, the long short-term memory network (LSTM) transmission line icing thickness prediction model and recurrent neural network (RNN) transmission line icing thickness prediction model are established. Finally, through simulation analysis, the mathematical and physical relationship between the ice thickness and the environmental temperature and instrument wind direction with the strongest correlation is obtained. The comparison chart of the real value and the predicted value of the ice cover of the two models is compared. The evaluation indicators are comprehensively analyzed, which verifies the accuracy and timeliness of the LSTM transmission line ice cover thickness prediction model. It can provide a theoretical basis for the prediction of transmission line ice cover thickness under low temperature rain, snow and freezing disasters.