An Early Warning Method for Transmission Line Ice Accretion and Galloping Based on LSTM Neural Network

To ensure the safe and stable operation of power transmission lines under ice accretion and galloping conditions, this paper proposes an early warning method based on Long Short-Term Memory (LSTM) neural network. The method formulates galloping warning as a supervised time-series prediction task. It incorporates meteorological features-such as wind speed, wind direction, conductor axis angle, relative humidity, and temperature-as input to train the LSTM model using historical galloping data. Case studies using real-world data from a province in China demonstrate that the model effectively identifies meteorological conditions conducive to galloping and provides timely risk warnings for transmission lines. Results show that the proposed method achieves high warning accuracy rate with low false warning rate and missed warning rate, offering strong support for grid dispatching and de-icing strategy planning.