Prediction of Equivalent Icing Thickness of Transmission Lines in Distribution Network

Aiming at the problem that the time series forecasting method will produce a certain degree of lag due to its series autocorrelation, the multivariate equivalent ice thickness prediction using meteorological data as the feature is used. Aiming at the problem that the prediction accuracy of ice thickness is not high, the attention mechanism is used to assign adaptive weights to multi-dimensional feature parameters and time steps to highlight important information in different feature dimensions and time steps. At the same time, due to the change of ice cover with time, a Bi-directional Long Short-term Memory (Bi-LSTM) is used to establish an ice cover prediction model to make use of the past and future information provided at each time point in the process, and for the parameter selection problem of Bi-LSTM, the sparrow search algorithm (SSA) is used to select the hyperparameters to reduce the error caused by human intervention, and then improve the accuracy of line equivalent ice thickness prediction, and finally verify the validity of the model through experimental results.