Research on ice thickness prediction of power transmission lines based on SO-CNN-LSTM algorithm

With the frequent occurrence of extreme weather, ice accumulation on power transmission lines has become a common phenomenon in winter, causing significant economic losses to power systems and attracting increasing attention. However, due to the complexity of ice formation on conductors, establishing a high-precision ice thickness prediction model is crucial for ensuring the safe and stable operation of the power grid. Therefore, this paper proposes a hybrid model that combines a snake optimization (SO) algorithm, convolutional neural network (CNN), and long short-term memory (LSTM) for ice thickness prediction. Firstly, temperature, humidity, wind speed, and rainfall are selected as feature parameters to establish an ice dataset by analyzing the factors contributing to ice formation using the Pearson correlation coefficient method, which then serves as the input for the power transmission line ice prediction model. Subsequently, CNN is utilized to learn data features, and the results are fed into LSTM where the SO algorithm optimizes the key parameters of the LSTM. Finally, experiments results show that the proposed model reduced the RMSE error by 9.25%, 13.39%, and 19.52% and decreased the MAE by 7.78%, 12.67%, and 17.45% compared to the CNN - LSTM, BiLSTM, and LSTM models, demonstrating good performance and significantly improving the accuracy of ice thickness predictions.