Optimization of LSTM-Based Icing Prediction Model for Transmission Lines

View at publisher Summarise with VIRAT
Authors: Hongfei Wang, Zepeng Sun, Yuzhi Shi, Kunpeng Ji, J T Li, Bin Zhao
Cited by: 0
References: 8

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper addresses the problem of transmission line icing prediction. Based on the Long Short-Term Memory (LSTM) model, it investigates transmission line icing prediction methods and optimizes the prediction model. The results show that the model exhibits excellent performance in short-term predictions (2-hour and 4-hour), with the Mean Absolute Percentage Error (MAPE) below 4% and the coefficient of determination (R2) exceeding 0.89, indicating a high degree of fit between the predicted and actual values. To address the issue of gradient dissipation in long-cycle predictions, the study finds that using a dataset with a 1-hour time interval can balance prediction accuracy and model generalization. This approach maintains controlled Root Mean Square Error (RMSE) while significantly improving the R2 metric, effectively solving the training failure problem caused by excessively large time steps. This achievement provides a high-precision and robust technical approach for transmission line icing prediction.

Fields of Study & Subjects

Icing and De-icing Technologies Thermal Analysis in Power Transmission Energy Load and Power Forecasting

Publication details

Journal / source
International Conference on Electrical Automation and Artificial Intelligence (ICEAAI)
Pages
54–60
Publication date
16 January 2026
Language
EN