湖北电力2025，Vol.49Issue(5)：25-34,10.DOI:10.3969/j.issn.1006-3986.2025.05.003
基于IPSO算法优化的CNN-LSTM模型及其输电线路覆冰厚度预测方法研究
Study on CNN-LSTM Model Optimized by IPSO Algorithm and Its Method for Predicting Ice Thickness on Transmission Lines
摘要
Abstract
To evaluate in real-time the safety impact of ice thickness on transmission lines under complex meteorological conditions on power systems,this paper proposes an intelligent prediction model that integrates Convolutional Neural Networks(CNN)and Long Short-Term Memory(LSTM)networks.The Improved Particle Swarm Optimization(IPSO)algorithm is employed to automatically optimize the hyperparameters within the hybrid neural network model,constructing an accurate prediction model for ice thickness on transmission lines.The research results show that the IPSO algorithm can effectively enhance the global search ability of particles,the fitness optimization amplitude is increased by 46.4%and 51.9%compared with the PSO algorithm and the DBO algorithm respectively.Compared with other models,the IPSO-CNN-LSTM hybrid model shows the best prediction performance on the test set,with ERMSE and EMAPE values of 0.408 3 mm and 3.132 6%,respectively,indicating a significant improvement in prediction accuracy.The correlation coefficient R² remains at 0.960 8,demonstrating a high correlation between predicted and actual values and strong data interpretability.In engineering applications,the prediction results of the model can provide reliable data and decision support for the prevention of freezing and thawing of power systems.关键词
输电线路/覆冰厚度预测/IPSO/CNN-LSTM神经网络Key words
transmission lines/ice thickness prediction/IPSO/CNN-LSTM neural network分类
信息技术与安全科学引用本文复制引用
聂汉文,顾崧民,易金桥..基于IPSO算法优化的CNN-LSTM模型及其输电线路覆冰厚度预测方法研究[J].湖北电力,2025,49(5):25-34,10.基金项目
湖北省自然科学基金项目(项目编号2024AFD069) （项目编号2024AFD069）
恩施州"赶超计划"科技计划项目(项目编号:D20230007). （项目编号:D20230007）
