An IHO-XGBoost Hybrid Intelligence Framework for High-Precision Prediction of Transmission Line Ice-Shedding
Conductor ice shedding can cause transient conductor jumps and reduce electrical clearances under combined icing and wind loading. This study presents an IHO-XGBoost framework for predicting maximum jump height within a specified transient window. A finite element model (FEM) of a 500 kV tower-line system defines a 1530-case database covering structural, icing, wind, and shedding conditions. The improved Hippopotamus Optimization (IHO) framework combines Logistic–Tent initialization, sine–cosine search with adaptive weighting, and Levy-flight exploration. Prediction is evaluated on 296 held-out cases, with MAE, RMSE, and R2 computed within each run and then averaged across 10 stochastic runs. The mean IHO-XGBoost test metrics are MAE = 0.420 m, RMSE = 0.560 m, and R2 = 0.995. Optimizer comparisons use 3000 actual objective evaluations per run, including initialization and additional candidate evaluations. XGBoost split-frequency importance identifies ice thickness and horizontal stress as the most frequently used physical inputs, accounting for 35.21% and 22.21% of the displayed split counts. The conclusions concern the investigated tower configuration, parameter ranges, and the 20 s dataset-generation window.
Authors
- Chen Zeng (ORCID: https://orcid.org/0000-0002-6950-5475)
- 杨展豪
- Xinglong Feng
- Baiyuan Wang
- Kai Liu
- Tian Wu
Institutions
- China Three Gorges University (CN)
- State Grid Corporation of China (China) (CN)
- Tonghua Normal University (CN)
- China Electric Power Research Institute (CN)
- State Grid Hubei Electric Power Co., Ltd. (China) (CN)
Publication Details
- Journal
- Energies
- Published
- 2026-11-16
- DOI
- https://doi.org/10.3390/en19225525
- Primary Topic
- Icing and De-icing Technologies
- Type
- article
- Field-Weighted Citation Impact
- 0.00