Multi-source Meteorological Dataset for Icing Forecasting of Partial 220kV Transmission Lines in Zhejiang Province

Second, data cleaning is carried out: abnormal extreme values caused by sensor failures are eliminated using the 3σ principle, short-term missing data due to communication interruptions are filled using linear interpolation, and temperature drift correction is performed on tension and tilt data based on tower design parameters. Finally, based on the wire mechanical equilibrium model, the wire tension and tilt data are converted into equivalent ice coating thickness, and key icing forecasting features such as icing growth rate and icing duration are derived to form structured time-series data.The overall data missing rate is less than 3%.

Missing data is mainly concentrated in the extreme rain and snow periods of remote mountainous lines, and has been filled using linear interpolation. The filled data can be normally used for time-series analysis and modeling. Data errors mainly come from two aspects: First, the inherent measurement errors of micro weather stations, tension sensors and tilt sensors, among which the temperature error is ±0.5℃, the wind speed error is ±0.3m/s, and the tension error is ±1%.

Second, the systematic errors caused by the simplification of the mechanical model in the calculation of equivalent ice coating thickness and the deviation of sensor installation positions. The overall calculation error of ice coating thickness does not exceed ±10%, which can meet the needs of engineering applications and academic research.All data files are in the universal CSV format and can be directly opened and processed using mainstream software such as Microsoft Excel, WPS Office, Python, and MATLAB, without the need to install additional niche tools.