1. Research on Short-term Prediction Method of Substation Bus-bar’s Voltage Trend Based on Multidimensional Time Series Data Mining
- Author
-
Ye Kang, Fei Xiao, Li Xiongli, Xiwu Leng, Zhu Licheng, and Hu Youlin
- Subjects
Computer science ,Busbar ,020209 energy ,02 engineering and technology ,computer.software_genre ,Grid ,Term (time) ,Filter (video) ,0202 electrical engineering, electronic engineering, information engineering ,020201 artificial intelligence & image processing ,Data mining ,State (computer science) ,Time series ,Cluster analysis ,computer ,Voltage - Abstract
The short-term prediction of voltage trend is the important technical foundation of the voltage exceeding intelligent alarm system in power substation. According to correlation between the adjacent lines in the same substation, the time series data of different bus-bars in same section, was constructed into a multi-dimensional time series data matrix model for prediction of voltage trend. Moreover, based on above matrix model, a novel prediction method based on multidimensional time series data was proposed, which transformed the multidimensional time series data matrix into a classical two-dimensional decision information system in the first stage via preprocessing and clustering. In the second stage, various classical machine learning algorithms are ensembled to forecast the short-term future trend of the specific bus-bars' voltages. The efficiency of the prediction method based on multi-dimensional time series data mining was validated by the implement in a 500 KV bus-bar's prediction in a Substation of the State Grid Shanghai Company. The results represented that the prediction method in this paper had sound precision in practice and can improve the functionality of voltage exceeding intelligent alarm system via assistance to filter plenty of fake alarms.
- Published
- 2018