Wide-Area Time Series Data Mining Based Transient Voltage Stability Assessment

ZHU Lipeng,LU Chao,HUANG He,SU Yinsheng,WANG Jifeng

DOI: https://doi.org/10.13335/j.1000-3673.pst.2016.01.024

2016-01-01

Abstract:In context of Energy Internet,big data analytics provides possibility of tackling an array of traditional troublesome problems in power systems.In terms of big data-based transient voltage stability assessment,precise capture of critical evolution trends or states is a great challenge.To cope with it,a time series shapelet methodology,reliably extracting sequence features from dynamic post-contingency PMU measurements,is introduced.Additionally,by setting distinct weights for stable/unstable cases,misclassifying costs of classification learning are incorporated into decision tree algorithm,so that the assessment model would avoid mistaking unstable cases as far as possible.Test results on Nordic system for whole assessment scheme demonstrate that the assessment model not only holds a high performance of classification,but also offers considerable insights into the results,providing guidelines for further comprehension of specific system instability patterns and online monitoring.