A framework for dynamic prediction of reliability weaknesses in power transmission systems based on imbalanced data

• Published in: International journal of electrical power & energy systems Vol. 117; p. 105718
• Main Authors: Sun, Chenhao, Wang, Xin, Zheng, Yihui, Zhang, Feng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.05.2020
• Subjects: Dynamic prediction; Imbalanced data; Power transmission fault event; Reliability weakness; Spatiotemporal distribution; Imbalanced data; Power transmission fault event; Reliability weakness; Dynamic prediction; Spatiotemporal distribution
• ISSN: 0142-0615; 1879-3517
• DOI: 10.1016/j.ijepes.2019.105718

•The spatiotemporal distribution of power transmission fault events is predicted.•The periods with fewer fault events and rarely occurred elements are incorporated.•Twofold relative weight is formed to probe the impact of elements and period.•An automatic self-adaption process is designed to dynamically modify parameters.•This framework is validated based on an empirical case. Power transmission systems are principal for energy supplies, and their reliability is remarkably threatened by fault events. The spatiotemporal distribution of such reliability weaknesses can provide crucial information for maintenance arrangement and operational scheduling in systems. It is also salutary for system operators to get sufficient preparation time. With such motivations, this paper presents original insights on the prediction of the spatiotemporal distribution of power transmission fault events. A framework based on the dynamic association rule mining with rare environmental elements and time series model is proposed. In this model, the rarely occurred environmental elements and fault causes, as well as the periods with fewer fault events, are incorporated and assessed explicitly. The twofold relative weights are developed to measure the influence of the different elements within the dissimilar periods on the reliability of the whole system. To further improve the prediction performance, an automatic self-adaption process is established to dynamically calibrate the current criteria and parameters in light of the consequences from the previous period. Finally, this framework is applied and testified via a practical instance, and the results of this empirical case demonstrate the flexibility and robustness of it during real applications.