Development of Statistical Fault Detection Algorithms for Modern Power Grid Networks

点击翻译按钮获取中文摘要

The national power grid has been undergoing transformational changes and facing new challenges over the last decade. The grid no longer consists of power suppliers, like coal, hydroelectric and nuclear plants, which supply power at a roughly constant rate. Renewable energy sources deliver power in a much less predictable manner due to their dependence on weather factors. Many retail and service facilities are no longer pure energy consumers but have become energy producers by installing large solar farms. In this new environment, it is important to develop new tools for timely detection of faults. Due to the less predictable power delivery and consumption, and possible attacks, fault detection must be based on the understanding of the random structure of the grid operation whose characteristics can be measured by a new generation of devices. These devices generate massive data sets, which have not yet been widely utilized. The PIs will address many specific questions related to the detection and identification of power grid faults. The PIs will use data obtained from modern measurements devices and develop state of the art statistical methodology. Advances made in this research will contribute to the energy security of the United States. The project will provide training to graduate and undergraduate students. The PIs will develop advanced statistical theory and large-scale computational tools that will form a foundation for engineering implementations aimed at detecting and identifying power grid faults. This will be done within a framework of change point detection methodology. Specifically, the PIs will 1) develop algorithms for screening sensor inputs that impact specific protection devices; 2) create algorithms for fault monitoring, with controlled significance level and expected time to detection; 3) develop theoretical and practical foundations for a posteriori change point testing in smart grid networks; 4) work out algorithms for the identification of subgrids affected by faults and the classification of faults; and 5) enhance PSCAD/EMTDC simulation tools by incorporating characteristics of high resolution sensor readings. It will be a collaborative effort combining the expertise of two statisticians working, respectively, in stochastic networks and large-scale computing (Wang) and time series analysis (Kokoszka), and a network engineer specializing in smart grids and signal processing (Yang).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.