Collaborative Research: AMPS Stochastic Algorithms for Early Detection and Risk Prediction of Hidden Contingencies in Modern Power Systems

Wayne State University, Detroit MI

Investigators

Abstract

Modern power systems (MPS) are complex systems involving conventional and renewable generators, smart distribution networks, and advanced information exchanges. High penetration of random low-inertia renewable energy sources, increased natural disasters such as the 2021 Winter storm Uri, and unprecedented man-made cyber-physical attacks have posed a threat to the reliability and security of MPS. Several cascading failures in MPS started with smaller undetected contingencies such as California's wildfires (e.g., Camp Creek Fire, Zogg Fire, and Dixie Fire) caused by equipment failures. Smaller contingency events, particularly on the distribution side of the grid, may not be directly detected. This project focuses on the early detection and risk prediction of hidden contingencies in MPS. The research fits within efforts to enhance the resilience of the U.S. power grid and move toward carbon-free energy infrastructure. Therefore, it has broader impacts on the carbon-free economy and social welfare. This project will also enhance teaching, training, and learning in mathematics and statistics, renewable energy, smart grids, and green technologies. The team plans to develop new courses for undergraduate and graduate students to facilitate the training of next-generation scientists and engineers. Every effort will be made to promote the participation of underrepresented students in the research project. This research project introduces a novel framework of stochastic prediction, estimation, and early detection (SPEED) for MPS. Covering a broad range of cyber-physical contingencies (CPC), this research will have the following distinct and novel aims and outcomes. First, the project introduces a new stochastic hybrid system (SHS) model, consisting of continuous dynamics and discrete events. Second, the project will develop new estimation and prediction computational methods. Starting from the Wonham filter for hidden Markov chains, to detect discrete jump changes, this research will focus on finding more computationally feasible schemes. Furthermore, rates of convergence of the algorithms will be obtained, and extensive numerical experiments will be performed. Third, fundamental concepts such as joint observability will be introduced. New estimation algorithms will be developed for joint estimation and prediction of CPC in SHS. Fourth, since early and quick detection of abrupt changes is vitally important for the risk management of MPS, this project will provide a new computable scheme based on Markov chain approximation for optimal stopping and will quantitatively predict risks of potential near-future cascading CPC. Fifth, evaluation and validation of the theoretical findings will be conducted through utility-level operational data, large-scale power grid simulations, and hardware-in-the-loop emulation on a microgrid. The synthetic operational and summary data of the distribution power grids and transmission systems will be incorporated into the validation and evaluation of the study. 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.