Risk Assessment of Power Systems to Extreme Events using Polynomial-Chaos-based Methods

点击翻译按钮获取中文摘要

Modern power systems involve numerous uncertainties induced by random load variations, renewable energy variations, and random outages of generating units, lines, and transformers. As a result, they have been subject to an increasing risk of cascading failures leading to large-scale blackouts. The computational burden current methods is prohibitively large for large-scale systems, restricting their practicability when assessing the risk of cascading failures under topology and power injection uncertainties. To address these challenges, this proposal will resort to polynomial-chaos methods and uncertainty quantification theory to develop a new risk assessment framework for power systems subject to extreme events. The developed framework will enhance power system planning, operations, online monitoring and local and global controls. Besides, the uncertainty quantification methods can be generalized to provide great improvement in the control and design procedures of many other engineering fields, such as aircraft design and control, vehicle and train control, among others. The developed framework will be validated on two real power systems, namely the Southern Brazil power system and the Dominion Virginia Power 500-KV transmission system. The project also contains an integrated educational agenda for K-12 students, undergraduates and graduate students who are interested in the STEM (Science Technology Engineering and Mathematics) area.Currently, Monte-Carlo-based methods are widely used in power system planning and operation. The computational burden of these methods for the problem of interest in this proposal is very high. In this project, a new risk assessment framework for power systems subject to extreme events and large uncertainties will be developed based on polynomial-chaos methods and uncertainty quantification theory. Specifically, our proposal will aim to 1) designing a new polynomial chaos-based algorithm that can handle arbitrary probability distributions assumed for the random inputs while providing accurate results for both short-time and long-time simulations, and 2) providing fast calculation speed when the developed algorithm is applied to a highly nonlinear system subject to high-dimensional uncertain inputs. This is achieved by initiating a hybrid approach to conduct dimension reduction for very high-dimension data, which consists of two steps. In the first step, a kernel-based principle component analysis (PCA) is applied to obtain intermediate dimension reduction results. In the second step, we use an analysis of variance (ANOVA) or sliced-inverse-regression to obtain a more active central subspace. Furthermore, scenario-based multi-element polynomial chaos expansion to handle system topology changes will be initiated.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.