Computer Simulation of Cascading Disturbances in Electric Power Systems

Executive Summary Hidden failures in protection systems for electric power networks can play significant roles in propagating small disturbances into wide-area disturbances. Hidden failures are incorrect operations of protection device that usually remain undetected until abnormal operating conditions occur. The objective of this research project was to investigate and demonstrate a new approach to numerical assessment of the vulnerability of a power system to hidden failures of individual relays. By identifying the most vulnerable locations in a power system, the approach identifies the protection system relays that could be upgraded to provide the greatest improvement in reliability within a constrained capital investment budget. Using parallel processing, the research advances practical use of computing resources in applying numerical techniques to protection system assessment. We examine these techniques in a case study of the New York Power Pool's (NYPP) 3000-bus system. Through computer simulations, we analyze the impact of consecutive relaying malfunctions, and define the protection system vulnerability and reliability to numerically characterize this impact. Protection system reliability and vulnerability can be reduced if upgraded relays with lower hidden failure probabilities are put into service. By sorting all the relays according to their vulnerabilities, we can locate the most vulnerable regions in the protection system. A heuristic random search algorithm is developed for fast, rare-event simulation of cascading outages. An optimal strategy for upgrading relays is proposed for the economical enhancement of protection system reliability under a limited capital budget. Using a 256-Processor Intel cluster at Cornell Theory Center, we simulated 41,053 NYPP blackouts that have load losses greater than 10 MW. From the simulation results, the vulnerability of each relay and the global protection system reliability are computed. The twenty-five most vulnerable relays in NYPP are identified. By solving the economic optimization problem, we determine the ten relays whose replacement can best improve the global reliability. Lack of computational resources and of efficient algorithms have been major obstacles in studying large blackouts. For large networks, the number of different disturbance paths would be quite large. It is difficult to simulate consecutive relay failures in large-scale power systems due to their inherently small failure probabilities and to their load-flow dependent nature. The heuristic random search algorithm presented in this report only simulates each important blackout once. It computes the probability afterwards based on the underlying hidden failure models. In addition, it is impossible to simulate all the paths on one single computer. …