Data-driven failure diagnosis in transmission protection system with multiple events and data anomalies

• Main Authors: Amir Gholami, Anurag K. Srivastava, Shikhar Pandey
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: Journal of Modern Power Systems and Clean Energy
• Subjects: Failure diagnosis; Transmission protection system; Protection mis-operation; Phasor measurement unit (PMU) data anomaly and cleaning; Ensemble method
• Online Access: https://ieeexplore.ieee.org/document/9028816/

To guarantee the reliable power supply, the expected operation of all the components in the power system is critical. Distance protection system is primarily responsible of isolating the faulty section from the healthy part of the grid. Failure in protection devices can result in multiple conflicting alarms at the power grid operation center and complex events analysis to manually find the root cause of the observed system state. If not handled in time, it may lead to the propagation of the faults/failures to the adjacent transmission lines and components. With availability of the synchronized measurements from phasor measurement units (PMUs), real-time system monitoring and automated failure diagnosis is feasible. With multiple adverse events and possible data anomalies, the complexity of the problem will be escalated. In this paper, a PMU based algorithm is presented and discussed to detect the root cause of the failure in transmission protection system based on the observed state, e.g. multiple line tripping, breaker failures. The failure diagnosis algorithm is further enhanced to come up with the fully functional version of the failure diagnosis tool, which is tailored for the cases in which the PMU anomalies are present. In the developed algorithm the validity of the PMU data is critical; however, such causes as communication errors or cyber-attacks might lead to the PMU data anomalies. This issue is well-addressed in this paper and some major types of anomaly detection methods suitable for PMU data are discussed. Results show that the ensemble approach has some distinct advantages in data anomaly detection compared to the previously used standalone algorithms. Additionally, the enhanced failure diagnosis method is developed to clean the inaccurate data in case of the anomaly in measured voltage magnitudes. Finally, both original and enhanced versions of the tool are tested on 96-bus test system using the real-time OPAL-RT simulator. The results show the accuracy of the enhanced tool and its advantages over the primary version of the tool.