Failure Analysis and Diagnosis Scheme in Distribution Systems

Abstract

Continuous and rapid technological advancements have transformed the modern day power system. Increased global inter-connectivity has made reliable power supply a critical requirement. A small outage can cascade into a blackout causing great inconvenience and significant monetary damage. These concerns highlight the need of an additional layer of proactive approach in conventional protection schemes. The focus of such an approach would be to shift from reacting to a failure to anticipating a failure. Anticipating a failure gives time to better prepare and mitigate the failure by efficient allocation of resources in order to limit the negative consequences. Starting from the inception of the event causing the failure to the final occurrence of the failure, the time-period in between is termed as the pre-failure period where the signatures of the incipient failure can be observed. The availability of high-resolution devices has improved monitoring of grid operations during this pre-failure period. Improved monitoring enhances situational awareness leading to easier detection of incipient failure signatures. Research conducted in this field has led to development of few failure anticipation techniques but the application potential of some are restricted to specific equipment or phenomena while that of others are restricted by resource requirements. There is a need of addressing the research gap of a comprehensive failure anticipation technique that fulfills three major criteria of low computational burden, wide applicability in different scenarios and installation compatibility with existing grid monitoring devices for economical implementation. The research conducted in this thesis aims to address this research gap by developing a comprehensive failure anticipation technology titled Failure Anticipation and Diagnosis Scheme (FADS) for AC distribution systems. FADS implementation broadly comprises of three functionalities. The first functionality is concerned with quick and accurate identification of incipient failure signatures. Almost all failure anticipation techniques rely on cross-referencing historical databases or identifying specific patterns in order to detect incipient failure signatures. However, incipient failure signatures seldom manifest in same patterns. Hence, FADS relies on the fundamental aspect that pure AC sinusoids are complex exponentials. Incipient failure signatures would invariably violate certain properties of complex exponentials and manifest as waveform distortions, which would be leveraged by FADS to detect the signatures. The second functionality involves the data processing of the distortion data. Several novel parameters are introduced in the second functionality that helps in processing the data obtained from the first functionality. The use of novel parameters helps in accurate assessment of the stress experienced by the grid operations due to the event causing distortions. Such an assessment help FADS to be robust to false positives or false negatives. Finally, the third functionality involves interpretation of the information obtained through data processing. The interpretation provides metrics to rank the severity of the damage the event can inflict on grid operations along with specific inputs on the event location. This interpreted and refined data helps to provide means to the Distribution System Operator (DSO) for informed decision-making and time-efficient resource allocation for failure mitigation purposes. The different FADS functionalities work in unison to detect incipient failure signatures and extract valuable information, which can be then used to plan mitigation strategies. The different functionalities of FADS are designed to be installed in a manner such that the incremental costs of widespread FADS implementation are minimal. The evaluation of FADS in this thesis is conducted through a series of stringent and realistic test cases. The test cases are simulated on the standard IEEE-13 and IEEE-34 node test feeders. The first set of simulation studies focusses on detecting High Impedance Faults (HIF) as conventional protection schemes mostly fail to detect it. The test cases comprise of several novel stringent scenarios to evaluate the capability of FADS to accurately distinguish and detect HIF events among multiple switching events and normal grid actions at different sections of the grid. The second set of simulation studies involves recreating transient behavior generated due to real life incipient equipment failure conditions in laboratory based simulations. Simulations are used to evaluate the ability of FADS to detect and assess the incipient failure before the equipment breakdown occurs. The next set of studies is focused on analyzing how the FADS performance in previous simulation studies could be translated to assess the improvement in major reliability indices, mainly System Average Interruption Duration Index (SAIDI). Improvement in reliability indices are a major area of concern for utilities and the results obtained from FADS implementation are further quantified to provide a range of possible improvement in SAIDI value in percentage terms. Finally, the proposed benefit of FADS is illustrated through implementation on real field data provided by the Dutch DSO, Stedin B.V. In the course of FADS implementation, few shortcomings were noticed and possibilities of further improvements were identified. The final chapters of this thesis discuss the shortcomings and recommend improvements for future research studies. The functionalities of FADS are flexible and mostly user-dependent and can be systematically improved over time to make FADS a global standard for industrial and research applications for failure anticipation in AC distribution systems.