Power System Alarm Processing And Fault Diagnosis: Analytic Models, Methods And Applications

When a fault occurs on a section in a power system, the corresponding protective relays (PRs) should operate to trip related circuit breakers (CBs) off, so as to isolate the fault section from the healthy part of the power system. If a malfunction or malfunctions are happening with these PRs and/or CBs, the outage area could be significantly extended. Modern integrated substation automation systems and dispatching automation systems are capable of monitoring, measuring and controlling all equipments in substations, and hence could reduce the burden for system operators. However, on the other hand, this has also brought a rapid increase in the volume of alarms. Especially when a fault or abnormality occurred on a power system, a flood of alarm information such as the PR operations, the CB tripping, and oscillography information, would be sent to a control center without any analysis. In these situations, it is extremely difficult for an operator to digest information and hence identify what has happened in a short time. A precise and effective method for power system alarm processing and fault diagnosis plays an important role in supporting rapid fault section identification and network restoration for dispatchers, so as to ensure the safe and steady operation of power system and enhance the reliability of power supply.Based on the existing analytic model-based approaches, precise and effective mathematic models and methods for power system alarm processing and fault diagnosis are developed, combined with several types of Artificial Intelligence (AI) technology. The temporal information of alarm messages as well as the oscillography information generated by Data Fault Recorder (DFR) are well utilized in this thesis. Moreover, the situations including multiple faults, malfunctions of PRs and CBs and the abnormality of alarm messages (missing or false alarms) are taken into account. Some significant research results are obtained as follows:1) Based on the temporal constraint network (TCN), a new analytic model is developed for alarm processing with temporal information taken into account. The TCN is employed to represent the randomness of event occurrences and hence the adaptability and the fault-tolerance are enhanced. The developed methods could find out not only what events caused the reported alarms and when these events happened, but also identify the missing or false alarm messages.2) A novel analytic model employing the temporal information of alarm messages is developed. The expression of fault hypothesis is extended considering the temporal information of alarm messages. The concept of related path is introduced to clearly feature the complicated coordination of protective relays of modern power systems with multiple protection configurations. According to the temporal relationship among the section faults, the actions of PRs and the tripping of CBs, the expected states of PRs and CBs are determined. The temporal relationship among the actions of PRs and CBs, and the different protection configurations in a modern power system can be reasonably represented by the developed model, and therefore, the diagnosed results will be more definite under different circumstances of faults.3) Based on the existing analytic model-based approaches, malfunctions of PRs and CBs are taken into account systematically. A new form of the fault hypothesis is presented, including the information about"the actual operating states of PRs and CBs (normal or malfunctioned)". The key issue is to determine the expected states of PRs and CBs corresponding to a given fault hypothesis. The fault diagnosis model could not only estimate fault section(s), but also identify the malfunctioned PRs and CBs, as well as the incorrect and missing alarms.4) A waveform matching and optimization based approach for fault diagnosis of a high voltage transmission line (FDHVTL) is proposed. The diagnosis problem is formulated as a mixed integer programming one with both discrete and continuous valuables. Harmony Search, an effective heuristic optimization algorithm developed in recent years, is employed to solve this problem. Fault type, fault location, fault time and indefinite variables such as ground resistance can be obtained at the same time.Finally, several conclusions are made based on the research outcomes, and directions for future research indicated.