Approach With Information Fusion Of Multiple Data Resources To Fault Diagnosis Of Power Grid

Along with the establishment of the information network for protective relaying and fault data recorders, the plentiful recorded fault data lay a foundation for further diagnosis. The rapidly growth of the physical complexity in power systems and the running of power market bring more and more unanticipated and indeterminate factors to power systems, that render the security and economy operation of power systems face up to new serious challenges. Therefore, fault diagnosis as an important part of power dispatch, has novel request in the precision and real time performance.When faults occur in power systems, fast and correct fault section diagnosis is very important to power system restoration. Progress in the areas of communication and digital technology has increased the amount of information available at supervisory control and data-acquisition (SCADA) systems. Although that information is very useful, during events that cause outages, the operator may be overwhelmed by the excessive number of simultaneously operated alarms. Further more, when the complex fault occurs, such as multiple faults, expanded faults, and a breaker or its associated relays fail to operate, fault diagnosis of power system turns more difficult. With the expansion of the size of the power system and enhancement of the real-time requirements, these issues will become increasingly complex.The previous researches of fault diagnosis are mostly based on digital information (the status of protection and circuit breaker). A number of techniques have been employed to solve this problem, including expert systems, neural networks, fuzzy set, artificial intelligence techniques and Bayesian method. But only digital information is used in those methods, the analogue information (the value of current and voltage) does not been fully used. However, the analogue information has the enormous advantage in accuracy, completeness and fault tolerance.In this paper, the trend of the development of fault daignosis is introduced. Then the element-oriented Petri net with time-stamp is proposed to perform faultdiagnosis. Next, based on digital and analogue information, fuzzy Petri net and wavelet analysis are used to extract the fault character, so the corresponding fuzzy fault degree and wavelet fault characteristics (wavelet fault degree, wavelet singularity degree, and wavelet energy degree) are obtained. Meanwhile, the improved D-S evidence theory is used in this paper to solve the inaccurate problem of fusion results which is caused by serious conflict among evidences. Then the fault characteristics are considered as the independent evidence to get the diagnosis results via information fusion. The proposed method analyzes multi-information sources, so the fault information can be expressed exactly, and the correct results can be made by information fusion. Moreover, the case study results verified validity of both the diagnosis model and method and the potential of their application into fault diagnosis.