| Summary: | 博士 === 國立臺灣大學 === 電機工程學研究所 === 88 === As an auxiliary tool, the purpose of this dissertation is to set up an on-line fault diagnosis system applied to automation substations. The proposed diagnosis system can provide correct fault information with rapid inference speed for operators in control center. Through correctly identify the fault sections and fault types, the faulted sections can therefore be isolated quickly and the restoration time is reduced. Most of the traditional methods used to develop diagnosis tools are heavily relied on the use of the rule-based Expert System, which has drawbacks in its slow inference speed and has difficulties in knowledge base revision or maintenance. In recent years, some researchers use Artificial Neural Networks as an alternative solution to these problems and demonstrate the feasibility in some related papers. However, some problems still remain unsolved in practical application so far, such as the systematic determination of the network parameters like hidden units, layers, learning rate etc. In addition, the training process of an Artificial Neural Network is tedious. When any configuration of the system changes, the related neural network needs to be re-trained. To overcome the drawbacks of the traditional methods, this dissertation proposes a new method by using the hybrid cause-effect network/fuzzy rule-based to implement an on-line fault diagnosis system. Since the cause-effect network is a graphic-modeling tool, it is much more useful for illustrating the relationship between faults and protective devices. Besides, the proposed algorithm has good transparency and fast inference speed as its advantages. To identify fault types, a fuzzy rule-based method is derived to improve the inadequate of selecting a fixed threshold value, which is unable to deal with the uncertainties involved in the process of clarifying faults in practical consideration. The proposed approach has been practically verified by testing on a typical Taiwan Power Company’s (Taipower) secondary substation. The experimental results reveal that the correct and rapid diagnosis is obtained even for the fault domains involving multiple faults with different types and failure operations of protective devices. Moreover, it is easy to implement and transplant into different substations.
|
|---|
