A Reasoning Chain Based Algorithm Fault Diagnosis In Power System

In 21 Century, with the construction of national infrastructure and the development of economics, electricity demand have raised step by step. Power system is getting bigger and more complex. The security and reliability of power grid have become a nationwide issue. Accidents will inevitably occur in the power systems because of natural and man-made factors that will result in huge economic losses as well as negative social impact if the power supply cannot be restored in a short time or crushed region cannot be limited on a small scale. Thus, how to diagnose fault quickly becomes a very important research subject.What we called fault diagnosis means using intelligence algorithms and rapid computations by computers to detect fault section, evaluate protection’s operation characteristic and provide related foundation for dispatchers’ decision-making in accordance with the gained information when a fault occurs in power systems. The reason for power system fault is not only related to primary equipment where the fault occurs, but also to relay protection device or communication system. After many years’ development, our relay protection device has evolved from electromagnetic protection to fast and convenient microcomputer protection. Even though, malfunctions and refused-operations of protection should not be ignored. Therefore, it’s extremely important to explain the whole process of fault during diagnosing trouble equipment and evaluating protection behavior.Recent years, a lot of artificial intelligent methods for the power system fault diagnosis have been proposed, including the expert system-based method, artificial neural network-based method, Petri-nets method, topology search method and cause-effect network-based method. However, the existing fault diagnosis methods do have some weakness. In some traditional methods, the nodes of analysis matrix contain all of the fault equipment and information of protective relays and circuit breakers. When the fault section is in a wide range, the matrix which is used to diagnosis fault is in a large scale and affects the rapidity and accuracy of diagnosis result. Moreover, those diagnosis methods have not considered comprehensive protection configuration, and closed loop will appear in the process of diagnosis.In such background, a reasoning chain based fault diagnosis that makes use of the timing sequence and logic behavior of fault data is presented. This method is better in explaining the cooperative relationship among fault equipment, protection relays and circuit breakers. It can also pre-process fault data, thus reducing the dimension of the problem and increasing the speed of diagnosis. The core of this reasoning chain based method is a chained data structure with a computation complexity of O(n), which is easy to accomplish the explanation of fault processing. This new method improved the fault tolerance of diagnosis by considering reclosing protection failure protection and the protection of general component, and implement redundant construction of reasoning chain. Redundant reasoning chain combinations is constructed to get the optimum result based on the likelihood function of the prior data.The last chapter solves a problem about omission of remote data in the fault diagnosis based on establishing virtual event sets in reasoning chain. This solution improves the accuracy and reliability of diagnosis even more. A realistic fault scenario from an actual power system is analyzed to prove the correctness of the presented reasoning chain based algorithm.