Research On Power Distribution Network Fault Diagnosis By Means Of Multi-Source Information

In modern electric power industry, energy consumption have continuously grown over decades, system failures have tremendously expanded among grid along with the highly extended power grid and more closely related networks. Distribution network, which plays a vital role in connecting between generator, transmission system and consumers, its safety and reliability is of great importance in regards the stability of the power supply. Once the failures happen in distribution system or operating condition is threatened by outside factors, the failures may develop into cascading failures if there are no prompt measures being taken, furthermore causes tremendous economic losses as well as inconvenience brought about to consumers. Up to date, the conventional fault diagnosis of distribution network is based on the breaker and relay protection information, which is unable to provide power system dispatchers with adequate basis for accident analysis. In recent years there are many data acquisition and monitoring systems have emerged aiming to monitor the dynamic information completely and acquiring fault information quickly. Therefore, research on fault diagnosis methods which is able to make the most use of multi-source information are of magnificent significance.In this thesis, source of fault information in the distribution network and coupling between fault data are analysed detailedly, and to the existing questions in data level fusion and feature level fusion, some corresponding solutions are proposed. Simultaneously, after analyzing the source of fault information in the distribution network, a dynamic hierarchical fault diagnosis method is proposed in this thesis, which is on the basis of distribution automation system, wide area measurement system and fault information system. There are three layers in this method containing switch layer, feeder layer and substation layer, and the diagnosis strategies can be adapted dynamically among various layers according to characteristic of fault information.In the switch layer, The depth-first searching method based on network incidence matrix is applied to determine the fault section quickly according to fault information of switches. And the Petri net theory is applied in the feeder layer to obtain the fault component accurately according to fault information of switch and protection. Simultaneously, the intuitionistic uncertainty rough sets theory is also applied in the substation layer to obtain the fault component completely according to fault information of switch, protection and electric quantity. This method can take advantage of fault information adequately to improve the accuracy and reliability of fault diagnosis, it also can improve the efficiency of diagnosis benefiting from the dynamic pattern.Taking the effect of uncertainty and inconsistency during the process of data acquisition into account, the intuitionistic uncertainty rough sets theory is applied in the substation layer in this thesis. After analyzing the defect of the fault diagnosis method based on rough sets theory, a method based on intuitionistic uncertainty rough sets theory is put forward. In this method, fuzzification has replaced discretization of continuous attribute, which can improve the ability of handling uncertainty.Finally, a fault diagnosis system which is based on multi-agent system is proposed in this thesis according to characteristic of fault information’s spatial distribution and diagnosis method’s gradational structure. The system adopts distributed structure which obtains diagnosis result from local agent to global agent, and the optimal coordinated agent is applied to help all agents achieve common goals. The functional requirement and structure model of various agents is given, and the interaction between agents is described in detail. The application of multi-agent system further improves the instantaneity and reliability of fault diagnosis.