Multiple Knowledge-based Power Transformer Fault Diagnosis Expert System

The complex structure of the power transformers results in its complex failure forms and electrical off-line examinations are usually used as main methods for their fault diagnosis. However, it is difficult to online detect failures, and it may cause excessive maintenance. Oil dissolved gas analysis(DGA) can be used for effectively detecting the faults of transformer, but it cannot locate the faults of the power transformer. Thus, motivated by the problem encountered when both detection and location of these faults were needed simultaneously, we developed a three-phase power transformer fault diagnosis system. The main results are shown as follows:1,A Huffman SVM based fault classifier was put forward. At first, the classifier was used in converter fault diagnosis,66 waveforms was exported from a three-phase converter whose model was set up in MATLAB, the feature value was extracted by combination with wavelet analysis and singular value decomposition, and then the waveform was classified by Huffman SVM. The experimental results show that the correctness of the proposed method exceeds 95%. Then the fault classifier was applied to fault diagnosis of power transformer, and ten sorts of transformer faults were classified. The experiment results show that the correctness can reach over 90%.2,An expert system based on knowledge base was proposed aimed at the complex failure forms of the power transformer. Three knowledge were established, including DGA knowledge base, insulation system examination knowledge base and water content detection knowledge base. In view of the large quantity of the DGA knowledge base, rule confidence was used to avoid the collision of the fault information. Because the fault probability is great at the centre but small in either end, a dynamic knowledge base was set up which was scanned firstly. The result of the experiment indicates that the expert system possesses a high correctness and high efficiency.3,A power transformer fault diagnosis system software was developed with VS2008 and SQL Server2005. The system software is made up of 6 modules, which include the running state module, data view module, user and equipment management, sampling plan and system log. The system integrated a transformer expert system based on knowledge base, which can automatically analyze and process the data received by the transformer, and then present the treatment recommendations. The application results show that the system can real-time monitor power transformers effectively.