| Techniques for on-line monitoring and fault diagnosis of transformer play an important role in improving the security and the stability of the power system, and that of on-line monitoring based on dissolved gas analysis in the transformer oil is the most popular and important techniques in the fields of on-line monitoring of transformer. Some universal defects exit in the running products, for instance, only a few kinds of gases can be detected, the accuracy is not enough, the monitoring equipment is large in volume, and the cost is high.The paper is concerned with the chromatography on-line monitoring and fault diagnosis system. To overcome the insufficiency of other on-line monitoring methods, this paper has made it improved. And based on the analysis, a gas monitoring system of transformer oil is designed in this paper. The system can accurately monitor the characteristic gases dissolved in the transformer oil, reflecting the running states of the transformer real-timely. This paper also studies and emulates the algorithm of the fault diagnosis. A method using micro-volume thermal conductivity detector based on MEMS technology is presented in this paper. To improve the sensitivity of the monitoring equipment, increase the segregation degree and stability, and reduce the volume of the equipment, based on theory analysis and experiments, considering the characteristic of the detector, the stationary phase and operating conditions, including column internal diameter, column length, column pressure and sample quantity are determined.According to the characteristic of the high voltage equipments and the on-line monitoring technology, an algorithm based on data fusion is put forward in this paper. The negligence errors are eliminated by the method of distribution diagrams, and the statistical method based on recurrence estimation are used to the sample data processing. The algorithm remains the local and global characteristics of the measured data. The results of experiments have proved the validity of the method.On the basis of getting the true and reliable monitoring datum, the methods of diagnosis algorithm based on state predicting are alsodiscussed in this paper. A method based on BP artificial neural network is presented. A diagnosis model is also designed and emulated. The results of the emulations show that the algorithm is obviously better than that of the traditional three-ratio method, and the algorithm can accurately makes good judgments on the faults base on qualitative and quantitative analysis. |
