Research On Transformer Condition Monitoring System Based On Dissolved Gas Analysis In Oil

Currently, a large number of transformer condition monitoring system put into use, however, according to statistics, in addition to their high failure rate of defect, fault diagnosis and status correctly predicted the low rate is also an obstacle to its role in the play. Based on the deep understanding of the complex mapping relationship between transformer fault types and characteristics, the paper not only improves the theoretical method of fault diagnosis module and state prediction module, also improves the hardware structure and the main IED of the transformer condition monitoring system, and has been successfully implemented in Weishan Dali substation. The main contents of this paper include:Fault diagnosis module: Firstly, based on historical data, a probability distribution transformer abnormal warning methods, namely by making statistical analysis of historical data, to determine the probability distribution, thereby accurately divide warning range, to achieve accurate early warning purposes, as troubleshooting provide a solid foundation, and by Dali Weishan 110 k V substation and 220 k V substation in Jinghong changed, indicating its advantages. Furthermore, the proposed fault diagnosis method based on a variety of combined ratio wears borda model, the use of the comprehensive improvement model borda three ratio, five common ratio Rogers ratio method, Japanese electric co-research method, IEC-60599, the ratio of non-coding method law, to take advantage of the traditional ratio method, but also avoid the single preference ratio method. Meanwhile, for the ratio of the boundary problem of false positives, is proposed based on informa tion entropy- transformer fault diagnosis method of weighted gray relevance, namely by improving the general gray correlation analysis method, the weighted gray correlation analysis, the difference between the standard spectrum factor technology, gray correlation coefficient distribution were weighted, giving each factor a twofold objective weights, make the most objective diagnosis, and by example to verify the advantages and feasibility of this approach.State prediction module: Initial data processing model for GM, GM model parameter optimization, error correction and other aspects were optimized to improve the accuracy of forecasting GM. cos(x)?Function Transformation of data preprocessing in order to improve the smoothness of the initial sequence; Changing the traditional form of the optimization of the development coefficient a and the ash content of b, optimization of gray background value generating coefficients, so that optimization is more reasonable and practical; for accuracy is high, the use of verhust model prediction errors, the prediction correction value. Through the analysis of the error, the comprehensive forecasting method of particle swarm optimization +verhust model GM model can improve the accuracy of prediction to a great extent.Condition monitoring system: according to the actual situation Dali Weishan substation, designed a set of integration can be achieved on the transformer monitoring system, and the main IED has been optimized to improve the speed of operation, optimizing the allocation of resources, has a good economic benefit.