Study On Smart Sensor Of The On-line Monitoring Gases Dissolved In Transformer Oil

Large power transformer is the important equipment in power system. Its running state is straightly related to the dependability of the whole power system. Dissolved gas analysis (DGA) method has been widely applied in power system, and become routine means of diagnosing power devices. In this paper DGA method is applied to diagnose the transformer fault, and an intelligent sensing system based on artificial neural network (ANN) pattern recognition technology and sensor-array technology is also presented to avoid the cross-sensitivity among gas sensors. This paper is mainly studied in followed:1) The condition and application prospect of the transformer oil dissolved gas on-line monitoring technique and the applied intelligent sense technology are analyzed deeply. The gas detection principle of gas sensors and the basic theory of gas sensor-array are presented. The principle, method and constitution of single, mixed gas identification of elements and concentration are also studied by using gas sensor-array and ANN. Generalization ability is adopted to train the neural network.2) The pattern recognition technical and basic theories in ANN are analyzed. Gas analysis method with the neural network of BP and RBF recognition is proposed.3) Single gas component quantitative analysis method is studied, which is based on the neural network model of BP and RBF. The method of the qualitative recognition is applied to study mixed gas using the neural network model of BP. A gas mixture analysis system based on ANN is designed, and the primary factors of low precision for gas analysis are investigated. There are the problems of the sample quality and the local minimum of network training. To solve these problems, a sample pre-treatment system and improved network training algorithms are proposed. The experiment results show that the precision of the gas analysis has been improved greatly.4) The intelligence gas sensors are produced, which is based on the gas sensor-array and pattern recognition algorithm. The qualitative analysis and quantitative analysis to mixed gas are realized.