Research On Cross-interference Characteristics And Sensing Method Of Transformer Light Gas Mixed Gas

Transformer is one of the most important equipment in power system.Whether it can work normally is related to the operation of the whole power grid.Gas relay is an important protection device on transformer equipment.When the fault occurs,the fault gas will alarm through the pressure wave.The light gas alarm also indicates that there may have been a fault inside the transformer.Nowadays,the method of artificial gas extraction can only know whether there is a fault and cannot judge the type of fault.In this paper,through the selection of sensors,a gas cross-interference detection test platform is built to study the qualitative identification and quantitative estimation of mixed gases by intelligent algorithms,and to solve the problems of inaccurate gas measurement and poor recognition ability under the influence of cross-interference between various mixed gases.The main work of this paper is to realize the on-line monitoring device of light gas combustible gas :(1)The gas sensors are selected for the three target gases of hydrogen,carbon monoxide and methane,which are common in the combustible gas of light gas.Combined with the principle,advantages and disadvantages of the sensor and the selection requirements,three different manufacturers are selected.The gas sensing experimental platform is composed of mixed gas,test chamber,gas sensor,detection unit and data receiver.The gas sensor that best meets the test requirements of this paper is selected for the study of its repeatability,concentration characteristics and dynamic response-recovery characteristics..(2)Using the principle of orthogonal test to establish an orthogonal table constructed by three factors and nine levels to select the ratio of light gas mixed gas,and using the gas sensitivity test platform to carry out a series of gas sensitivity tests to obtain test data.(3)A field fast mixed gas pattern recognition model constructed by the crossinterference compensation correction method of mixed gas is established,and the preprocessed sample data is used to test the example.The results show that this method can reduce the cross-interference between the mixed gases and realize the rapid detection of the transformer on site,but it cannot meet the requirements of on-line monitoring of the transformer.(4)A BP neural network based on genetic algorithm is established to quantitatively analyze hydrogen,carbon monoxide and methane gas.It is suitable for quantitative identification of light gas combustible gas.The results show that the GA-BP method can effectively reduce the cross-interference between the mixed gases,indicating that the algorithm has high accuracy and reliability for the prediction of the three target gases.