Data Modeling Of Gas Compressors

With the constant expansion of industrial scale and increasing complexity of industrial processes, centrifugal compressors are widely used in the industry fields such as metallurgy, petroleum, chemical industry, natural gas transportation, refrigeration etc. It has lots of advantage points, but it also has some insurmountable shortcomings such as unstable in low flow area, narrow steady operating area, easy surge etc. Surge is the inherent characteristics of the centrifugal compressor, which will endanger the safety of the compressor seriously. The judge must be made quickly so that compressor will escape from surge conditions in order to ensure stable and long cycle of centrifugal compressor operation.The gas system of combined cycle power plant of Baosteel is took as the research object in this thesis. In order to deal with the above-mentioned problems, the mathematical model of centrifugal compressor is established by data modeling and mixed modeling from the actual view of system operation. Because the system has many variables, nonlinear characteristics, as the accuracy of the traditional data modeling method can not accurately deal with the problem of nonlinear precision, the kernel partial least squares method is used in this thesis. Because low dimensional space nonlinear regression is put into high dimension space linear regression by using kernel functions, KPLS method can overcome the negative impact which the nonlinear factors have on the prediction model in the actual production process.Centrifugal compressor model based on KPLS method is established by using field data with the gas composition, mass flow, the inlet pressure, the inlet temperature as input variables. The centrifugal compressor mixed model is established in order to solve big error and low accuracy of the mechanism model, mixed model which is trained by KPLS method is made up of the mechanism model and error compensation device. The simulation analysis shows that compressors model based on the KPLS method is accurate and reliable.