| Logging is of great significance in oilfield exploration and development.Accurate logging results can increase oil and gas mining efficiency and production.In my country,logging analysis technology is mainly based on gas chromatography.However,this technology has shortcomings such as long analysis period,limited detection gas type,and many auxiliary equipment and complicated operation.In addition,Raman spectroscopy has the advantages of fast analysis speed,non-contact type and high detection efficiency in gas detection,but it is not perfect enough for logging analysis and it is expensive to purchase foreign Raman spectroscopy analyzer for logging analysis.Therefore,this dissertation firstly studies the quantitative analysis and detection technology of logging gas based on Raman spectroscopy,and then the radial basis neural network and genetic algorithm are combined to realize the quantitative analysis of the components of logging gas.The main research contents and results of this dissertation are as follows:1.Design of logging Raman spectral acquisition systemRaman spectroscopy acquisition system is mainly composed of Raman spectroscopy,embedded controller,computer,gas pretreatment unit and so on.The system mainly controls the constant temperature,pressure and flow of the Raman spectroscopy acquisition system through the embedded control unit,and then uses the computer to complete the experimental data processing,thereby realizing the collection and analysis of the logging gas Raman spectroscopy data.The results show that the Raman spectral signal is obvious,and the signal-to-noise ratio meets the detection requirements.2.Design of multi-channel calibration device systemThe multi-channel calibration device is mainly composed of a pressure regulating valve,a mass flow meter,a mixing chamber and so on.The device can mix standard alkane elemental gases in a certain ratio to obtain a mixed gas of known component concentrations.The experimental sample is prepared for establishing the GA-RBFNN model,and the configured experimental samples are verified by the Agilent 6820 chromatograph.Results show that the average error of the sample gas prepared by the multi-channel calibration device is 2.4%,which meets the experimental requirements.3.Quantitative analysis of each component of logging gasIn this dissertation,many sets of test samples are designed by orthogonal experiment,and the actual samples are configured by multi-channel calibration device.After that,the Raman spectrum of these samples is obtained by Raman spectroscopy acquisition system,and finally the quantitative analysis model of GA-RBFNN is established to predict the content of logging gas(CH4,C2H6,C3H8,n-C4H10,i-C4H10,n-C5H12),based on the radial basis neural network and genetic algorithm.The results show that the correlation coefficients between the predicted concentration and the actual concentration of the logging gas components are 0.9888,0.9897,0.9882,0.9886,0.9897,and 0.9902,respectively,and the predicted root mean square error RAMSEP is 0.482,0.648,0.569,0.521,0.729,and 0.642,respectively,which indicating that the experimenta l results meet the requirements. |
PDF Full Text Request