Design Of Infrared Spectrum Detection System For Logging Gas Based On Deep Learning

Logging technology is the "eye" of oil and gas exploration and development.Fast and accurate logging analysis technology can improve the efficiency of oil and gas resource development and utilization.At this stage,most of the methods used in China's logging analysis are gas chromatography.Although this analysis method can detect and analyze logging gas,it is difficult because of the shortcomings of many auxiliary equipment and long analysis time.Meet the needs of fast-growing logging technology.Infrared spectroscopy analysis technology can quickly and non-destructively detect gas,and has been widely used in many fields of life,but the application in the field of logging analysis is not perfect.Therefore,this paper studies the quantitative analysis technology of logging gas based on Fourier Transform Infrared Spectrometer,and uses the convolutional neural network in deep learning to quantitatively analyze the content of each component gas in the logging body.First,a logging gas infrared spectrum detection system is designed.The system is mainly composed of a gas pretreatment unit,a gas distribution unit,a gas sensor unit,a Fourier Transform Infrared Spectrometer,an embedded control unit,and a computer communication unit.The embedded control unit detects the temperature,pressure and dangerous gas concentration in the entire instrument and transmits it to the computer for real-time monitoring.The infrared spectrum signal collected by the Fourier Transform Infrared Spectrometer is directly transmitted to the computer for processing.Secondly,this article uses orthogonal experiments to design multi-component experimental samples,uses a multi-channel gas mixing device to configure the experimental gas,uses the KS algorithm to group the acquired infrared spectrum data,and divides the divided training set and The prediction set data was trained by a convolutional neural network,and a deep-learning quantitative gas analysis model was established based on deep learning.The quantitative analysis of methane,ethane,propane,n-butane,and npentane was completed.Finally,through experimental verification,the Fourier Transform Infrared Spectrometer detection system designed in this paper can accurately detect the infrared spectrum of the logging gas,and the quantitative analysis model of the logging gas established can identify 99% of the pretreated elemental gas.The accuracy rate,the recognition accuracy rate of the pre-processed mixed gas infrared spectrum reaches 98%,indicating that the deep learning-based infrared gas spectrum detection model established in this paper can quickly and accurately perform quantitative analysis of the gas.The logging gas infrared spectrum detection system can accurately detect the simple alkane gas and multi-component mixed alkane gas,which is of great significance to the exploration and development of oil and gas resources.A fast and accurate analysis of the gas composition of the well logging can avoid the omission of thin oil layers and lean oil layers to the greatest extent and improve the supply capacity of China's oil and gas resources.