Study On Rapid Detection Of Oil Yield Of Oil Shale Based On NIR Technique

Oil shale is one kind of unconventional oil and gas resources, its global reservevolume is very rich and it is an important alternative energy to oil and gas. With therapid development of our economy and the growing consumption of oil, the oil shale isdrawing various department’s attention. Oil yield is an important indicator to evaluateoil shale, and it is an important factor is to achieve industrial use. The traditionalmethod for determination oil yield of oil shale is aluminum screening method. Thismethod requires the laboratory environment, cumbersome testing procedures,andcosts a long time for sample analysis. It can not meet the need of oil shale sampledetection in the exploration and exploitation.Near Infrared Spectroscopy has been in the rapid development these years andhas been widely used in the oil reconcile, reforming and crude oil distillation processin the field of petrochemical industry, etc. However the current application of oil shaleresearch using laboratory spectrometers are all full spectrum scanning, whose speedis slow, is generally more than30seconds. Scanning process is prone to interferenceand is easily affected by many factors.In this paper, we study the kaolin mixed machine oil as the standard sampleanalysis, using near-infrared spectroscopy and chemometrics analysis, and propose arapid detection method by near-infrared oil yield of oil shale. We use the method of acombination of selected characteristic wavelengths point to greatly reducing thespectral scanning time and reduce interference in the measurement process andsimplify the modeling data. By this way, it can also provide the basis for the choice ofspectral components of the discrete NIR spectroscopy.First, we use the laboratory spectroscopy instrumentation combined withchemometrics methods to select the best combination of wavelength: According todiffuse reflectance spectra data obtained by spectrometer, we achieve the judgment and removement of abnormal spectrum, multivariate scatter correction,Savitzky-Golay smoothing to remove noise; based on the correlation coefficient figureof spectral data and oil content, we initially select20characteristic wavelengths points;we use the combination generation algorithm to obtain all combinations of theselected20wavelengths from2~9points, and use multiple linear regression toestablish the calibration model, and compare each coefficient and the relativestandard deviation of the modeling set and the prediction set, and determine the best4wavelength combination to detect the oil yield of oil shale. Finally we set thelaboratory spectrometer to a mode of single point detection, re-detect the data of theselected wavelengths, and re-model the spectral data with the oil content data, for thesame time, we compare the results of the untreated one and multiplicative scattercorrection one. We configure application sample and analysis it to verify the feasibilityof the wavelength combination to forecasting oil content.