Analysis Of Sulfur In Coal Based On Partial Least Squares Regression By Near-infrared Spectroscopy

The sulfur in coal has a negative effect on the equipment and production in the related processes.Moreover,the sulfur oxides from coal combustion seriously pollute nature environment,endanger human health and destroy ecosystem system.In order to solve these issues effectively,it is necessary to determine the sulfur content of coal.Currently,the common method to determine sulfur is chemical analysis of samples in laboratory,which needs a long time and causes the lag of detection results.As a fast and on-line analysis technology,near-infrared spectroscopy can measure the sulfur content of coal in a very short time.So,it is of great significance to optimize the usage conditions in time and to realize the full combustion of coal.This thesis introduced the principle,characteristic,algorithm and evaluation index of near-infrared spectroscopy,and then fast and non-destructive detection of sulfur content in coal with diffuse reflectance near-infrared spectrum analysis technology was studied in this thesis.Collecting diffuse reflectance near infrared spectra of 221 coal samples and detecting their sulfur contents while control the experimental conditions of laboratory.These coal samples included ultra-low sulfur coal,middle low sulfur coal and middle sulfur coal.Choosing 176 coal samples from the whole as calibration set.Calibration set was used to build the regression model of sulfur.And the rest were prediction set,which was used for detecting the model's prediction capability and regression effect.Firstly,selecting partial least squares regression to build the regression model,and the prediction capability and regression effect of the model was good.Then different outlier detection methods was used to optimize the model,and the results revealed that the mahalanobis distance had The best optimization capability.On this basis,the optimization ability of different spectroscopy pre-processing for the model was investigated,and the results showed that the auto-scaling method was most suitable for the model.Finally,after the two phase optimization,the effects of different band selecting algorithm on the model were compared,the result showed that selecting bands according to the relativity of spectra and coal's sulfur content can optimize the model.The regression model of sulfur in coal,which was established by partial least squares regression method combined with these three best methods,had the best regression effect prediction capability.