Analysis Of Coal Properties By Near Infrared Spectroscopy

Moisture,ash content,volatile and fixed carbon of coal is basic indicator of coal quality.But it is complicated and time-consuming,which is unfavorable for coal-fired power plants to control the coal.Therefore,looking for a quick analysis method is very necessary.Near infrared spectral analysis technology is a kind of on-line analysis technology.It can realize on-line detection of coal quality,which is convenient and quickly.It is significance for coal-fired power plant coal to detect the quality of coal.Near-infrared spectroscopy(NIRS)is an on-line analytical technique,which can easily analysis on-line detection of coal quality without loss of samples and pollution.It is very important to detect coal quality in coal-fired power plant.Near Infrared Spectroscopy(NIRS)provides a new rapid detection method for coal quality.This paper studies the application of near infrared spectral analysis technology to the coal of moisture,ash,volatile and fixed carbon.This article selects 184 coal samples to build prediction model.Application the K-S classification method to select 147 samples as a calibration sample set and 37 samples as predictio n sample set.Under strictly control the temperature and humidity of the laboratory,measured the spectral data and values of each sample.Selection the partial least squares algorithm as the main algorithm for modeling,respectively for moisture,ash cont ent,volatile matter and fixed carbon of coal samples.Screening the best main factor for each model and the optimal spectral preprocessing methods.Using Mahalanobis Distance method and Student's Residual method eliminate outliers in data,to ensure that the model has higher prediction ability.Results show that the model of moisture's best main factor is 7 and the optimal spectral preprocessing methods is OSC;the model of volatile's best main factor is 8 and the optimal spectral preprocessing methods is NAS;the model of ash's best main factor is 9 and the optimal spectral preprocessing methods is the mean centralized method;the model of fixed carbon's best main factor is 7 and the optimal spectral preprocessing methods is OSC;the best method for the four model is Mahalanobis Distance to eliminate outliers in data.The model of moisture and volatile prediction ability is superior to ash and fixed carbon.