Coal Proximate Analysis Method Research And Equipment Development Based On Laser-induced Breakdown Spectroscopy

The coal quality parameters are important for the safe and economic operation of thermal power plants,and at the same time they play an important role in achieving energy conservation and emission reduction.However,due to the lack of on-line measurement technologies,especially coal quality characteristics,the development of optimization operation technologies is restricted in the thermal power plant.The application of laser induced breakdown spectroscopy(LIBS)is expected to achieve on-line analysis of coal quality parameters.The existing study of LIBS-based coal analysis generally require pressing the pulverized coal into pellet,which limit the timeliness of LIBS technology.Therefore,this paper focuses on the direct measurement of particles flow pulverized coal by LIBS.Based on the spectral characteristics of particle flow,a series studies of data pretreatment methods and quantitative analysis model establishment are carried out.Finally,the development and integration work of LIBS coal quality analyzer is carried out.The spectral data of LIBS measurement inevitably fluctuates due to fluctuation of the particle flow.In order to improve the repeatability of measurement,the method based on the Markov's characteristic distance and the statistical distribution based on the normal distribution are used to eliminate the outlier data on the basis of obtaining a large amount of measurement data.The qualitative analysis results show that after the removing of outlier data,the intensity distribution of the characteristic lines is closer to normal distribution,the relative deviation of the characteristic lines is significantly reduced.The effect of the two data elimination methods are equivalent.The problem of baseline drift are solved with BEADS algorithm.Finally,the partial least squares(PLS)is introduced as a quantitative analysis model to fitting the relationship between ash content and LIBS spectrum.The results show that the comprehensive use of data elimination and baseline correction methods can improve the accuracy and repeatability of the ash content measurement of coal.Because of the complexity of the composition and structure of the coal,there are certain matrix effects and self-absorption effects when measuring using LIBS.Artificial neural network(ANN)is introduced as a quantitative analysis model aiming at fitting the nonlinear relationship between spectral data and proximate analysis.The genetic algorithm(GA)was used to optimize the network structure parameters.To reduce the effect of random initialization of ANN,we establish multiple models and take the average output as the finall predicted value.The results show that ANN has a stronger ability of fitting and generalization compared with the linear PLS model.In order to meet the application requirements of industry,the results of the laboratory study and the above data processing methods and quantitative analysis are put into practice.We carry out the development and integration work of LIBS coal quality analyzer.The performance of coal quality analyzer is tested and the results showed that the prediction accuracy of the ash content and gross calorific value was close to the national standard of the neutron activation analyzer.