Research On On-line Soft-sensor Technique Of Coal Particle Size Based On Support Vector Machine

Pulverized coal particle size on the efficiency of coal-fired boiler,boiler combustion efficiency and safety impact is essential,however,because of technical and economic reasons to make the coal particle size of the on-line detection can not be achieved,at present most of the power plant still stay in the off-line sampling of coal particles on a regular basis for measurement.This paper draws on the successful experiences of soft measurement technology in the measurement of carbon content,flue gas oxygen content and ball mill load forecasting in power plant,based on the support vector machine soft measurement method,study the particle size of coal powder on-line measurement.The paper introduces the basic principle and modeling method of the soft measurement,finds the impact of coal particle size of the relevant factors from the milling system analysis of the mechanism.Based on the actual operating history data of Guohua Sanhe Power Plant,correlation analysis method was used to preprocess the original data,and then the auxiliary variables used in the soft-sensing model were determined.The feasibility of this method is analyzed theoretically,basing on the SVM support vector machine,the support vector regression(SVR)method is used to model and implement the soft-sensing scheme.Grid search method,Genetic algorithm and Particle Swarm Optimization(PSO)are used to optimize the parameters.And then the realization process of soft-sensing method based on SVM is introduced in detail.Finally,after a number of experimental comparison,the best soft sensor model parameter values is found.The model is used to forecast the test sample and compare it with the actual coal particle size data.The conclusion that the model can accurately estimate the value of coal particle size by grid searching and genetic algorithm is obtained.which fully demonstrates the feasibility of on-line measurement of coal particle size by using support vector machine(SVM)method.It is proved that the method has the condition and value in practical production.