Study On Soft Measurement Method Of Carbon Content In Fly Ash

The carbon content in fly ash seriously affects the economic benefits of boiler combustion in thermal power plants.High carbon content will increase the heat loss of combustion,decrease the combustion efficiency,and even cause t a safety hazard to the boiler unit.At present,the carbon measurement instruments based on physical measurement methods are difficult to adapt to industrial sites and difficult to maintain.The emergence and development of soft measurement technology provides a new measurement method for the carbon content in fly ash.Based on the boiler combustion system,this paper studies the boiler combustion process,selects the auxiliary variables and extracts the feature matrix,and studies the soft measurement model and optimzation algorithm of the carbon content in fly ash.The main research contents are as follows:(1)Analyze the boiler combustion system process,auxiliary variables such as coal characteristics,boiler load,coal feed amount of coal feeder and burner swing angle are selected.Establish a partial least squares regression model to extract the auxiliary variables.According to the cross-validation method,determine the optimal number of principal components of the feature matrix,and finally determine the 7-dimensional feature matrix as the input variable of the soft-measurement model.(2)Establish BP neural network model,and use improved PSO algorithm to optimize weights and thresholds.The model establishment and model verification are carried out in MATLAB environment,and the results show that the improved PSO-BP model of carbon content in fly ash has high precision and good reliability,and can effectively avoids the PSO algorithm falling into premature convergence problem.The mean square error of the test sample is 0.0049.(3)Establish LSSVM and improve the carbon content in fly ash model of PSO-LSSVM.The results show that the stability and prediction accuracy of the LSSVM model which optimized the regularization parameter and kernel parameter by grid search method are better than BP neural network,and the accuracy of improved PSO-LSSVM model is better than LSSVM model.The above four models are evaluated by means of mean square error,average relative error and maximum relative error.It is concluded that the improved PSO-LSSVM model of carbon content in fly ash has the highest accuracy,and the mean square error of test samples is 0.0032,which satisfies the measurement requirements of carbon content in fly ash.(4)Propose an online soft measurement scheme for carbon content in fly ash based on OPC technology.Data preprocessing and model calculation are carried out under MATLAB,and real-time data communication with WinCC software is carried out through OPC technology to realize online soft measurement of carbon content in fly ash.