Research On Boiler Nitrogen Oxide Prediction And Reduction Based On Data Mining

Boiler operation optimization can improve the boiler operation efficiency and ensure the operation safety of the boiler.It generally includes the optimization of equipment and the optimization of operation parameters.Due to the numerous boiler components and complicated operation process,there are many difficulties in the current boiler operation optimization.In recent years,the power plant SIS system has been widely used in power plants and it can fully record the operating data of power plant boilers.Finding the best parameters of boiler operation from the large amount of data accumulated by the system can save the cost of on-site commissioning and save manpower and material resources,which has important practical significance.However,the power station data has strong volatility,many parameters,and large data volume,it is not realistic to use manual analysis.With the gradual maturity of data mining,it has become a research hotspot to apply data analysis to power station boilers.This paper used data mining to analyze the SIS system data of the power plant,aimed to predict the nitrogen oxide concentration of the SCR entry,optimize the boiler operating parameters and then ensure a green and efficient operation.Mainly carry out the following work:Perform data preprocessing.First,we need to select the parameters used in the modeling to achieve the purpose of reducing the dimension of the data.Second,the isolated forest algorithm is used to detect the outliers and then the outliers are deleted,and the vacancy values are directly deleted.In addition,there will be noise interference in the power station boiler data,and the wavelet soft threshold is used to remove noise.Finally,the unsteady data of the power station boiler can not participate in the model establishment.In this paper,the improved adaptive polynomial filtering algorithm is used to extract the steady state data,and the data when burning common coal types is selected.This data will be used for subsequent modeling.The BP neural network and support vector machine are used to establish the NO_X regression prediction model respectively.The particle swarm optimization algorithm(PSO)and the differential evolution algorithm(DE)are used to optimize the model respectively.It is found that the support vector machine is better than the BP neural network.The difference optimization support vector machine in the vector machine model is relatively stable,so DE-SVM is adopted.After debugging,it is found that when the two parameters c and g of the support vector machine are 0.78995 and 3.9628 respectively,the model has the highest accuracy.The average relative error at this time is 2.283%,and the maximum relative error is 9.35%,which is in line with industrial requirements.Data processing is performed using fuzzy association rules.Firstly,the non-equal fuzzy clustering is used to divide the oxygen content of the SCR inlet into five fuzzy intervals,and the SCR inlet NO_X concentration and the pre-export outlet fly ash carbon content are divided into three fuzzy intervals.Then use the association rules to process data,it is concluded that when the oxygen content of the flue gas is[2.86%-3.12%],the concentration of NO_X at the inlet of the SCR and the carbon content of the fly ash at the pre-exhaust outlet are in a lower range,which is in line with the actual situation.