Study On Online Measuring And Control Of PCDD/F From Organic Waste Incineration

With the increasing proportion of incineration for the disposal of organic waste and the increasingly stringent polychlorinated dibenzo-p-dioxins and-furans(PCDD/F)emission legislation limits,PCDD/F emissions from organic waste incineration are the focus of attention of the government,the publics,and industrial managers.The current measurement and control of PCDD/F emission are far away from the requirement of the government and the public.Therefore,the development of online measurement and feedback control for PCDD/F is imperative.To study correlation mechanism between PCDD/F emission and parameters under various operations and air control pollution system,to make up for the deficiency of accurate and applicable predicting model of dioxin emissions,and to address the problem of the feedback control of dioxins and industrial application,in this thesis,the prediction and control of dioxin emissions from organic waste incineration are systematically studied.The main contents are listed as follows:Firstly,to address the problem that the correlation between dioxins,chlorobenzene,and operating parameters is ambiguous under different operating conditions in municipal solid waste incineration,the formation mechanism of dioxins and correlation between dioxins and chlorobenzene and macropollutant under transient operations and normal operation are studied at the level of individual PCDD/F homologs.Combined with statistical analysis and characteristic PCDD/F homologs signal intensity,the dominant formation mechanism of dioxins and the influence of operating conditions on the correlation between PCDD/F and parameters are revealed.For the vague influence of the air pollutants control devices of hazardous waste incinerators on the correlation between PCDD/F and chlorobenzenes,the levels of PCDD/F,heavy metals,and chlorobenzene along with the air pollutants control devices are comprehensively studied.Combined with statistical analysis and PCDD/F characteristic homolog signal intensity.Furthermore,the formation mechanism and migration mechanism of PCDD/F,the key parameters affecting PCDD/F emissions,and the influence of air pollutants control devices on the correlation between PCDD/F and operating parameters were revealed.All in all,combined with the association mining of formation mechanism analysis,the key parameters affecting dioxin emission are mined to support the construction of credible PCDD/F data sets and the development of dioxin emission prediction and control methods.Then,based on the key factors affecting the formation of PCDD/F,credible datasets of PCDD/F emissions and key parameters are constructed through long-term measurement and data collection.For the long-term operation of the municipal solid waste incinerator,it is difficult to achieve accurate on-line measurement of PCDD/F emissions.Combined with the chlorobenzene measured by Tunable Laser Ionization coupled with Time-of-Flight Mass Spectrometry(TLI-TOFMS),based on the cluster analysis between the signal value of the characteristic homolog contribution of the specific formation path and the PCDD/F emission,the hybrid modeling method of formation path recognition clustering and Box-Cox transformation is proposed.Compared with linear regression,the prediction performance is significantly improved,and the average absolute error decreased by 42.6%.To solve the problem that the current model for predicting PCDD/F emissions can only be applied to specific incinerators due to high deviations or systematic errors,based on the data set of operation parameters and PCDD/F emissions,the proposed Bayesian support vector regression method achieves accurate prediction of PCDD/F emissions.The performance of Bayesian support vector regression is significantly better than that of ridge regression,multiple regression,and univariate regression.To aim at the online prediction of PCDD/F emissions from hazardous waste incineration and the online prediction of PCDD/F homolog concentration,based on the data set of operating parameters,chlorobenzene,and PCDD/F,a sequence feature selection-hybrid discriminative model is proposed to realize the accurate prediction of PCDD/F emissions and PCDD/F homolog concentrations.Significant supports for the general application of PCDD/F online measurement are provided.Furthermore,based on the online PCDD/F measured by TLI-TOFMS and best discriminative model,the large data sets of PCDD/F and operating parameters for grate furnace incineration and fluidized bed incineration are constructed.The methods of Bayesian optimized e Xtreme Gradient Boosting(XGBoost),Light Gradient Boosting Machine(Light GBM),and Cat Boost are studied and constructed to accurately predict(R~2=0.992,RMSE=0.031)the PCDD/F emissions based on easy-to-measure incineration parameters on the grate furnace incineration and fluidized bed incineration data sets.The generalization performance the proposed method is significantly better than that of bidirectional long-term and short-term memory neural network and convolutional neural network.The divide-and-conquer Boosting method has excellent fitting generalization performance and applicability,and significantly reduces the cost of PCDD/F n emission prediction.Then the best prediction model is explained to reveal the importance,correlation,and interactivity of inputs,to guide PCDD/F feedback control and emission reduction.Finally,to clear complex formation mechanism of PCDD/F in complex systems,to explain the black box characteristics of model prediction and to establish PCDD/F feedback control strategy,based on the constructed optimal discriminant model and through Shapley additive explanation(SHAP),Local Interpretable Model-Agnostic Explanations(LIME)and Partial Dependence Plots(PDP)methods,the importance of the features is explained,and the mechanism of characteristics affecting PCDD/F emissions,homolog concentration and the interaction between input features are revealed.Thus,the optimal quantitative operating conditions of incineration parameters were constructed to achieve the lowest PCDD/F emissions.Aiming at the high PCDD/F emission under the start-up operation,combined with the activated carbon adsorption PCDD/F model and the equation for PCDD/F and time,the PCDD/F emission under the start-up operation can meet the national legislation limit.Finally,the feedback control application of PCDD/F emission is realized in large-scale incinerators.By adjusting the parameters such as incinerator temperature and pressure,the PCDD/F emissions can be reduced to the national legislation limit during long-term operation.