High-fidelity Simulation Of Devolatilization And Combustion Of Coal And Biomass

China’s unique energy reservoir structure determines that in the foreseeable future,China’s energy structure will always be dominated by coal.In China,most of the coal is used for power and heat generation in coal-fired power plants.However,a large number of pollutants,such as NO_x,SO_x,heavy metal mercury and dust particles,are produced in the process of coal com-bustion,which are the important reasons for the formation of photo-chemical smog,acid rain and haze.Therefore,the development of high-efficiency and clean utilization technology of pulverized coal is a research hotspot in the field of energy science.At present,there are some advanced technologies of high-efficiency and clean utilization of pulverized coal,such as the technology of coal classification and multi-generation,oxy-coal combustion and biomass/coal co-firing.While pulverized coal combustion is a very complex physical and chemical process,in-cluding volatile release,char combustion and complex gas-phase combustion,involving cou-pling and interaction of many multi-scale and multi-physical basic processes,such as complex gas-solid two-phase heat transfer,mass transfer,momentum exchange,phase change,turbu-lence,chemical reaction and interaction between turbulence and chemical reaction.In order to develop a new clean coal combustion technology,optimize and control the existing coal com-bustion system,it is of great significance to study the unsteady multi-scale physical interaction in different clean coal utilization technologies.Based on those backgrounds,this paper has carried out in-depth research on several key issues with machine learning and high-fidelity nu-merical simulation method,such as the developing devolatilization models for pulverized coal and biomass,exploring the gas-solid two-phase turbulent combustion characteristics in oxy-coal combustion and biomass/coal co-firing,the formation mechanism of pollutants and the modelling of interaction between turbulence and chemical reaction.Firstly,a general model which can accurately predict the devolatilization rate and compo-nent distribution of coal devolatilization is developed by using the artificial neural network.The model uses a two-step kinetic framework,but allows its kinetic parameters to be dynamically adjusted with the change of coal type and heating conditions.The regulation of adjustment is learned by artificial neural network from the widely collected experimental data and the predic-tion results of complex pyrolysis model,such as Chemical Percolation Devolatilization（CPD）model.The model can well reproduce the experimental data and the prediction results of the complex model,but the calculation cost is only 1/4 of the complex network devolatilization model,and only 15%more than the traditional two-step model.At the same time,the method is extended to biomass devolatilization,and the artificial intelligence models of biomass de-volatilization are developed.These models can well reproduce the devolatilization behaviour of a wide range of biomass types under wide ranges of heating rates,and have obvious ad-vantages over the traditional polynomial fitting model.In addition,this paper also developed artificial intelligence models which can predict biomass higher heat value and chemical compo-sition.For the prediction of the three main chemical components of biomass,the random forest model developed in this paper can give a better prediction for a wide range of biomass types,and the relative error of prediction is less than 20%.For the prediction of biomass higher heat value,whether based on element analysis or proximate analysis,the artificial neural network and random forest model developed in this paper can give much better prediction results than the traditional polynomial fitting model,especially the performance of random forest model in the training and validation data base is the best.It is worth noting that when using the results of element analysis as training data,even if there are few samples,the accuracy of the trained models is higher than that of proximate analysis.At the same time,based on the developed biomass devolatilization model and sensitivity analysis,the phase diagram of biomass direc-tional devolatilization in full parameter space is obtained,which provides effective guidance for the directional selecting biomass types and changing reactor operating conditions to achieve the biomass directional devolatilization.Then,some basic issues of pulverized coal combustion,pulverized oxy-coal combustion and coal/biomass co-firing（CBCF）are studied by direct numerical simulation（DNS）.For pul-verized coal combustion,this paper first studies the influence of gas temperature fluctuation on the devolatilization of pulverized coal particles by directly coupling the CPD model.The re-sults show that the gas temperature fluctuation can significantly enhance the devolatilization of particles,and the smaller particles are more sensitive to gas temperature fluctuation compared with large particles.Then,DNS study of laminar flow retarded pulverized coal flame is carried out.Two common devolatilization models,three devolatilization components hypotheses and three reaction mechanisms are compared and evaluated.The results show that when the single-step model fitted from the CPD model is used,the devolatilization products are assumed to be a mixture of light gas or heavy hydrocarbon,and the corresponding detailed mechanism is used for their oxidization,the best agreement with the experimental data can be obtained;assuming the devolatilization product as methane is unreasonable,even through the detailed mechanism is used to describe its combustion,and the gas phase temperature and particle properties are overestimated;the simple two-step mechanism can not reproduce the prediction results of de-tailed mechanism,and the predicted flame shape and combustion characteristics show obvious errors;the devolatilization model also has an important impact on the pollutants formation.For the oxy-coal combustion,this paper uses DNS coupled with detailed mechanism to explore the influence of oxygen concentration on the oxy-coal ignition,gas and particle,and pollutant for-mation characteristics,and also to explore how the oxygen concentration changes the path of NO formation.The results show that the increase of oxygen concentration can obviously ad-vance ignition,increase combustion intensity,gas temperature,heat release rate and CO mass fraction,and reduce CO₂mass fraction,which increases the difficulty of CCS（Carbon capture and store）;with the increase of oxygen concentration,the devolatilization and char combustion are also enhanced;corresponding to the increase of combustion intensity under high oxygen concentration,the NO mass fraction also increases.For CBCF flames,the combustion charac-teristics and pollutant formation of CBCF jet flames with different blending ratios are studied with DNS coupled with detailed mechanism.The results show that CBCF can significantly prompt ignition,and enhance the devolatilization and char combustion when the blending ratio is 20%and 40%,while when the blending ratio is 50%,char combustion is reduced due to the local lacking of oxygen and cluster effect of large particles.In addition,with the increasing of the blending ratio,the NO production routes are obviously reduced,while the NO consumption routes are slightly affected,therefore resulting in an obvious reduction of NO emission.Finally,the coal flamelet model is extended to the oxy-coal condition,and the large eddy simulation of Cambridge coal flames,one of the target flames in Coal and Biomass Conversion Workshop（CBC Workshop）is carried out to evaluate the model performance.The results show that the large eddy simulation can well reproduce the flow and combustion fields.This work lays a solid foundation for its application to the realistic industrial-scale oxy-coal burners.