Modeling And Optimization Of The Combustion Characteristics In Low NO_x Combustion Pulverized Coal-fired Boilers Under Coal Blending Conditions

Combustion optimization is an effective approach in improving the combustion efficiency and reducing pollutant emissions of pulverized coal-fired boilers.In this study,modeling and optimization of the combustion characteristics in coal-fired boilers based on data mining and numerical simulation were studied.A coal property prediction model and low NOx emission combustion optimization method based on data mining were developed.A numerical model was developed for modeling the combustion of deep air-staging boilers.In addition,the combustion modeling and optimization methods were applied in two coal-fired utility boilers with different capacity.Considering that the measurement of coal calorific value may not always be accessible in coal-fired plant.A total of more than 4000 Chinese and U.S coal samples were employed to study the correlation between the proximate analysis and calorific value.A calorific value model was developed using the nonlinear support vector machine.The model significantly improves the prediction accuracy of the calorific value compared with the traditional multiple linear regression models.The ignition and burnout characteristics of coal is also important for the operation of boilers.A total of 46 coals were collected and analyzed to study the correlation between the coal basic analysis and ignition,burnout characteristics.,A coal ignition temperature and burnout temperature were established by combining the sequence backward search algorithm and support vector machine.The average relative error of the developed model in predicting ignition temperature and burnout temperature is only 3.02%and 2.93%.The model provides a foundation for modeling and optimization of the combustion under coal blending conditions.Historical operational data contains a wealth of knowledge.The combustion modeling and optimization approach based on data mining was studied by taking the low NOx combustion optimization of a 700MW boiler as an example.A combustion optimization approach combining the extreme learning machine and harmony search algorithm was proposed.A NOx emission prediction model was developed based on the extreme learning machine.The relative error of the NOx prediction model is only 1.4%,which is smaller than that based on the traditional neural network modeling method,and is comparable with that based on support vector machine.While the modeling response is far superior to the neural network and support vector machine.Based on the NOx prediction model,two typical cases were selected to be optimized using the harmony search algorithm.The NOx emissions of the two cases reduced 16.5%and 19.3%respectively according to the optimization.The effectiveness of the optimization results was verified by field tests.For the low oxygen concentration and relatively high carbon dioxide,water vapor concentration in primary combustion zone of low NOx combustion boiler,the adaptability of different numerical combustion models was studied.It is found that the combustion model without char gasification reaction is difficult to accurately predict the CO emission at the outlet of the furnace.The char combustion rate in the primary combustion zone is small and the char combustion concentrates at the upper part of the furnace,leading to the low burnout rate.The NO distribution in furnace is also different from the actual situation.The combustion model containing char gasification reactions showed better agreement with the field test.Based on the established numerical model,the combustion characteristics of the low NOx combustion boiler were studied.The CO concentration is about twice as high as that before modification.The peaks of char burning rate transferred from the elevation of AA and CCOFA to the elevation of CCOFA and SOFA.With the increase of operating oxygen,CO emissions and unburned carbon in fly ash decreases significantly,NOx emissions rises slightly.Reducing the proportion of over-fire air is beneficial to improve the combustion efficiency,but the NOx emissions increases dramatically.The combustion modeling and optimization method based on data mining were applied to optimize the operation of a 660MW low NOx combustion boiler.The ignition temperature and burnout temperature of the used coals are calculated by using the established ignition temperature and burnout temperature prediction model.Subsequently,a NOx emission,CO emission and unburned carbon in fly ash prediction model is developed based on the extreme learning machine,which accurately predict the NOx emission and the trend of CO and unburned carbon in fly ash under different operation conditions.The power generation cost significantly decreases according to the optimization result by the harmony search algorithm.The developed numerical simulation method was applied to study the combustion characteristics of co-firing sludge in a 100MW boiler,and the blending ratio of sludge and the water content of sludge were synergistically optimized.It was found that when the blending ratio of sludge is over 10%,the primary combustion zone of the wall heat flux and ignition characteristics decreased significantly.Reducing the moisture content of the sludge is beneficial to improve the flue gas temperature,ignition characteristics,wall heat flux,flame stability and burnout rate,but the improvement is limited.Considering both the technical and the economic analysis,co-firing 10%sludge at the moisture content ranging from 40%to 56%is an optimal coal-sludge co-combustion solution for the coal-fired power plant.