Combustion Properties Of Power Coal Blends Characterized With Nonlinear Theories

Power coal blending is a kind of clean coal technology which meets our national conditions, it helps improve the coal heat energy utilization rate, strengthen the security and reliability of the coal combustion equipment and lower the emission of pollutants during the coal combustion process. In this paper, thermgravimetric analysis indicates that the relationship between the combustion characteristic and coal properties of the power blending coal shows strong nonlinear characteristic, tibe early-stop technology is applied to prevent the neural network model from over fitting and the combustion characteristic of the power blending coal is precisely predicted, and the fractal theory is used to analyzes the fractal structure of the coal coke pores which is formed during coal pyrolysis and the coal ignition characteristic. A nonlinear multivariate optimization power coal blending expert system is developed based on the conditions in Zaozhuang coal yard and is applied in the demonstration project successfully.The Early-stopping back-propagation (ESBP) neural network precisely predicted the ignition temperature and activation energy of 16 kinds of typical power blending coal and 48 kinds of blending coal, the early-stopping technology is used to solve the over-fitting problem of the training data when applying neural network, and this helps improve the network generalization ability and the accuracy of predicting new data obviously. Three-layer neural network model is built based on the coal quality analysis, the average errors of the ignition temperature and activation energy predicted by the neural network are 0.29% and 1.28% respectively, which are far lower the average errors of 2.24% and 5.91% by applying second nonlinear regression method.The ESBP neural network precisely predicted the maximum combustion rate and fixed carbon burnout rate of 16 kinds of typical power blending coal and 48 kinds of blending coal, the early-stopping technology is used to solve the over-fitting problem of the training data when applying neural network, and this helps improve the network generalization ability and the accuracy of predicting new data obviously. Three-layer neural network model is built based on the coal quality analysis, the average errors of the maximum coal combustion rate and fixed carbon burnout rate predicted by the neural network are 1.97% and 0.91% respectively, which are far lower the average errors of 7.06% and 4.03% by applying second nonlinear regression method.Analysis has been done to look into the coal coke pore fractal structure formed during the coal pyrolysis process and the ignition characteristic. When the total content of the volatile and moisture in the air-dried based raw coal increases from 15.22% to 39.49%, the activation energy of pulverized coal pyrolysis will decrease leading to an increasing of the coal coke pore fractal dimension from 2.30 to 2.84, the peak differential specific surface area increases when the pore diameter is 3.7 nm, however the average particle pore diameter decreases at the same time. Accordingly, the coal coke ignition temperature decreases from 617℃ to 486 ℃, and the coke combustion activation energy decreases which increases the fixed carbon burnout rate from 84% to 91%.The characteristics of the hydrothermal catalytic reduction reaction of CO2 in the coal-fired flue gas are researched, and thermodynamics analysis on the four possible ways of CO2 hydrothermal reduction is conducted only to find out that the easiest way to go is the reaction: HCO3-+2Hâ‰'HCOO-+H2O. The experiment indicates that Cu shows a better catalytic characteristic than Ni does in the CO2 hydrothermal reduction reaction. When 16mmol Cu and Al mixture powder is added to the solution, the reduction product formic acid has a concentration of 6694ppm, and the CO2 conversion efficiency in the hydrothermal reduction reaction reaches 29.1%.The one dimensional furnace combustion characteristic of the nonlinear multivariate optimization power coal blending in the power plant is studied. Five kinds of blending coal indexes are identified according to the designed coal rank of the the boilers in state grid Hebei Longshan power plant and the real coal source, a optimized coal blending scheme is drawn by the calculation done by the nonlinear multivariate optimization coal blending model. The one dimensional furnace is used to do the combustion characteristic experiment of optimized blending coal, and the differential thermal difference method is applied to get the ignition temperature of the blending coal in sedimentation furnace test. The ignition distance of the main coals (Xiyang coal and Guoyang coal) is decreased by optimizing the blending coal, the ignition temperature is lowered and the emission of SO2 and NOx is reduced as well.A nonlinear multivariate optimization expert system is developed based on the conditions of Zaozhuang coal yard. ESBP neural network is used to predict the ignition and burnout characteristics of the power blending coal and a multivariate optimized power coal blending calculation model is built, a method that can calculate the blending coal comprehensive evaluation index based on multiple coal quality index weight is put forward. The nonlinear multivariate optimization power coal blending core calculation program is realized, with the design of coal yard store sales management module, a computer expert system software is developed and is applied to the 2 million ton/year power coal blending demonstrative project in Zaozhuang successfully. The combustion test of the optimized power coal blending scheme method calculated by the expert system is conducted on a 6t/h chain furnace, according to the test by a third party, the thermal efficiency of the chain furnace is improved from 64.9% to 71.4%, with a coal saving rate of 10%.