NO_x Emission Of Coal-fired Boiler Model And Optimization Based On Intelligence Algorithm

With the development of economy, more and more serious environmental pollution.Coal-fired power generating is still the main power generation equipment, introduced in 2011 "power plant air pollution emission standard" requirement coal-fired power plants run more efficient and low pollution.Coal-fired are mounted device in China recent years, denitration equipment make the NOx emission value is lower than the national standard, so inevitably affects efficiency of coal-fired power plants, thus for boiler combustion optimization operation is still worth studying.Making mixed coal blending hot test on a 660 MW coal-fired power plant boiler, select three conditions which are the plant always running, in accordance with the laboratory efficiencyand pollution emission test of the coal blending plan.Experimental results show that under the 660 MW load, adopting scheme two ways of coal boiler in variable oxygen test boiler thermal efficiency is highest 93.57% when the oxygen content was 2.5%;Scheme is applied in 600 MW load, four ways of coal boiler in variable oxygen test boiler thermal efficiency is highest 93.68% when the oxygen content was 2.0%;Scheme is applied in 550 MW load under six way of coal, the boiler in variable oxygen test boiler thermal efficiency is highest 93.60% when the oxygen content was 2.5%.Boiler NOx emission concentration increased with the increase of oxygen content increases, therefore in the process of operation should guarantee the stable oxygen in oxygen best value to make the boiler operate high-efficiency and low pollution.Based on thermal state test, using BP neural network model boiler emission characteristics, good results are obtained. Network is good at mapping relationship between input and output, NOx emission model of the average relative error is 0.73%, the biggest relative error in 9 samples, maximum relative error is 4.6%.The relative error of three test samples were 0.46%, 0.59% and 2.34%,it’s average relative error was 1.13%.Boiler efficiency average error is 0.13%, the network of the average relative error was 0.4%.Combined with genetic algorithm to optimize the established network model, the optimized model accuracy and generalization ability have improved. after optimization Average net error is 0.18%, the optimized network is greatly reduced from 0.73% before optimization, the relative error of calibration samples were 0.39%, 0.51%, 0.8%, average error is 0.57.optimization results show that the genetic algorithm optimize the initial weights of BP network training is effective and can improve the precision and generalization ability of network. On the basis of the establishment, optimizing the boiler NOx emissions, the habit of working conditions before NOx concentration is 458.4mg/m3, NOx concentration is 329.7 mg/m3 after adoptingthe optimized operation, reduced by 28%, the effect is obvious. The optimizedoperation mode can reflect fuel boiler combustion classification and grade of air distribution, thus inhibiting the formation of NOx.