Accurate Ammonia Spraying Technology Based On NO_x Concentration Field Prediction By

Selective catalytic reduction(SCR)is widely used in coal-fired power plants as an efficient nitrogen oxide control method.Recently,coal-fired power plants have put forward higher requirements for NOx emission limits.SCR system has exposed problems such as large ammonia escape,low denitration efficiency,and difficulty in stably controlling NOx concentration at SCR outlet below the set value during off-design operation.After analyzing,the main reason for the above problems lies in the uneven distribution of NOx concentration in the flue.The conventional SCR uniform ammonia spraying strategy is difficult to accurately match the NOx concentration distribution,which results in large differences in ammonia-nitrogen molar ratio in different regions.At the same time,because of the SCR control system has hysteresis quality,the ammonia spraying amount is difficult to follow the change of NOx concentration in real time.In order to solve the above problems,an accurate ammonia spraying technology based on NOx concentration field prediction is proposed in this paper.Firstly,taking the cold-state SCR flue gas denitration test bench in the laboratory as the object,this paper optimizes the zoning strategy by means of numerical simulation.The details are as follows: The flow field distribution in the flue is analyzed,and the Ammonia Injection Grid(AIG)is optimally divided into 10 zones according to the flow field distribution.According to five typical NOx concentration distribution conditions,the ammonia spraying amount in each zone is adjusted,and the ammonia-nitrogen molar ratio at the inlet of the catalyst layer is optimized,so that the distribution standard deviation coefficient is less than5%.The standard deviation coefficient of the measured ammonia-nitrogen molar ratio distribution under the simulated zoning ammonia spraying rate for these five typical working conditions is less than 6.5%.The accuracy of the simulation model and the effectiveness of the optimized zoning ammonia spraying are verified.Secondly,from the previous experiments,it can be seen that it is difficult to adjust the amount of zoning ammonia spraying in different zones through simulation to meet the requirement of adjusting the ammonia-nitrogen molar ratio and making the distribution deviation less than 5% when NOx distribution changes.In order to achieve accurate ammonia spraying under different NOx concentration distributions,based on the experimental bench simulation model,a series of different AIG zoning ammonia spraying amount combinations is calculated through simulation,and corresponding NH3 concentration at 25 measuring points is obtained,and then a neural network calculation model with NH3 concentration at 25 measuring points as input parameters and 10 zoning ammonia spraying amount as output parameters are established.The trained BP neural network is used to calculate the zoning ammonia spraying amount under the preset working conditions.The amount of zoning ammonia spraying is input into the simulation.It is found that under different NOx concentration distributions and standard deviation coefficients,the ammonia-nitrogen ratio standard deviation coefficients of the first 25 measuring points at the inlet of the first layer catalyst are below 3.6%,meeting the standard requirement of less than 5%.This method can be used to accurately and quickly calculate the zoning ammonia spraying amount under different NOx concentration distributions.Finally,for the hysteresis of SCR control system,by setting the corresponding relationship between the coal feeding rate of the coal mill and the NOx concentration distribution at the SCR inlet,GA-BP neural network is used to establish the NOx distribution prediction model based on different coal feeding rates,which is used as the feedforward adjustment of the control system to reduce the hysteresis of the control system.It is verified that the average relative error of the model is less than 2% under different NOx concentration distribution characteristics.After adding different random deviations to the test data,the prediction models all ensure the stability of the output results.