Study on NOx emission estimation method of chain furnace based on EDMS model

In order to overcome the problems of low estimation accuracy and large relative error of traditional NOx emission estimation methods, a new NOx emission estimation method of chain furnace based on EDMS model is proposed in this paper. Firstly, the overall structure of the chain furnace is analyzed, and the EDMS model is used to count the NOx emission data of the chain furnace. Secondly, SVM model is used to train the emission sample data obtained from the above statistics. Finally, the NOx emission estimation objective function of chain furnace is constructed to complete the emission estimation. The experimental results show that compared with the traditional emission estimation methods, the estimation accuracy of this method is higher, and the relative error of estimation can be reduced.


Introduction
Under the background of rapid social and economic development, industrial economy has also developed rapidly.In this context, higher requirements are put forward for the performance of industrial processing, especially coal-fired boilers [1] .Although China's clean energy power generation has been effectively developed, coal-fired boiler is still the main way of application in industrial processing.Chain furnace has been widely used in industrial processing because of its good energy-saving and combustion performance [2][3][4] .In the environment of sustainable development, higher requirements are put forward for the emission of combustion pollutants from industrial boilers.In the combustion process of industrial boilers, nitrogen oxides (NOx) are the main pollutants emitted, which will produce greenhouse effect and cause smog.In order to reasonably control the pollutant emission of industrial chain furnace, it is necessary to accurately estimate the NOx emission of chain furnace [5] .Reference [6] proposes a NOx emission estimation method based on maximum information coefficient variable selection.Firstly, this method expounds the basic concept of variable selection method, and introduces the maximum information coefficient to improve the variable selection method.The improved variable selection method is used to construct the NOx emission model, and the output of the model is the emission estimation result.However, there is a big gap between the estimated results and the actual results.Reference [7] proposes a NOx emission estimation method based on multi model clustering integration, divides NOx emission data into high-dimensional data and low-dimensional data, and uses the combination of variable weight and information entropy to determine the clustering parameters of different dimensional data.After completing the clustering processing of NOx emission data, the multi model clustering integration algorithm is used to construct the membership matrix of emission estimation, judge the NOx emission threshold and complete the NOx emission estimation.However, this method has the problem of low accuracy of emission estimation.Reference [8] proposes a NOx emission estimation method based on generalized regression neural network.This method is based on the historical data of boiler NOx emission, uses generalized regression neural network to construct NOx emission estimation model, and uses genetic algorithm to solve the estimation model to obtain NOx emission estimation results.However, the estimation result of this method has the problem of large relative error.In order to solve the problems of low estimation accuracy and poor estimation relative error of the above traditional emission estimation methods, a NOx emission estimation method of chain furnace based on EDMS model is proposed.

Statistics of NOx emission data of chain furnace based on EDMS model
Chain furnace is a kind of front feeding furnace.The coal combustion process is completed in movement.It has stable combustion conditions, high thermal efficiency, convenient operation, low labor intensity and low smoke emission concentration.The structure of chain furnace is shown in Figure 1.
Fig. 1 Structure of chain furnace EDMS model is a model developed abroad for gas pollutant emission inventory and emission statistics, which is composed of emission model and diffusion model [9] .The calculation formula of NOx emission of chain furnace in one week under EDMS model is: In the formula, u represents the type of NOx pollutants, v represents the working mode of chain furnace, and , u v E represents the total amount of NOx pollutants.The calculation formula of total NOx pollutants is: In the formula, p represents the type of chain furnace, q represents the power type of chain furnace, p n represents the number of air chambers of chain furnace, , p q s represents the number of cycles of chain furnace, , , p q v F represents the coal consumption rate, , , , p q u v E represents the emission factor, and , v p t represents the working time of chain furnace [10] .
Since the NOx emitted from the chain furnace will show a diffusion trend with the flow of air, it is necessary to calculate the diffusion concentration of NOx pollutants from the chain furnace: In the formula, O represents the emission of NOx pollutants from the chain furnace, u represents the average value of the wind speed at that time, zs  represents the vertical diffusion concentration of NOx pollutants, es h represents the height of the flue outlet of the chain furnace, ieff z represents the diffusion height of NOx pollutants, and y F represents the lateral diffusion distribution parameters of NOx pollutants from the chain furnace [11,12] .In order to obtain accurate NOx emission data of chain furnace, the above calculation results are combined with the diffusion concentration results to make statistics on the total NOx emission of chain furnace: ( , , ) In the formula, t F represents the load factor of coal combustion of chain furnace, t U represents one working cycle of chain furnace, and it I represents the emission factor of NOx pollutants.The above data are used to estimate the NOx emission of chain furnace, so as to obtain high-precision estimation results.

Estimation of NOx emission from chain furnace
Based on the above statistical total NOx emission data, SVM model is used to train the samples to obtain the emission estimation results.In the SVM model, i x is defined to represent the input NOx emission sample data, and represents the sample space, and i y represents the output result [13] .The key of SVM model training is to construct the nonlinear mapping function between input space and output space, so as to get the final output result.The expression of linear regression function is: (5) In the formula,  represents the weight in the input space, ( ) represents the nonlinear mapping function, and b represents the threshold of the linear regression function.If the loss parameter of SVM model is defined as  , the insensitive loss function of SVM model can be: When the NOx emission sample of the chain furnace is outside the area, the error between the model output result and the actual estimation result can be ignored.At this time, the gap between the emission estimation can be punished [14,15] .Build empirical risk function: 1 ( ) Aiming at minimizing the error of emission estimation, the constraint optimization function of emission estimation is constructed: In the formula,  and *  represent the upper and lower limits of the output error of the model, and there is * , 0 The kernel function is introduced to transform the objective function into a dual solution problem: . ., 0, 0 , Construct regression estimation function under kernel function: In the formula, ( , ) represents the kernel function.
It is assumed that the data used in this emission estimation is the data of recent m years, and the number of emission indicators each year is n .The objective function of NOx emission estimation is constructed with the objective of minimizing the relative root mean square error of NOx emission estimation of chain furnace: In the formula, ( ) f x represents the estimated NOx emission of chain furnace under parameters , , Through the above calculation, the estimation of NOx emission of chain furnace is completed, and the effectiveness of the estimation results will be further verified in the subsequent contents.

Experimental verification
In order to verify the actual estimation effect of the proposed NOx emission estimation method of chain furnace based on EDMS model, simulation comparative verification experiments are carried out.Before the experiment, in order to ensure the effectiveness of the experimental results, the main parameters of the chain furnace are set first.The main parameters of chain furnace are shown in Table 1.After setting the parameters of the chain furnace, select the operation data of the chain furnace in a factory in our city.The chain furnace operates continuously for 24 hours to estimate the NOx emission generated by the chain furnace.Set the overall experimental scheme: take the estimation accuracy and relative error of NOx emission as the experimental comparison index, and compare and verify the method in this paper with the methods in reference [6] and reference [7] .NOx emission estimation accuracy: NOx emission estimation accuracy refers to the consistency between the NOx emission estimation results of different methods and the actual emission results.The higher the NOx emission estimation accuracy, the stronger the effectiveness of the estimation method.The calculation formula of relative error of NOx emission estimation is: In the formula, q represents the estimated value and Q represents the real value.

NOx emission estimation accuracy
As the key index for the evaluation of the estimation method, the NOx emission estimation accuracy can effectively prove the estimation performance of the NOx emission estimation method in this paper, and compare this method with the traditional reference [6] and reference [7] .The comparison results of NOx emission estimation accuracy of the three methods are shown in Figure 2.
Fig. 2 Comparison results of NOx emission estimation accuracy From the comparison results of NOx emission estimation accuracy shown in Figure 2, it can be seen that the NOx emission estimation results of this method are basically consistent with the actual emission results, reaching more than 99%, while the NOx emission estimation results of reference [6] and reference [7] methods are quite different from the actual emission results, especially the results of reference [7] , in the early and late stages of the experiment, the difference in emission results is even doubled.Therefore, the above experimental results show that this method can accurately estimate the NOx emission of chain furnace.

Relative error of NOx emission estimation
The relative error of NOx emission estimation can also reflect the performance of different emission estimation methods.The smaller the relative error, the stronger the effectiveness of the estimation results.If the relative error is too large, the actual availability of the method will be reduced.Therefore, the relative error of NOx emission estimation is selected as the experimental index, and the method in this paper is compared with the methods in reference [6] and reference [7] .The comparison results of relative error of NOx emission estimation of the three methods are shown in Figure 3.
[1 Observe the comparison results of the relative error of NOx emission estimation shown in Figure 2. The relative error results of the method in this paper basically remain within the range of ±0.01 and can be basically ignored.The relative error between reference [6] and reference [7] is large.The maximum relative error of reference [6] method is -0.12 and that of reference [7] method is -0.065.Therefore, this method can reduce the relative error of NOx emission estimation.

Conclusion
In order to improve the effectiveness of NOx emission estimation of chain furnace, a NOx emission estimation method of chain furnace based on EDMS model is proposed, and the performance of the method is verified from both theory and experiment.This method has high emission estimation accuracy and low relative error in NOx emission estimation.Specifically, compared with the method based on multi model clustering integration, the estimation accuracy of NOx emission of chain furnace in this method is significantly improved, which is basically more than 99%; Compared with the method based on generalized regression neural network, the relative error of emission estimation in this method is significantly reduced.Therefore, it fully shows that the proposed emission estimation method based on EDMS model can better meet the requirements of NOx emission estimation of chain furnace.

Fig. 3
Fig. 3 Comparison results of relative error of NOx emission estimation

Table 1
Main parameters of chain furnace