Nitrogen oxide prediction method in urban solid waste incineration process

Abstract

The invention provides a method for predicting nitrogen oxides in an urban solid waste incineration process, and relates to the technical field of urban solid waste incineration, and the method comprises the following steps: determining a combustion area and a temperature interval in an incinerator based on an MSWI process; based on the combustion area and the temperature interval, a pollutant emission data set is determined according to equipment parameters and operation parameters of the incinerator; according to the pollutant emission data set and the manipulated variables, a virtual numerical simulation mechanism data set under multiple working conditions is obtained through an orthogonal experiment; and obtaining a nitrogen oxide prediction result according to the virtual numerical simulation mechanism data set and a nitrogen oxide prediction model. The problem that in the prior art, the NOx generation concentration cannot be accurately predicted through the air distribution and material distribution variable is solved.

Description

Technical Field

[0001] The invention relates to the technical field of municipal solid waste incineration, and in particular to a method for predicting nitrogen oxides in a municipal solid waste incineration process. Background Art

[0002] The global total amount and per capita production of municipal solid waste (MSW) are increasing every year with the continuous economic growth. According to statistics, the total amount of MSW in 2020 has reached 235 million tons. If it is not properly disposed of, it is estimated that the growth rate of MSW in 2050 will be twice the growth rate of the global population. Landfill treatment can no longer meet the demand. MSW incineration (MSWI) with the advantages of harmlessness, reduction and resource utilization has become the method currently promoted in my country. The MSWI process will inevitably generate some pollutants, such as NOX, SO2, HCl, CO2 and other gases. Among them, NOx under sunlight will react with hydrocarbon compounds to generate photochemical smog, which irritates human eyes and causes conjunctivitis; it also has a significant irritating effect on the nose, throat, trachea and lungs, causing breathing difficulties; in addition, NOx entering the stratosphere will damage the ozone layer, thinning it or even forming holes. At present, mitigating NOx emissions has become an urgent problem that needs to be solved internationally. Research on the formation mechanism and inhibition technology of NOx for the MSWI process has always been a hot issue in industry and academia.

[0003] In current research, NO and NO2 are collectively referred to as NOX, and the NOX generated during the combustion process is divided into thermal, fuel and rapid types, and there are corresponding models in numerical simulation software. Thermal NOx is formed by the oxidation of nitrogen in the combustion air, rapid NOx is formed by the rapid reaction of the flame front, and fuel NOx is formed by the oxidation of nitrogen in the fuel. Since the proportion of rapid NOX generated during the combustion process is very small, most studies only consider the fuel and thermal generation mechanisms. In view of the above reasons, most domestic and foreign researchers currently analyze the combustion process through numerical simulation models to gain insight into the combustion mechanism and reduce pollutant emissions from the perspective of optimized design. There are no reports on the study of the "wind and material distribution" variables on the NOx generation concentration from a control perspective.

[0004] In summary, the existing technology cannot use the "wind and material distribution" variables to accurately predict the NOx generation concentration. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a method for predicting nitrogen oxides in the process of urban solid waste incineration. The present invention solves the problem in the prior art that the "wind and material distribution" variables cannot be used to accurately predict the NOx generation concentration.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for predicting nitrogen oxides in a municipal solid waste incineration process, comprising:

[0008] Based on the MSWI process, determine the combustion area and temperature range in the incinerator;

[0009] Based on the combustion area and temperature range, determine a pollutant emission data set according to equipment parameters and operating parameters of the incinerator;

[0010] According to the pollutant emission data set and the manipulated variables, a virtual numerical simulation mechanism data set under multiple working conditions is obtained by using an orthogonal experiment;

[0011] The nitrogen oxide prediction result is obtained according to the virtual numerical simulation mechanism data set and the nitrogen oxide prediction model.

[0012] Preferably, the manipulated variables include:

[0013] Feed rate, grate speed, primary air flow, primary air drying temperature and primary air combustion temperature.

[0014] Preferably, the determining of the pollutant emission data set based on the combustion area and the temperature range according to the equipment parameters and operation parameters of the incinerator comprises:

[0015] Construct FLIC model and Aspen PLus model;

[0016] Determine the MSW composition, primary air volume and boiler parameters corresponding to the combustion area and each temperature range;

[0017] Inputting the MSW composition, primary air volume and boiler parameters into the FLIC model to obtain an initial output result;

[0018] The initial output results are input into the AspenPLus model to obtain a pollutant emission data set.

[0019] Preferably, the pollutant emission data set includes:

[0020] N2, O2, CO2, H2O, CH4, CO, H2, NO, HCN and NH3.

[0021] Preferably, the nitrogen oxide prediction model includes:

[0022] Feature mapping layer, latent feature extraction layer, feature enhancement layer and incremental learning.

[0023] Preferably, the output of the feature mapping layer is expressed as:

[0024]

[0025] The present invention discloses the following technical effects:

[0026] The present invention provides a method for predicting nitrogen oxides in the process of incineration of urban solid waste. Based on the MSWI process, the combustion area and temperature range in the incinerator are determined; based on the combustion area and temperature range, the pollutant emission data set is determined according to the equipment parameters and operation parameters of the incinerator; according to the pollutant emission data set and the manipulated variables, a virtual numerical simulation mechanism data set under multiple working conditions is obtained by orthogonal experiment; and the nitrogen oxide prediction result is obtained according to the virtual numerical simulation mechanism data set and the nitrogen oxide prediction model. The present invention considers the influence of different manipulated variables on the NOx generation concentration based on the actual situation of the MSWI plant. First, based on the MSWI process, the generation, combustion, reduction and re-combustion mechanism of nitrogen-containing pollutants in the furnace are analyzed, and the area and temperature range are divided; then, according to the actual MSWI plant incinerator equipment parameters, operation parameters and other conditions, a conventional pollutant emission concentration numerical simulation model that meets the actual pollutant emission range is constructed; finally, virtual data is obtained through orthogonal experiments, and a mapping model between manipulated variables and NOx generation concentration is constructed using BHFR, and the NOx generation concentration is accurately predicted using the "wind and material distribution" variables, which provides effective support for the realization of pollution reduction and intelligent control of the MSWI process. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0028] Figure 1 A flow chart of a method for predicting nitrogen oxides in a municipal solid waste incineration process provided by an embodiment of the present invention;

[0029] Figure 2 A schematic diagram of the MSWI process description for NOx generation provided in an embodiment of the present invention;

[0030] Figure 3 A schematic diagram of a solid phase combustion zone and a gas phase combustion zone facing NOx provided in an embodiment of the present invention;

[0031] Figure 4 The high temperature combustion zone NO provided by the embodiment of the present invention X Generate schematic diagrams;

[0032] Figure 5 A schematic diagram of the influence of manipulated variables on NOx provided in an embodiment of the present invention;

[0033] Figure 6 A schematic diagram of a modeling and analysis strategy provided by an embodiment of the present invention;

[0034] Figure 7 AspenPlus model flow chart provided for the embodiment of the present invention;

[0035] Figure 8 A schematic diagram of gas phase combustion results on a grate provided in an embodiment of the present invention;

[0036] Fig. 9 A schematic diagram of the distribution results of N2, HCN and NH3 components on the grate provided in an embodiment of the present invention;

[0037] Fig.10 A schematic diagram of a test set prediction curve provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] like Figure 1 As shown, the present invention provides a method for predicting nitrogen oxides in a municipal solid waste incineration process, comprising:

[0041] Step 100: Determine the combustion area and temperature range in the incinerator based on the MSWI process;

[0042] Specifically, the MSWI process includes multiple process stages such as solid waste storage and transportation, solid waste combustion, waste heat exchange, steam power generation, flue gas purification, and flue gas emission. Taking a grate-type MSWI in Beijing as an example, from the perspective of NOx generation, some of its process flows are as follows: Figure 2 shown.

[0043] Depend on Figure 2It can be seen that the original MSW contains a small amount of nitrogen-containing compounds and various N elements. The feeder pushes the MSW into the incinerator, and the combustible components are dried, burned by combustion grate 1, combustion grate 2 and ember grate. In order to ensure that the harmful gases generated by the solid phase combustion of MSW are completely decomposed in the furnace, the gas phase combustion process must meet the technical requirements of strictly controlling the flue gas temperature above 850°C, the high-temperature flue gas residence time in the furnace exceeding 2s, and ensuring that the flue gas has sufficient turbulence. The high-temperature flue gas generated in the furnace is pumped into the waste heat exchange system by the induced draft fan, and the flue gas exchanges heat with the liquid water in the boiler drum to produce high-temperature steam. After cooling, the temperature of the flue gas at the outlet of the waste heat boiler (i.e., flue gas G1) is lower than 200°C.

[0044] From NO X From the perspective of the generation mechanism, it mainly includes: the solid phase combustion zone that releases the NOx contained in MSW, the gas phase combustion zone that generates a large number of intermediates, the high temperature combustion zone that generates a large amount of NOx, and the SNCR purification zone; the generated NO X Contains fast-acting NO X 、Thermal NO X 、Fuel type NO X , among which fast NO X After being generated, it will decompose rapidly. Therefore, the relevant reactions occur in stages, and the present invention only considers the thermal NO X and fuel-type NO X .

[0045] Further, the NOx generation process is described as follows:

[0046] Figure 3 The left side shows NO in the solid phase combustion zone. X Schematic diagram of the generation process, Figure 3 It can be seen that when MSW is burned in the solid phase, it releases the NO contained in it. X , generating a large amount of nitrogen oxides, HCN, NH3, etc., which are the latter stage of NO X An important intermediate in the removal of NO X In addition, this area will also produce conventional pollutants such as CO2, HCl, SO2, etc.; gas phase combustion such as Figure 3 As shown on the right, the temperature is 300-500℃. The nitrogen oxides produced by solid phase combustion in the grate decompose in different areas, and NO X Reduction reaction, such as CO reducing NO to N2, the final product is a small amount of NO X , N2 and a large number of intermediate products HCN and NH3; the temperature of the high-temperature combustion zone is 500-800℃, which is located between the gas phase combustion zone and the SNCR purification zone. Among them: the intermediate NH3 and HCN formed in the previous stage are decomposed in large quantities in the lean fuel zone and produce fuel-type NO XIn the fuel-rich area, N2 is produced. In addition, solid nitrogen (Char-N) will gradually decompose to produce NO in this high-temperature area. The specific pathways are as follows: Figure 4 As shown; in the MSWI plant, the SNCR purification and denitrification system is usually arranged in the furnace. At the optimal operating temperature of 850℃~1150℃, NH3 can react with NOx in the flue gas, thereby achieving the purpose of reducing NOx emissions. In addition, there is a waste heat recovery device after the SNCR, which cools the flue gas by using a three-stage superheater and a three-stage economizer. The cooling water enters from the economizer, and after contacting with the high-temperature flue gas, the temperature rises and turns into steam and dissipates from the superheater. The flue gas after cooling should be lower than 200℃.

[0047] Step 200: Based on the combustion area and temperature range, determine the pollutant emission data set according to the equipment parameters and operation parameters of the incinerator;

[0048] Step 300: obtaining a virtual numerical simulation mechanism data set under multiple working conditions by using an orthogonal experiment according to the pollutant emission data set and the manipulated variables;

[0049] Step 400: Obtaining a nitrogen oxide prediction result according to the virtual numerical simulation mechanism data set and the nitrogen oxide prediction model.

[0050] Furthermore, the manipulated variables include:

[0051] Feed rate, grate speed, primary air flow, primary air drying temperature and primary air combustion temperature.

[0052] Specifically, the present invention mainly studies the influence of five manipulated variables, namely, feed rate, grate speed, primary air flow rate, primary air drying temperature and primary air combustion temperature, on the NOx generation concentration at the G1 flue gas. The corresponding qualitative influence diagram is shown in FIG. Figure 5 shown.

[0053] like Figure 5 As shown, V feed Indicates the feed rate; T dry Indicates the primary air drying temperature; T burn Indicates the temperature of the primary air combustion section; V grate Indicates the grate speed; Q primary Indicates the primary air flow.

[0054] MSW is dried, dehydrated and incinerated with the entry of primary air. The dried MSW undergoes pyrolysis reaction, gradually decomposing the NOx it contains, and the remaining N elements appear in the form of compounds such as HCN and NH3. As the temperature gradually rises, the compounds begin to react with oxygen with the entry of primary air, and HCN and NH3 are oxidized to NOx. As the combustion process continues to deepen and the temperature continues to rise, some char-N that is difficult to decompose also decomposes and generates a large amount of NO. At the same time, as the combustion process proceeds, reducing gases such as CO generated in the furnace react with the generated NOx to reduce it to N2. Finally, the flue gas enters the purification stage for treatment.

[0055] Furthermore, the pollutant emission data set is determined based on the combustion area and temperature range according to the equipment parameters and operation parameters of the incinerator, including:

[0056] Construct FLIC model and Aspen PLus model;

[0057] Inputting the MSW composition, primary air volume and boiler parameters into the FLIC model to obtain an initial output result;

[0058] The initial output results are input into the AspenPLus model to obtain a pollutant emission data set.

[0059] Specifically, based on the MSWI process and NOx generation mechanism, the following control variables were proposed for the five manipulated variables that affect NOx in the incinerator: feed rate, grate speed, primary air flow, primary air drying temperature, and primary air combustion temperature: Figure 6 The modeling and analysis strategies shown include a benchmark working condition numerical simulation model construction module, a multi-working condition virtual numerical simulation mechanism data generation module based on orthogonal experiments, and a model construction module based on width mixed forest regression.

[0060] Depend on Figure 6 It can be seen that the modeling and analysis strategy is divided into three modules, and its functions are as follows:

[0061] The module for building the numerical simulation model of the benchmark operating conditions includes two parts: FLIC and AspenPLus models. The former is used to simulate the solid-phase combustion in the bed, and the latter is used to simulate the gas-phase combustion process in the furnace and the generation and decomposition process of NOx.

[0062] Virtual numerical simulation mechanism data generation module under multiple working conditions based on orthogonal experiments: virtual numerical simulation mechanism data under multiple working conditions are obtained through orthogonal experimental design and experimental implementation oriented to manipulated variables.

[0063] Model building module based on width hybrid forest regression: This module uses the BHFR algorithm consisting of a feature mapping layer, a potential feature extraction layer, a feature enhancement layer, and an incremental learning layer to build a NOx generation model.

[0064] The model uses a multi-software coupling simulation method to simulate the combustion process in the incinerator, which is divided into the solid phase combustion process on the bed and the gas phase combustion process in the furnace. The main process is: the solid phase combustion process on the bed is numerically simulated using FLIC, and the calculation results are imported into Aspenplus in the form of a file as the boundary condition of the grate bed inlet, and the calculations are performed in sequence and iterated according to the model process until the NOx emission concentration at the boiler outlet G1 tends to be stable. For the entire combustion process, the following simplified assumptions are made: (1) The incinerator is in a stable operating state, the reaction in the furnace can reach equilibrium, the temperature and pressure in each reactor are constant, and the pressure loss and heat loss are not considered; (2) During the incineration process, MSW and air are fully mixed and evenly distributed, and the effect of MSW particle size on combustion is ignored; (3) The main elements of MSW are C, H, O, N, S and Cl, among which: H, O, N, S and Cl are all converted to gas phase; N element is divided into two categories, gaseous N directly participates in the gas phase combustion reaction, and solid N decomposes at the highest temperature and participates in the gas phase reaction; (4) It is assumed that the product does not contain tar with complex components.

[0065] Furthermore, the pollutant emission data set includes:

[0066] N2, O2, CO2, H2O, CH4, CO, H2, NO, HCN and NH3.

[0067] Specifically, solid phase combustion can be divided into four consecutive sub-processes: water evaporation, volatile matter release, volatile matter combustion and char combustion. The mathematical model of each process is described as follows:

[0068] When MSW enters the incinerator from the feeder, it is first heated by radiation from the bed, and moisture is released as it evaporates. The release rate can be expressed as:

[0069]

[0070] In the formula, R evp is the water evaporation reaction rate; S a is the particle surface area; H evp Indicates the heat of evaporation of solid water; T s is the MSW temperature; h s is the convection mass transfer coefficient between solid and gas; C w,s and C w,g are the water concentrations in the solid phase and the gas phase, respectively; Q crThe heat absorbed by the solid for convective and radiative heat transfer is as follows:

[0071]

[0072] In the formula, ε s represents the solid emissivity; σ b is the Boltzmann radiation constant; T g is the gas temperature; T env Indicates ambient temperature; h' s is the convective heat transfer coefficient between solid and gas.

[0073] After the water evaporates, the dry MSW begins to release volatiles at high temperatures. The volatile gases are mainly composed of hydrocarbons (CmHn), CO, CO2, H2, O2 and other trace compounds, as follows:

[0074] MSW→Volatile(C m H n ,CO,CO2,H2O)+Char (3)

[0075] Using a relatively simple and accurate one-step global reaction model, the release rate of MSW volatiles is proportional to the remaining volatiles and temperature, as shown in the formula.

[0076]

[0077] In the formula, R V represents the reaction rate of volatile matter; v ∞ is the final yield of volatile matter; v For volatile matter t The amount released at the time, R is the universal gas constant, A V The pre-exponential factor for the burning rate of volatile matter, E V is the activation energy.

[0078] When the gas released by high-temperature MSW in the furnace mixes with the surrounding air, it burns rapidly. The simplified gaseous combustible volatiles include CmHn and CO, and the combustion reaction is as follows:

[0079] C m H n +(m / 2+n / 4)O2→mCO+n / 2H2O (5)

[0080] CO+1 / 2O2→CO2 (6)

[0081] The burning rate or kinetic rate in relation to temperature is:

[0082]

[0083] The combustion of flammable volatile gases is limited not only by the reaction kinetics, but also by the mixing rate of the flammable gas with the surrounding air. Assuming that the mixing rate in the bed is proportional to the energy loss through the bed, the mixing rate is expressed as:

[0084]

[0085] In the formula, C mix is an empirical constant, C is the mass fraction of the gaseous reactant, D g is the mass diffusion coefficient of the gas, V g is the air velocity, ρ g is the gas density, d p is the particle diameter, φ is the porosity of the bed, and S is the stoichiometric coefficient in the reaction.

[0086] The actual combustion reaction rate of the volatile matter takes the minimum value of the temperature-related kinetic rate and the mixing rate with oxygen, as shown in the formula:

[0087] R=min(R kinetics ,R mix ) (10)

[0088] When volatilization is carried out, MSW particles will form coke, and the main products of further gasification are CO and CO2. The reaction equation can be expressed as:

[0089]

[0090] In the formula, the coefficient α The range is 0.5-1 and can be determined according to Arthur's law.

[0091] The rate of coke combustion can be expressed as:

[0092]

[0093] In the formula, R C is the reaction rate of coke combustion, is the partial pressure of oxygen, k r and k d are the rate constants due to chemical kinetics and diffusion, respectively.

[0094] The products of the model are N2, O2, CO2, H2O, CH4, CO, H2, NO, HCN and NH3, and the inputs are parameters such as MSW composition, primary air volume, and boiler parameters.

[0095] Based on the uncertainty of MSW composition and the complexity of the combustion process, the following assumptions are used for steady-state analysis: (1) The incinerator is in a stable operating state, the reaction in the furnace can reach equilibrium, the temperature and pressure of each reactor are constant, and pressure loss and heat loss are not considered; (2) During the entire incineration process, it is assumed that the air is evenly distributed and can fully react with MSW; (3) The main components of MSW are C, H, O, N, S and Cl, among which C is converted into gas and ash, and ash does not participate in the reaction; gas-phase N elements react to generate gas compounds, and solid N reacts alone to generate NO; (4) Due to the complex composition of tar, tar-related reactions are not included in the simulation process. The overall block diagram of gas phase combustion simulation based on AspenPlus is shown in the figure. Figure 7 shown.

[0096] Depend on Figure 7 It can be seen that the model includes solid phase combustion zone, gas phase combustion zone, high temperature combustion zone and SNCR purification zone. Its combination with FLIC is to obtain the solid phase combustion flue gas components from FLIC, input them into the B3 module, and then mix them with the primary air volume before entering the subsequent AspenPlus simulation experiment. Figure 7 The functions of each module in are shown in Table 1.

[0097] Table 1 Figure 7 Summary of the functions of each module

[0098]

[0099] In this module, wet MSW first enters the drying module B1, and the moisture is discharged from B2 through the drying module; the dried MSW flows into B3 to mix with the primary air, and then enters the splitter B4 for splitting and flows into the reactor B solid 1-B solid 5 solid phase combustion simulation, 5 simulation reactors simulate the generation of NO2, NO, HCN, NH3 and other conventional gas products, among which NH3 and HCN are NO X The gas precursor, HCN, NH3 generation ratio is 9, all the simulator products flow into the mixer B5, the relevant reaction of the process is:

[0100] MSW→NO2+NO+HCN+NH 3+ ... (13)

[0101] This area simulates the reactions that occur from the time when MSW enters the grate to the time when the temperature rises to 500°C. The main source of NOx in this stage is the release of NOx contained in the compounds and other components in MSW after pyrolysis. In this process, the N element volatilizes from the MSW in the form of various intermediates. This includes the production and decomposition reactions of a small amount of HCN gas, as well as a small amount of NO reduction reaction. The products of the solid phase combustion zone flow into separator B6, which discharges solid slag such as ash, and the char-N is separately diverted and flows into B gas 5 reactor, the remaining gas is mixed with the primary air and flows into the splitter B8, and then flows into B com 1-B com 5 Simulate gas phase combustion reaction, all products flow into mixer B9 for mixing. The reaction is as follows:

[0102] 3H2+N2=2NH3 (14)

[0103] NH3+CO=HCN+H2O (15)

[0104] NO+CH4=H2O+H2+HCN(16)

[0105] This area simulates the process of furnace temperature at 500-800℃. In this process, the intermediate HCN and NH3 gas generated by fuel N decompose into NO in large quantities, and at the same time, NO reduction reaction occurs in some areas where oxygen is insufficient. Below 800℃, the forward reaction of NO is always greater than the reverse reaction. Therefore, the concentration of NOx gradually increases in this temperature range, and the related reactions are as follows:

[0106] HCN+O2→NO+... (17)

[0107] HCN+NO→N2+... (18)

[0108] NH3+O2→NO+... (19)

[0109] NH3+NO→N2+... (20)

[0110] NO2 will also be oxidized to NO above 600℃. In addition, this area contains some reducing gases that will reduce NOx, such as the reduction reaction of CO to NO:

[0111] 2CO+2NO→N2+2CO2 (21)

[0112] In addition, this module also includes the reaction of converting fixed nitrogen into NO. It is generally believed that all Char-N is converted into NO, and the reaction formula is as follows:

[0113] Char-N→NO (22)

[0114] This module is the entire MSWI process NO X The position with the highest concentration is because at this temperature, most of the intermediate products HCN and NH3 of NOx are decomposed into NOx, and fixed N also undergoes conversion reaction.

[0115] The products of the gas phase combustion zone flow into the separator B10 together with the secondary air (SA) to mix and separate HCN and NH3 from other conventional gases. The separated NH3 and HCN flow into B gas 1 and B gas 4. The remaining gas flows into B after being diverted by diverter B11 gas 1-B gas 4 reactors for high temperature combustion simulation, separation in the gas phase combustion zone and flow into B gas 5 Char-N also starts to decompose, and finally all the products flow into the mixer B12 together.

[0116] In MSWI, SNCR denitrification is arranged in the furnace. At the optimal operating temperature of 850℃~1100℃, NH3 can react with NOx generated in the flue gas, thereby achieving the purpose of reducing NOx emissions; when the temperature is higher than 1100℃, NH3 is easily oxidized to NO. The relevant reactions are as follows:

[0117] 4NO+4NH3+O2=4N2+6H2O (23)

[0118] 2NO+4NH3+2O2=3N2+6H2O (24)

[0119] 6NO2+8NH3=7N2+12H2O (25)

[0120] The treated gas flows into the waste heat recovery device and enters the economizer and heat exchanger, where the temperature is reduced to 170°C under the action of water, and then enters the chimney and is discharged into the atmosphere.

[0121] The products of the high-temperature combustion zone enter the B14 yield reactor, along with the entry of water and NH3 to simulate the SNCR denitrification reaction; then the gas treated by SNCR enters B15-B20 to simulate the cooling effect of the three-stage superheater and the three-stage economizer; finally, the gas after cooling is discharged through the Comper.

[0122] Furthermore, combined with the experience of experts in the field of MSWI in Beijing, five manipulated variables, including feed rate, grate speed, primary air flow rate, primary air drying temperature and primary air combustion temperature, were selected as experimental factors for a seven-level orthogonal experimental design. The level values ​​of each factor are shown in Table 2.

[0123] Based on the above experimental design and combined with the numerical simulation model under the previous benchmark conditions, the scheme was implemented, the experimental results were recorded, and the virtual numerical simulation mechanism data set under multiple conditions was obtained by preprocessing.

[0124] Table 2 Experimental design of each factor level value

[0125]

[0126] Furthermore, the nitrogen oxide prediction model includes:

[0127] Feature mapping layer, latent feature extraction layer, feature enhancement layer and incremental learning.

[0128] Specifically, the BHFR algorithm is used here to build the NOx model, which includes four main parts: feature mapping layer, potential feature extraction layer, feature enhancement layer and incremental learning layer, as follows:

[0129] For the original breadth learning system (BLS), a mixed forest group consisting of RF and CRF is used to replace the neurons in the BLS system. First, J training subsets are obtained through the Bootstrap and random subspace method (RSM) sampling strategies, and RF and CRF modeling are performed respectively. Among them, RF uses Bootstrap and RSM to construct a parallel integration model of decision trees with split growth using square error as the loss function, and CRF uses Bootstrap and RSM to construct multiple parallel integration models of decision trees with completely random split growth. Finally, the output of the feature mapping layer is expressed as:

[0130]

[0131] A fully connected strategy is used to realize information transfer between the feature mapping layer, feature enhancement layer, and incremental learning layer to avoid overfitting caused by information loss during information transfer. The potential features of the fully connected mixing matrix extracted by PCA are selected through a feature selection algorithm based on mutual information to achieve dimensionality reduction. Finally, the obtained A new dataset {X′,y} with potential features.

[0132] First, the new data set is sampled to obtain the training subset of the mixed forest algorithm. Then, taking the construction of the jth RF in the kth mixed forest group as an example, the RF model in the kth mixed forest group in the feature enhancement layer is expressed as follows:

[0133]

[0134] Then, the CRF model in the kth mixed forest group in the feature enhancement layer is expressed as:

[0135]

[0136] Combining the above models, the kth hybrid enhancement feature is obtained as follows:

[0137]

[0138] In the formula, Indicates the kth mixed forest group for the Nth in the new data Raw The enhanced map of samples.

[0139] Finally, the output H of the feature enhancement layer K It is expressed as follows:

[0140]

[0141] The BHFR model at this time is expressed as follows: Y = G K W K =[Z1,Z2,…,Z N |H1,H2,…,H K ]W K (31)

[0142] In the formula, G K represents the combination of the feature mapping layer and the enhancement layer output; W K Represents the corresponding weight, which is calculated using the following formula:

[0143] W K =(λI+[G K ] T G K ) -1 [G K ] T Y (32)

[0144] When one mixed forest group is added, the outputs of the feature mapping layer, feature increment layer, and incremental learning layer are as follows:

[0145]

[0146] For G K+1 The inverse matrix of is recursively updated as follows:

[0147]

[0148] The update matrix of the weights between the feature mapping layer, feature incremental layer, incremental learning layer and output layer is:

[0149]

[0150] Where W K =(λI+[G K ] T G K )-1 [G K ] T Y.

[0151] From the above, it can be seen that the pseudo-inverse update only needs to calculate the pseudo-inverse matrix of the mixed forest group of the incremental learning layer. Furthermore, adaptive incremental learning is achieved based on whether the convergence degree of the training error reaches the set threshold.

[0152] Furthermore, the MSW samples used in the experiment came from a MSWI power plant in Beijing. The industrial analysis and elemental analysis results are shown in Table 3. X The corresponding primary air, reaction temperature and other parameter settings of the relevant modules are shown in Table 4.

[0153] Table 3 Industrial analysis and elemental analysis of MSW from industrial sites

[0154]

[0155]

[0156] Table 4 Specifications of incinerator and operating parameters of base condition

[0157]

[0158] The present invention adopts FLIC to simulate solid phase combustion. The main operating parameters of FLIC under the benchmark conditions are shown in Table 5.

[0159] Table 5 Main operating parameter settings of FLIC

[0160]

[0161] Under the benchmark conditions, the gas phase components of solid phase combustion simulated by FLIC were selected as boundary conditions to provide fuel for Aspen Plus to simulate gas phase combustion. The specific module B3 setting parameters are shown in Table 6.

[0162] Table 6 Gas compositions input to Aspen Plus

[0163]

[0164]

[0165] The process rate of solid phase MSW combustion on the grate and the distribution of flue gas components are as follows: Figure 8 shown.

[0166] The distribution results of N2, HCN and NH3 components on the grate are as follows: Fig. 9 shown.

[0167] Depend on Fig. 9It can be seen that after MSW enters the incinerator, it undergoes pyrolysis as the temperature rises, and the N element it contains is released in the form of N2. Then the concentration gradually decreases, reaching the lowest value at about 3.3 grate length. After that, the concentration continues to increase as the combustion continues. The HCN and NH3 contained in the MSW itself are completely released at the beginning of combustion and remain unchanged during the subsequent combustion process.

[0168] The present invention studies the NO at the flue gas (G1) at the solid phase combustion zone outlet (Ga), gas phase combustion zone outlet (Gb), high temperature combustion zone outlet (Gc), SNCR purification zone outlet and boiler outlet under the benchmark working conditions. X The simulation results of these four points are shown in Table 7.

[0169] Table 7 Different positions NO X concentration

[0170]

[0171] It can be seen from Table 7 that when MSW starts to burn, it first releases a small amount of NO X As the combustion continues, a small amount of intermediates in the gas phase combustion zone undergo oxidation reaction to generate N2 under the condition of sufficient fuel; NO is generated in some fuel-poor areas. X , and with the production of reducing gases such as CO, it will react with NO X A reduction reaction occurs, reducing it to N2; as the temperature increases, in the high-temperature heat exchange zone, the oxidation reaction of the intermediate gradually accelerates, reaching a peak when the temperature reaches 800°C, at which time 90% of HCN and NH3 are converted into NO X , although it is still accompanied by NO X The reduction reaction occurs, but the forward reaction rate is much greater than the reverse reaction; in addition, char-N will gradually react to generate NO at this high temperature. X The concentration reaches a peak value; then the gas enters the SNCR purification area, and the NO in the flue gas after treatment X The concentration and temperature decrease, and finally, the flue gas flows into the waste heat boiler, where it is cooled by water to 170°C and finally discharged into the atmosphere through the exhaust chimney.

[0172] Considering the five factors of feed rate (A), grate speed (B), primary air flow (C), primary air drying temperature (D) and primary air combustion temperature (E), 72 groups of experiments were conducted to obtain multi-condition virtual numerical simulation mechanism data. The 72 groups of orthogonal simulation mechanism experimental data after preprocessing were divided into the training set, the validation set and the test machine each accounted for 1 / 4, and the detailed information is shown in Table 8.

[0173] Table 8 Statistical results of experimental data

[0174]

[0175] The root mean square error (RMSE) is used to evaluate the performance of the model. Its expression is as follows:

[0176]

[0177] In the formula, N represents the number of test samples, y i Indicates the actual value of NOx generation concentration, Represents the predicted value.

[0178] Hyperparameter setting of BHFR in the present invention: minimum number of samples N for decision tree leaf nodes sample is 3, the number of decision trees is N tree is 7, the number of RSM feature selections is 3, and the regularization coefficient is 2 -10 , the number of hybrid forest groups N in the feature mapping layer and feature enhancement layer Forest are all 6, and the potential feature contribution rate threshold η is 0.87. Based on the above parameters, the prediction curves of the training set and the test set are as follows Fig.10 shown.

[0179] Depend on Fig.10 It can be seen that the BHFR algorithm has a good prediction effect on the training set and the test set, where the RMSE of the training set reaches 1.41 and the RMSE of the test set reaches 5.09. This result shows that the BHFR algorithm has a good fitting effect for NOx generation concentration modeling.

[0180] RF, DT, BPNN and the BHFR algorithm proposed in this invention are compared, and the parameters are set as follows: 1) RF, the minimum number of samples is N sample 9, the number of RSM feature selections is 6, the number of decision trees is N tree 50; 2) DT, the minimum number of samples of decision tree leaf nodes N sample is 3, the number of RSM feature selection is 6, and the number of decision trees is N tree is 1; 3) BPNN, maximum convergence times 1500, convergence error 0.001, learning rate 0.01.

[0181] The above method was repeated 20 times under the same conditions. The statistical results and test set prediction curves are shown in Table 9 and Fig.10 shown.

[0182] Table 9 Algorithm comparison experimental results

[0183]

[0184] It can be seen from Table 9 that 1) BPNN has the smallest RMSE in the training set and the test set, and has the best model fit for the NOx generation concentration data set in the incinerator; 2) DT and RF are inferior to BPNN and BHFR in terms of accuracy, but DT's modeling accuracy in the test set is better than RF, but still worse than BPNN; 3) RF is slightly better than DT in the modeling accuracy of the training set, but its model accuracy in the test set is worse than BPNN and DT; 4) BHFR has the best RMSE indicator mean statistical results in both the training set and the test set, indicating that BHFR has good modeling accuracy, good generalization performance and stability for the NOx concentration data set in the incinerator.

[0185] In summary, the BHFR of the present invention has better learning ability than RF, DT and BPNN, and has obvious advantages in constructing a prediction model for NOx generation concentration in an incinerator.

[0186] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0187] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting nitrogen oxides in the process of municipal solid waste incineration, characterized in that: include: Based on the MSWI process, determine the combustion area and temperature range in the incinerator; Based on the combustion area and temperature range, determine a pollutant emission data set according to equipment parameters and operating parameters of the incinerator; According to the pollutant emission data set and the manipulated variables, a virtual numerical simulation mechanism data set under multiple working conditions is obtained by using an orthogonal experiment; The nitrogen oxide prediction result is obtained according to the virtual numerical simulation mechanism data set and the nitrogen oxide prediction model.

2. The method for predicting nitrogen oxides in the process of municipal solid waste incineration according to claim 1, characterized in that: The manipulated variables include: Feed rate, grate speed, primary air flow, primary air drying temperature and primary air combustion temperature.

3. The method for predicting nitrogen oxides in the process of municipal solid waste incineration according to claim 1, characterized in that: The pollutant emission data set is determined based on the combustion area and temperature range according to the equipment parameters and operation parameters of the incinerator, including: Construct FLIC model and Aspen PLus model; Determine the MSW composition, primary air volume and boiler parameters corresponding to the combustion area and each temperature range; Inputting the MSW composition, primary air volume and boiler parameters into the FLIC model to obtain an initial output result; The initial output results are input into the AspenPLus model to obtain a pollutant emission data set.

4. The method for predicting nitrogen oxides in the process of municipal solid waste incineration according to claim 1, characterized in that: The pollutant emission data set includes: N2, O2, CO2, H2O, CH4, CO, H2, NO, HCN and NH3.

5. The method for predicting nitrogen oxides in the process of municipal solid waste incineration according to claim 1, characterized in that: The nitrogen oxide prediction model includes: Feature mapping layer, latent feature extraction layer, feature enhancement layer and incremental learning.

6. The method for predicting nitrogen oxides in the process of municipal solid waste incineration according to claim 5, characterized in that: The output expression of the feature mapping layer is: Among them, Z N Contains N Raw samples and 2N-dimensional features.

Images