CN115758824A - Data-driven three-dimensional soft measurement method for NOx concentration of boiler furnace - Google Patents
Data-driven three-dimensional soft measurement method for NOx concentration of boiler furnace Download PDFInfo
- Publication number
- CN115758824A CN115758824A CN202211444645.5A CN202211444645A CN115758824A CN 115758824 A CN115758824 A CN 115758824A CN 202211444645 A CN202211444645 A CN 202211444645A CN 115758824 A CN115758824 A CN 115758824A
- Authority
- CN
- China
- Prior art keywords
- data
- nox concentration
- boiler
- working condition
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Regulation And Control Of Combustion (AREA)
Abstract
The invention provides a data-driven boiler furnace NOx concentration three-dimensional distribution soft measurement method, which comprises the steps of obtaining operation parameters of all typical working conditions under the stable operation of a boiler according to the actual operation condition of a power plant, drawing a geometric model of the boiler and carrying out grid division, obtaining a boiler furnace three-dimensional NOx concentration distribution data set under each typical working condition through numerical simulation, reconstructing a modeling data set and a verification data set, training model parameters based on a training data set, constructing a boiler furnace NOx concentration three-dimensional distribution soft measurement model based on an extreme learning machine algorithm, and carrying out on-line prediction on a boiler furnace NOx concentration three-dimensional value by using the trained model. The method can be used for soft measurement of the three-dimensional distribution condition of the NOx concentration in the boiler hearth, has the advantages of high soft measurement precision, high convergence speed and the like, and is beneficial to improving the boiler efficiency and reducing the boiler emission.
Description
Technical Field
The invention belongs to the technical field of thermal power generation, and particularly relates to a three-dimensional distribution soft measurement method for NOx concentration of a boiler hearth based on data driving.
Background
Increased emissions of Nitrogen Oxides (NOx) have had a tremendous impact on both the ecological environment and human health. However, the combustion process of the boiler is a black box state, and the three-dimensional distribution of the NOx concentration of the boiler cannot be known, so that the operating personnel cannot adjust the boiler parameters in time, the operating state of the boiler deviates from the design state, and the NOx emission concentration is increased.
The concentration of NOx obtained at present is mainly the concentration of NOx at a measuring point at the tail part of a flue, and the three-dimensional distribution condition of the concentration of NOx in a boiler hearth cannot be obtained. The three-dimensional distribution condition of the NOx concentration of the boiler hearth can be obtained through Fluent numerical simulation calculation, the simulation result is well matched with the actual field measurement result, the operation condition of the boiler can be accurately reflected, and particularly in the related aspects of flow, combustion, pollutant release and the like in the boiler, the results can be used for guiding the design and operation of the boiler. However, the numerical simulation relates to complex processes such as turbulence, the calculation time is long, the convergence rate is slow, and the requirement of rapidity of real-time monitoring of the concentration of NOx cannot be met.
Disclosure of Invention
Based on the problems, the three-dimensional distribution soft measuring method for the NOx concentration of the boiler hearth based on data driving can quickly and accurately measure the three-dimensional NOx concentration distribution situation and meet the actual operation requirement.
A three-dimensional distribution soft measurement method of NOx concentration of a boiler furnace based on data driving comprises the following steps:
step 1: according to the actual operation condition of the power plant, obtaining the operation parameters of all typical working conditions under the stable operation of the boiler, wherein the operation parameters comprise: the system comprises a boiler load, a total air quantity, a total coal quantity, a total primary air quantity, a total secondary air quantity, a coal machine operation mode and a secondary air temperature, wherein the typical working condition refers to a working condition capable of covering the operation state of a power plant;
step 2: drawing a geometric model of the boiler and carrying out grid division;
and step 3: numerical simulation is carried out by adopting Fluent software to obtain a three-dimensional NOx concentration distribution data set M of a boiler furnace under various typical working conditions i (ii) a The method comprises the following steps:
step 3-1: selecting a mathematical model required by simulation, and setting simulation boundary conditions;
step 3-2: setting corresponding operating parameters of each typical working condition and carrying out simulation calculation;
step 3-3: obtaining three-dimensional NOx concentration distribution data set M of boiler furnace under each typical working condition through simulation i I =1,2, F is the total number of the determined typical working conditions, and the corresponding data is expressed as a matrixWhereinX-coordinate values representing data points of simulation results,Y-coordinate value representing a data point of a simulation result,Z-coordinate value representing a data point of a simulation result,Representing simulated values of NOx concentration at corresponding three-dimensional coordinate points, and n represents the total number of data points obtained by the simulation.
And 4, step 4: for typical working condition data set M obtained by simulation i Classifying and selecting the reference working condition of each type; the method comprises the following steps:
step 4-1: classifying all the determined typical working conditions according to the opening degree of the burnout air door and the opening degree of the secondary air door to obtain N classifications;
step 4-2: calculating Euclidean distance d of data corresponding to any two typical working conditions o and p in each classification op :
Step 4-3: calculating the sum of the Euclidean distances between any working condition and all other working conditions in each classification;
step 4-4: and selecting the Euclidean distance and the minimum working condition as reference working conditions, namely obtaining one reference working condition for each type, and obtaining N reference working conditions in total.
And 5: typical working condition data set M obtained according to simulation i Reconstructing the modeling dataset and the verification dataset; the method comprises the following steps:
step 5-1: randomly selecting 1 typical working condition in each classification except the reference working condition, and making a difference value with data corresponding to the reference working condition to construct a verification data set;
step 5-2: at each data set M i Randomly selecting J data point coordinates from the n data points;
step 5-3: extracting data from the remaining L-2 working conditions and the reference working conditions according to the J data coordinate points selected in the step 5-2 except the 1 typical working condition and the reference working condition selected in the step 5-1, and then performing difference value to construct a modeling data set;
step 5-4: carrying out normalization processing on the modeling data to obtain a training data set;
wherein h is i In order to model the data set in a model,for the normalized training data set, h min 、h max The minimum value and the maximum value of the corresponding variable in the modeling data set are obtained.
And 6: training model parameters based on a training data set, and constructing a three-dimensional distribution soft measurement model of the NOx concentration of the boiler furnace based on an Extreme Learning Machine (ELM) algorithm; and carrying out on-line prediction on the three-dimensional value of the NOx concentration of the boiler furnace by using the trained model.
Input weight ω of ELM i And offset b i Is randomly generated and only the output weight beta needs to be determined i For a single hidden layer neural network with L neurons, this can be expressed as:
wherein j =1,2, …, N; g (x) is an activation function; omega i Is the input weight; beta is a i Is the output weight; b i Bias for the ith hidden layer; omega i ·X i Represents omega i And X i Inner product of (2); o j Is output by the neural network. The learning of the ELM neural network is mainly to minimize the output error as much as possible, and can be expressed as:
wherein t is j Is real data. I.e. the presence of beta i 、ω i And b i When equation (4) is satisfied, it can be expressed as:
the method is simplified as follows:
Hβ=T (6)
wherein H is the neuron output; beta is the output weight; and T is the output of the neural network.
Training the single hidden layer neural network to obtain optimal input weights, bias and output weights such that:
the final ELM-based in-furnace NOx soft measurement model can be written as:
C s =f ELM (x s ,ω i ,b i ,β i ) (8)
wherein, C s For soft measurement results, x s Modeling samples; omega i Is the input weight; beta is a i Is the output weight; b i Biasing for the ith hidden layer,f ELM () Is a soft measurement model.
The beneficial effects of the invention are:
the invention provides a data-driven boiler furnace NOx concentration three-dimensional distribution soft measurement method, which adopts a method of fusing typical working condition numerical simulation data and actual operation data, can overcome the limitation of a traditional method on data source, takes ELM as a modeling algorithm, can soft-measure the three-dimensional distribution condition of the NOx concentration in the boiler furnace, has the advantages of high soft measurement precision, high convergence rate and the like, and is beneficial to improving the boiler efficiency and reducing the boiler emission.
Drawings
FIG. 1 is a flow chart of a data-driven boiler furnace NOx concentration three-dimensional distribution-based soft measurement method provided by an embodiment of the invention;
FIG. 2 is a comparison graph of soft measurements for a first type of operating condition provided by an embodiment of the present invention;
FIG. 3 is a comparison graph of soft measurements for a second type of operating condition provided by an embodiment of the present invention;
FIG. 4 is a comparison graph of soft measurements for a third type of operating condition provided by an embodiment of the present invention;
FIG. 5 is a comparison graph of soft measurements for a fourth type of operating condition provided by an embodiment of the present invention;
FIG. 6 is a comparison graph of soft measurements for a fifth type of operating condition provided by an embodiment of the present invention;
FIG. 7 is a comparison graph of soft measurements for a sixth type of operating condition provided by an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and specific examples.
As shown in FIG. 1, a data-driven three-dimensional soft measurement method for NOx concentration in a boiler furnace comprises the following steps:
step 1: according to a certain 350MW supercritical lignite boiler, 120 operating parameters of typical working conditions of the boiler are obtained, and the operating parameters totally comprise 7 operating parameters: boiler load, total air volume, total coal volume, total primary air volume, total secondary air volume, coal mill operation mode and secondary air temperature; the typical working condition selected in the embodiment is the working condition that the boiler stably runs under 100% load, 75% load and 50% load under different coal mill running conditions;
step 2: according to the size of a hearth of a certain 350MW boiler: the height is 64.8 meters, the width is 14.6273 meters, the depth is 14.6273 meters, the depth of a tail rear flue is 6.82 meters, the elevation of a cold ash bucket is 6.5 meters, a geometric model of a research object is drawn, grid division is carried out by using Gambit software, structured grid division is carried out on a cold ash bucket area, unstructured grid division is carried out on a combustor area, and 2776652 grid points are finally obtained;
and step 3: numerical simulation is carried out by adopting Ansys Fluent software to obtain a three-dimensional NOx concentration distribution data set M of the boiler furnace under the ith typical working condition i ,i=1,2,···,120;
Step 3-1: selecting a mathematical model required by simulation, and setting simulation operation parameters;
selecting a Realizable k-epsilon model as a turbulence model, selecting a two-step competition reaction model as a volatilization analysis model, selecting a dynamics/diffusion reaction rate model to simulate a coke combustion process, and selecting a DO model to calculate the radiation heat exchange between gas and solid;
step 3-2: setting boiler coal quality and simulation operation parameters corresponding to each typical working condition according to actual operation parameters acquired by the coal furnace in an actual operation state, wherein the boiler coal quality and the simulation operation parameters comprise boiler load, total air volume, total coal volume, total primary air volume, total secondary air volume, coal mill operation mode and secondary air temperature, and performing simulation calculation;
wherein the coal quality parameters are shown in table 1:
TABLE 1 coal quality parameter table
The boiler operating parameters are shown in table 2:
TABLE 2 operating parameters Table
Determining a basic conservation equation which is satisfied by a pulverized coal combustion process in a boiler hearth, wherein the basic conservation equation comprises a mass conservation equation, a momentum conservation equation and an energy conservation equation and is expressed as follows:
mass conservation equation (continuity equation):
conservation of momentum equation:
wherein rho is the density of the medium; t is time; v is a velocity vector; ρ g i Is gravity; f i Is an external force; tau is ij Is the viscous stress tensor;
energy conservation equation:
wherein E is energy; k is a radical of eff Effective thermal conductivity; s h Is a chemical reaction heat source item;
step 3-3: simulating to obtain a boiler furnace three-dimensional NOx concentration distribution data set M under the ith typical working condition i The corresponding data is expressed as a matrixWhereinX-coordinate values representing data points of simulation results,Y-coordinate value representing data point of simulation result,Z-coordinate value representing data point of simulation result,And expressing simulated values of the NOx concentration at the corresponding three-dimensional coordinate point, and obtaining 2776652 groups of data in each working condition.
And 4, step 4: for typical working condition data set M obtained by simulation i Classifying and selecting a reference working condition of each classification;
step 4-1: classifying the typical working conditions according to the opening degree of the burnout air door and the opening degree of the secondary air door to obtain 6 classes in total, wherein each class contains 20 working conditions, and the classification result is shown in table 3;
TABLE 3 Classification results Table
Specific results of the opening degree of the burnout air door and the opening degree of the secondary air door are shown in table 4;
step 4-2: calculating Euclidean distance d of data corresponding to any two typical working conditions o and p in 20 working conditions in each class op ,
Step 4-3: calculating the sum of the Euclidean distances between any working condition and all other working conditions in each classification;
step 4-4: and selecting the Euclidean distance and the minimum working condition as reference working conditions, namely obtaining one reference working condition for each type, and obtaining 6 reference working conditions in total.
And 5: reconstructing the modeling dataset and the verification dataset;
step 5-1: randomly selecting 1 working condition in each classification except the reference working condition, and making a difference value with the reference working condition to construct a verification data set; obtaining 6 groups of data sets in total, and carrying out modeling verification;
step 5-2: randomly selecting 50000 coordinate points from 2776652 coordinate points;
step 5-3: in addition to the randomly selected working conditions and the reference working conditions in the step 5-1, data are extracted from the remaining 18 working conditions and the reference working conditions according to coordinate points and then difference values are made, a modeling data set is constructed, 900000 groups of data are obtained in total, and the structure of the data is shown in table 4;
TABLE 4 modeling data set Table
Step 5-4: carrying out normalization processing on the modeling data set to obtain a training data set;
wherein h is i In order to model the data set in a model,for the normalized training data set, h min 、h max The minimum value and the maximum value of the corresponding variable in the modeling data set are obtained.
And 6: constructing a three-dimensional distribution soft measurement model of the NOx concentration of the boiler furnace based on an Extreme Learning Machine (ELM) algorithm based on a training data set; establishing a three-dimensional NOx concentration distribution condition soft measurement model of the rapid boiler furnace in each type according to a training data set to obtain 6 soft measurement models in total; wherein the ELM has 10 input nodes, 1 output node, and the number of nodes of the hidden layer is 200;
and predicting the three-dimensional NOx concentration value of the boiler furnace under each classification by using the trained soft measurement model according to the collected actual data of the typical working conditions under each classification.
The modeling error of the three-dimensional NOx concentration distribution soft measurement model of the 6 classified boiler furnace is shown in the table 5.
TABLE 5 modeling error
Wherein
N is the number of samples;is a soft measurement of the model; y is i The simulation value is obtained by numerical simulation;is the average of simulated values.
From table 5 and fig. 2 to 7, it can be seen that the method has higher model accuracy, can quickly and soft measure the three-dimensional NOx concentration distribution of the boiler furnace, better meets the actual production requirement, and has wide application prospect.
Claims (5)
1. A three-dimensional distribution soft measurement method of NOx concentration of a boiler furnace based on data driving is characterized by comprising the following steps:
step 1: obtaining operation parameters of all typical working conditions under the stable operation of the boiler according to the actual operation condition of the power plant;
step 2: drawing a geometric model of the boiler and carrying out grid division;
and 3, step 3: numerical simulation is carried out by adopting Fluent software to obtain a three-dimensional NOx concentration distribution data set M of a boiler furnace under various typical working conditions i ;
And 4, step 4: for typical working condition data set M obtained by simulation i Classifying and selecting the reference working condition of each type;
and 5: typical working condition data set M obtained according to simulation i Reconstructing the modeling dataset and the verification dataset;
step 6: training model parameters based on a training data set, and constructing a boiler furnace NOx concentration three-dimensional distribution soft measurement model based on an extreme learning machine; and carrying out on-line prediction on the three-dimensional value of the NOx concentration of the boiler furnace by using the trained model.
2. The data-driven boiler furnace NOx concentration three-dimensional distribution soft measuring method based on the claim 1 is characterized in that the operation parameters in the step 1 comprise: the system comprises a boiler load, a total air quantity, a total coal quantity, a total primary air quantity, a total secondary air quantity, a coal machine operation mode and a secondary air temperature, wherein the typical working condition refers to a working condition capable of covering the operation state of a power plant.
3. The data-driven three-dimensional distribution soft measurement method for the NOx concentration of the boiler furnace according to claim 1, wherein the step 3 comprises the following steps:
step 3-1: selecting a mathematical model required by simulation, and setting simulation boundary conditions;
step 3-2: setting corresponding operating parameters of each typical working condition and carrying out simulation calculation;
step 3-3: obtaining three-dimensional NOx concentration distribution data set M of boiler furnace under each typical working condition through simulation i I =1,2, F is the total number of the determined typical working conditions, and the corresponding data is expressed as a matrixWhereinX-coordinate values representing data points of simulation results,Y-coordinate value representing data point of simulation result,Z-coordinate value representing data point of simulation result,Representing simulated values of NOx concentration at corresponding three-dimensional coordinate points, and n represents the total number of data points obtained by the simulation.
4. The data-driven three-dimensional distribution soft measurement method for the NOx concentration of the boiler furnace according to claim 1, wherein the step 4 comprises the following steps:
step 4-1: classifying all the determined typical working conditions according to the opening degree of the burnout air door and the opening degree of the secondary air door to obtain N classifications;
step 4-2, calculating Euclidean distance d of data corresponding to any two typical working conditions in each classification op :
Step 4-3: calculating the sum of the Euclidean distances between any working condition and all other working conditions in each classification;
step 4-4: and selecting the Euclidean distance and the minimum working condition as reference working conditions, namely obtaining one reference working condition for each type, and obtaining N reference working conditions in total.
5. The data-driven three-dimensional distribution soft measurement method for the NOx concentration of the boiler furnace according to claim 1, wherein the step 5 comprises the following steps:
step 5-1: randomly selecting 1 typical working condition in each classification except the reference working condition, and making a difference value with data corresponding to the reference working condition to construct a verification data set;
step 5-2: at each data set M i Randomly selecting J data point coordinates from the n data points;
step 5-3: extracting data from the remaining L-2 working conditions and the reference working conditions according to the J data coordinate points selected in the step 5-2 except the 1 typical working condition and the reference working condition selected in the step 5-1, and then performing difference value to construct a modeling data set;
step 5-4: and carrying out normalization processing on the modeling data to obtain a training data set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211444645.5A CN115758824A (en) | 2022-11-18 | 2022-11-18 | Data-driven three-dimensional soft measurement method for NOx concentration of boiler furnace |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211444645.5A CN115758824A (en) | 2022-11-18 | 2022-11-18 | Data-driven three-dimensional soft measurement method for NOx concentration of boiler furnace |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115758824A true CN115758824A (en) | 2023-03-07 |
Family
ID=85373099
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211444645.5A Pending CN115758824A (en) | 2022-11-18 | 2022-11-18 | Data-driven three-dimensional soft measurement method for NOx concentration of boiler furnace |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115758824A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116610170A (en) * | 2023-07-17 | 2023-08-18 | 北京中科润宇环保科技股份有限公司 | Efficient SNCR intelligent three-dimensional temperature partition control method |
-
2022
- 2022-11-18 CN CN202211444645.5A patent/CN115758824A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116610170A (en) * | 2023-07-17 | 2023-08-18 | 北京中科润宇环保科技股份有限公司 | Efficient SNCR intelligent three-dimensional temperature partition control method |
CN116610170B (en) * | 2023-07-17 | 2023-10-13 | 北京中科润宇环保科技股份有限公司 | Efficient SNCR intelligent three-dimensional temperature partition control method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105864797B (en) | Real-time prediction system and method for boiler entering heat value of circulating fluidized bed household garbage incineration boiler | |
CN111339716B (en) | Boiler high-temperature flue gas flow field online agent model construction method | |
CN103676822B (en) | The control device in thermal power plant and control method | |
CN111222685B (en) | Boiler wide-load NOx emission concentration prediction method based on model migration | |
CN110032747A (en) | A kind of coal-burning boiler NOxDischarge On-line Measuring Method | |
Tang et al. | A deep learning model for measuring oxygen content of boiler flue gas | |
CN115758824A (en) | Data-driven three-dimensional soft measurement method for NOx concentration of boiler furnace | |
Hu et al. | Multi-model ensemble prediction model for carbon efficiency with application to iron ore sintering process | |
Wilkinson et al. | Approximating computational fluid dynamics for generative tall building design | |
Thummar et al. | Machine learning for vortex flowmeter design | |
Wang et al. | Prediction of swirling flow field in combustor based on deep learning | |
Glumac et al. | A multi-fidelity wind surface pressure assessment via machine learning: A high-rise building case | |
Xu et al. | A new online optimization method for boiler combustion system based on the data-driven technique and the case-based reasoning principle | |
CN116542164A (en) | Combustion system optimization design method based on physical driving parameterized proxy model | |
Zha et al. | An improved extreme learning machine with self-recurrent hidden layer | |
Ilamathi et al. | Predictive modelling and optimization of nitrogen oxides emission in coal power plant using Artificial Neural Network and Simulated Annealing | |
Rajabi et al. | Intelligent prediction of turbulent flow over backward-facing step using direct numerical simulation data | |
CN116085823A (en) | Boiler combustion control method and system | |
Liu et al. | Numerical simulation and mathematical models of ash deposition behavior considering particle properties and operating conditions | |
Lv et al. | Prediction of the 3D Distribution of NOx in a Furnace via CFD Data Based on ELM | |
Jo et al. | Validation of computational fluid dynamics (CFD) platforms for the early stages of architectural design | |
Cao et al. | Efficient optimization design of flue deflectors through parametric surrogate modeling with physics-informed neural networks | |
CN115510638A (en) | Rotary kiln multi-target parameter optimization method based on GRNN model | |
Shi et al. | Online prediction of unburned carbon content in fly ash with clustering LS-SVM models | |
CN116959595A (en) | Prediction method for combustion efficiency of biomass cogeneration system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |