Infrared absorption detection model for CO2 concentration in boiler flue tail gas and construction method thereof

Abstract

The application discloses a boiler flue tail gas CO2 concentration infrared absorption detection model and a construction method thereof. The construction method comprises the following steps: calibrating experiments on standard gases with different CO2 concentrations to obtain relative light intensity at different wavelengths; analyzing and processing experimental data by different methods; comparing and analyzing processing methods by comparing prediction results of CO2 concentration prediction models of the standard gases obtained by different analysis and processing methods; mixing the standard gases with water vapor to prepare mixed sample gases with different CO2 concentrations; performing infrared detection on the mixed sample gases after condensation and dehumidification and controlling relative humidity of the mixed sample gases to target values to obtain relative light intensity of the mixed sample gases with different CO2 concentrations after dehumidification; and analyzing and processing experimental data by the screened analysis and processing method to obtain a CO2 concentration prediction model of the mixed sample gases after dehumidification. The application can reduce the interference of water vapor in the tail gas on the infrared absorption detection of the CO2 concentration of the boiler flue tail gas, and improve the detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of CO2 detection technology in boiler flue gas. Specifically, it relates to an infrared absorption detection model for CO2 concentration in boiler flue gas and its construction method. Background Technology

[0002] Factory flues emit large amounts of CO2 into the atmosphere, a major source of global carbon emissions. To facilitate source monitoring and control of carbon emissions, online CO2 concentration detection devices are needed in industrial sites. Currently, most industrial boilers still require raw emission factors to calculate carbon emissions, inevitably leading to high uncertainty in the obtained figures. Flue gas composition is complex, containing not only CO2 but also N2, nitrogen oxides, sulfur oxides, CO, and water vapor. The presence of these components negatively impacts CO2 detection. Especially when using infrared detection methods, water vapor content significantly affects the accuracy of the results. Therefore, the influence of water vapor content on the detection results must be considered when performing concentration detection. Thus, it is necessary to construct a model for detecting CO2 concentration in boiler flue gas using infrared absorption to reduce water vapor interference and improve the accuracy of infrared detection results for CO2 concentration in flue gas. Summary of the Invention

[0003] Therefore, the technical problem to be solved by the present invention is to provide an infrared absorption detection model for CO2 concentration in boiler flue gas and a method for constructing it, so as to reduce the interference of water vapor in the flue gas when detecting CO2 concentration in boiler flue gas by infrared absorption and improve the accuracy of detection.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] The method for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas includes the following steps:

[0006] Step (1): Calibration experiments were conducted on standard gases with different CO2 concentrations at room temperature and pressure to obtain the relative light intensity at different wavelengths;

[0007] Step (2): The experimental data obtained in the calibration experiment in step (1) are analyzed and processed using different analytical methods to obtain standard gas CO2 concentration prediction models. The prediction results of different standard gas CO2 concentration prediction models are compared to select the analytical processing method for the experimental data.

[0008] Step (3): Mix the standard gas with water vapor to prepare mixed sample gases with different CO2 concentrations;

[0009] Step (4): After the mixed sample gas is dehumidified by condensation, its relative humidity is controlled to reach the target value and then infrared detection is performed to obtain the relative light intensity of the mixed sample gas with different CO2 concentrations after dehumidification.

[0010] Step (5): The experimental data obtained in step (4) is analyzed and processed using the experimental data analysis and processing method selected in step (2) to obtain a prediction model for CO2 concentration of the mixed sample gas after dehumidification.

[0011] Step (6): Calculate the CO2 concentration of the boiler flue gas tail gas based on the CO2 concentration prediction model of the mixed sample gas after dehumidification.

[0012] In the above method for constructing the infrared absorption detection model of CO2 concentration in boiler flue gas, in step (1), the detection frequency band for infrared detection of CO2 concentration in boiler flue gas is 1.45μm.

[0013] In the above method for constructing the infrared absorption detection model of CO2 concentration in boiler flue gas, in step (1), the CO2 concentration in the standard gas is 0, 10%, 20%, 30%, 40%, and 50%; the water vapor content in the standard gas is 0.

[0014] In the above method for constructing the infrared absorption detection model for CO2 concentration in boiler flue gas, step (2) involves analysis and processing using the least squares method and a neural network model.

[0015] The above-mentioned method for constructing the infrared absorption detection model for CO2 concentration in boiler flue gas involves using a neural network model to build a standard gas CO2 concentration prediction model. For each concentration of standard gas, 20 experiments are conducted. Temperature, humidity, and transmitted light intensity are used as three input variables, and CO2 concentration is used as the output variable to establish the model and calculate the final sample gas concentration. Eighteen sets of data are selected for each concentration, totaling 102 sets, for model training. The remaining two sets of data for each concentration are used to validate the detection model. When using the least squares method to construct the standard gas CO2 concentration prediction model, the number of experiments for each concentration of standard gas is greater than or equal to five.

[0016] In the above method for constructing the infrared absorption detection model for CO2 concentration in boiler flue gas, in step (2), when constructing the standard gas CO2 concentration prediction model using the least squares method, the fitting equation of the prediction model is assumed to be:

[0017] g(x)=c1 f1(x)+c2 f2(x)+c3 f3(x)+…+c n f n (x);

[0018] Where f1(x), f2(x), f3(x)...f n Let (x) be a known function, and c1, c2, c3…cn Undetermined coefficients:

[0019] (x1,y1), (x2,y2), (x3,y3)…(x n ,y n If ) represents the data measured in the experiment, then:

[0020] A×c=y;

[0021] c = [c1, c2, ..., c n ] T ;

[0022] Thus, a standard gas CO2 concentration prediction model can be calculated.

[0023] In the above method for constructing the infrared absorption detection model of CO2 concentration in boiler flue gas, in step (3), the CO2 concentration in the mixed sample gas is 0%, 10%, 20%, 30%, 40%, and 50%; the relative humidity of the mixed sample gas with different CO2 concentrations is 50%.

[0024] The above method for constructing the infrared absorption detection model of CO2 concentration in boiler flue gas includes step (4), where the relative humidity of the mixed sample gas with different CO2 concentrations after dehumidification is controlled at 22-24% (the humidity is set in the range of 22-24% because excessive humidity after dehumidification will affect the detection accuracy, but low humidity is difficult to achieve); and the temperature of the mixed sample gas is controlled to be the same as the temperature before dehumidification when performing infrared detection.

[0025] In the above method for constructing the infrared absorption detection model for CO2 concentration in boiler flue gas, step (6) involves the following formula for calculating the CO2 concentration in boiler flue gas:

[0026] C = (V 样 ×C0) / V 总 ;

[0027] C represents the CO2 concentration in the boiler flue gas, and V 总 V is the total volume of the mixed sample gas. 样 C0 represents the remaining volume of the mixed sample gas after dehumidification; C0 represents the CO2 concentration of the mixed sample gas after dehumidification.

[0028] The infrared absorption detection model for CO2 concentration in boiler flue gas is characterized by being obtained using the aforementioned method for constructing the infrared absorption detection model for CO2 concentration in boiler flue gas; that is, the formula for calculating the CO2 concentration in boiler flue gas is: C = (V 样 ×C0) / V 总 ;

[0029] C represents the CO2 concentration in the boiler flue gas, and V 总 V is the total volume of the mixed sample gas.样 C0 represents the remaining volume of the mixed sample gas after dehumidification; C0 is the CO2 concentration prediction model of the mixed sample gas after dehumidification obtained in step (5). The CO2 concentration prediction model of the mixed sample gas after dehumidification is as follows:

[0030] C0 = 11.0178RH - 0.0331I; C0 is the CO2 concentration of the mixed sample gas after dehumidification; RH is the initial relative humidity of the mixed sample gas; I is the relative light intensity measured in the experiment.

[0031] The technical solution of the present invention achieves the following beneficial technical effects:

[0032] The infrared absorption detection model for CO2 concentration in boiler flue gas constructed in this invention is based on the principle that the boiler flue gas is dehumidified before infrared detection to reduce interference from water vapor. However, the dehumidified flue gas still contains water vapor, therefore, water vapor parameters need to be considered when constructing the CO2 detection calculation model. The infrared absorption detection model for CO2 concentration in boiler flue gas constructed using the method of this invention fully considers the influence of temperature and humidity on the infrared absorption detection results of CO2 concentration in boiler flue gas, and provides a CO2 concentration conversion formula based on changes in water vapor concentration and sample gas volume. Verification shows that the model has a maximum relative error of 9% and a minimum of 1.16%, exhibiting high accuracy in concentration detection. Attached Figure Description

[0033] Figure 1 Structural block diagram of the infrared detection system for CO2 concentration in boiler flue gas in this embodiment of the invention;

[0034] Figure 2 Flowchart of infrared detection of CO2 concentration in boiler flue gas in this embodiment of the invention;

[0035] Figure 3a A physical model diagram of the infrared detection system for CO2 concentration in boiler flue gas in an embodiment of the present invention;

[0036] Figure 3b Flow chart of infrared detection process for CO2 concentration in boiler flue gas in an embodiment of the present invention;

[0037] Figure 3c A partial control electrical schematic diagram of the heating wire (heating tube) in an embodiment of the present invention;

[0038] Figure 3d PLC control block diagram in an embodiment of the present invention;

[0039] Figure 3e External wiring diagram of S7-1500 PLC in this embodiment of the invention;

[0040] Figure 3f The main program design flowchart in this embodiment of the invention;

[0041] Figure 3g Flowchart of the CO2 concentration detection subroutine in this embodiment of the invention;

[0042] Figure 3h Flowchart of the sample gas flow control subroutine in this embodiment of the invention;

[0043] Figure 3i Flowchart of the dehumidification system subroutine in this embodiment of the invention;

[0044] Figure 3j Flowchart of the sample gas temperature detection subroutine in this embodiment of the invention;

[0045] Figure 4a In the embodiments of the present invention, the CO2 near-infrared absorption peak is shown.

[0046] Figure 4b In the embodiments of the present invention, CO2, CO, and NO are used. x Near-infrared absorption peak;

[0047] Figure 4c In the embodiments of the present invention, H2O and CO2, CO, NO x Comparison of absorption peaks in the near-infrared band;

[0048] Figure 4d Schematic diagram of gas infrared absorption in an embodiment of the present invention;

[0049] Figure 5 Neuron model diagram of artificial neural network in this embodiment of the invention;

[0050] Figure 6 Neural network flowchart in an embodiment of the present invention;

[0051] Figure 7 Flowchart of the concentration conversion subroutine in this embodiment of the invention;

[0052] Figure 8 Curves showing test results at different concentrations in embodiments of the present invention;

[0053] Figure 9 Concentration prediction results diagram in the embodiments of the present invention;

[0054] Figure 10 A diagram showing the neural network calculation results in an embodiment of the present invention;

[0055] Figure 11 A diagram of the experimental setup for the effect of humidity in this invention embodiment;

[0056] Figure 12 The flow rate curve is shown in the embodiment of this invention;

[0057] Figure 13The ambient temperature curve is shown in the embodiment of this invention.

[0058] Figure 14 The humidity curve of the detected environment is shown in the embodiment of the present invention;

[0059] Figure 15 The results of tests at different concentrations after humidification in this embodiment of the invention are shown in the figure. Detailed Implementation

[0060] Part 1: Boiler Flue Gas CO2 Concentration Detection Scheme and Related Theories

[0061] 1.1 Testing Requirements

[0062] Monitoring the boiler flue gas environment requires numerous sensors, and the automated detection process necessitates a core controller to perform control functions. A sampling and control system for CO2 in the flue gas needs to be designed to collect various environmental values ​​and control parameters. CO2 concentration detection employs infrared absorption, requiring the design of components such as the light source, gas chamber, and sensors to acquire data and select a suitable spectrometer for sample gas detection. The infrared detection system detects the processed, dried sample gas. Due to the removal of water vapor, the CO2 concentration in the dried sample gas differs from that in the flue gas; therefore, the measured results need to be converted to obtain the CO2 concentration within the flue gas.

[0063] 1.2 Overall Scheme Design

[0064] A comprehensive scheme design was developed for online CO2 detection in boiler flue gas. Based on detection requirements and similar online gas monitoring schemes, the system employs a sampling detection method for online detection. To complete gas sampling, a sampling gun is connected to an external centrifugal pump, which extracts the gas from the flue to obtain sample gas. Temperature and humidity sensors and a flow meter are placed inside the flue to monitor temperature, humidity, and flow rate. A flow meter is placed outside the flue to control the sample gas flow rate. To reduce sample gas humidity, the sample gas processing module uses an electronic condenser and activated carbon dehumidification to obtain dry sample gas. Temperature and humidity sensors are placed after the sample gas processing system to detect the temperature and humidity of the dry sample gas. The temperature of the processed dry sample gas must remain consistent with that before filtration; this is achieved by heating with a heating rod. Infrared detection uses a gas chamber and light source combined with a near-infrared spectrometer to collect relative light intensity data for CO2 concentration calculation. Therefore, this system includes a main control module, a sampling module, a sample gas processing module, an infrared detection module, and parameter detection modules. The overall design block diagram is shown below. Figure 1 As shown.

[0065] The overall system flow is as follows: the main control module controls the operation of each sensor. The sampling module primarily samples the gas in the flue to obtain sample gas. The collected sample gas enters the sample gas processing module, where an electronic condenser separates water vapor from the gas. The dried sample gas then enters the gas chamber, where an infrared detection module detects the gas. Finally, the CO2 concentration in the flue gas is calculated. The temperature, humidity, and flow control module uses sensors to detect the temperature and humidity of the gas in the flue, the flow rate of the gas outside the flue, and the temperature and humidity after condensation. The display module displays the collected signals, allowing real-time monitoring of gas temperature and humidity. These modules work together to ultimately detect the CO2 concentration in the mixed gas inside the flue. The detection process is as follows: Figure 2 As shown.

[0066] First, sample gas collection and parameter detection are performed. This step mainly involves using a gas pump to collect gas from the flue to obtain sample gas. The sample gas contains CO2, N2, water vapor, and other gaseous components. Simultaneously, the temperature and humidity of the flue gas and the sample gas flow rate are monitored. Next, sample gas processing is performed. This primarily involves water vapor removal, requiring an electronic condenser and activated carbon. The electronic condenser removes most of the water vapor, but also lowers the sample gas temperature. A heating module is used to heat the sample gas back to its original temperature, maintaining consistent parameters. Next, spectral absorption detection is performed. The processed sample gas enters the gas chamber and is detected by a near-infrared spectrometer. Calibration, modeling, and testing are conducted to detect the CO2 concentration in the sample gas. Finally, concentration conversion is performed. An infrared detection module detects the sample gas in the gas chamber. An infrared light source is used to illuminate the gas through the entrance hole of the gas chamber, and light source information is collected at the exit hole. Data is acquired and converted to predict the CO2 concentration in the mixed gas. This embodiment uses the following... Figure 3a and Figure 3b The boiler flue gas CO2 concentration detection system shown was tested.

[0067] Figure 3b In this system, the flue gas temperature and humidity sensor and the sample temperature and humidity sensor in the adjacent gas chamber are both wall-mounted digital tube 485 type temperature and humidity transmitter sensors, model PR-300SMG-WS-N01; the flue gas flow meter is a Pitot tube; the sample flow meter near the electronic condenser is a remote gas flow meter; the electronic condenser is a CS-5A type electronic condenser; the infrared light source is a halogen tungsten lamp; the gas chamber is a reflective single gas chamber with a White cell structure, and both the light source inlet and outlet of the gas chamber are connected to the optical fiber via an SMA905 interface; the photoelectric sensor is a ULTRAMAT23 gas analyzer. A programmable logic controller (PLC) is used to control the entire system; the PLC is a Siemens S7-1500 PLC, model CPU1513-1PN. The heating wire is heated by a solid-state relay.

[0068] The electrical schematic diagram for the local control of the heating wire is shown below. Figure 3c See the simplified diagram of PLC control. Figure 3d See the external wiring diagram for Siemens S7-1500 PLC. Figure 3e The control flowcharts of the programmable logic controller for each part of this system are shown below. Figures 3f to 3j .

[0069] 1.3 Basic Theory of Infrared Absorption

[0070] Infrared absorption spectrum (IR), also known as molecular vibrational rotation spectrum, is a type of molecular absorption spectrum. Its basic theory is that when infrared light of continuously varying frequency irradiates a gas being tested, some frequencies of radiation are absorbed by the gas molecules, causing energy level transitions. The corresponding light intensity decreases, and the resulting curve containing information such as gas concentration is the infrared spectrum. The infrared region lies within the wavelength range of (0.8–1000) μm, including the near-infrared region (0.8–2.5) μm, the mid-infrared region (2.5–50) μm, and the far-infrared region (50–1000) μm.

[0071] The two conditions for infrared absorption are as follows: (1) the energy is the same, that is, the energy of the infrared radiation photon is the same as the energy required for the transition of the molecular vibrational energy level; (2) the dipole moment changes, that is, there is a coupling effect between radiation and matter.

[0072] The polarity of a molecule is generally characterized by its dipole moment μ = qd, where q is the charge at the positive and negative charge centers, and +q, -q, and d is the distance between the centers of the positive and negative charges. If the centers of the positive and negative charges in a molecule do not coincide, the atoms within the molecule will be in a state of continuous vibration around their equilibrium positions, and the distance d between the centers of the positive and negative charges will change accordingly. The dipole moment will also lengthen or contract accordingly; that is, the molecule has a definite frequency of dipole moment change. Considering a group with a changing dipole moment as a dipole, when the radiation frequency matches the dipole frequency, the molecule undergoes vibrational coupling interaction with the radiation, resulting in increased vibrational energy, increased amplitude, and energy level transitions. Fundamental and overtone peaks are formed due to energy level transitions. Both fundamental and overtone peaks are characteristic peaks for qualitative and quantitative analysis of gases. However, in the near-infrared region, overtone peaks are often used to analyze gas types and concentrations.

[0073] 1.4 Infrared Absorption Characteristics of CO2 Gas

[0074] In the field of infrared gas absorption spectroscopy, the high-resolution transmission molecular absorption database (HITR) is often used to calculate parameters such as the absorption intensity of gas molecules. By selecting the gas composition to be measured and the wavenumber range based on the parameters provided by the database, the absorption frequency band and absorption intensity can be viewed and analyzed. Figure 4a This is a curve showing the absorption peak of CO2 in the near-infrared band.

[0075] In experiments, the fundamental absorption peak of a gas is usually chosen as the frequency band for gas detection because its absorption is strongest, making it easier to detect changes in light intensity during measurement, thus improving the detection limit and accuracy. The fundamental absorption peak of CO2 is in the near-infrared band, around 1.95 μm and 1.6 μm, with a particularly noticeable absorption peak near 1.45 μm in the near-infrared.

[0076] When selecting the absorption peak of CO2, one should not only consider the absorption intensity of CO2, but also, based on the research background, whether other gases also have absorption peaks near the CO2 absorption peak. If multiple gas absorption peaks overlap, it will have a significant impact on the measurement results and be difficult to eliminate. The research background of this paper is the detection of CO2 in boiler flue gas. According to the data, the flue gas produced by boiler combustion contains SO2, NOx, CO, CO2, and water vapor, etc. Therefore, this embodiment, based on the research background of boiler flue gas detection, investigates and eliminates interfering gas absorption peaks, such as... Figure 4b and Figure 4c The figure shows the absorption peaks of different gas molecules and water vapor in the near-infrared band within the flue gas.

[0077] Near 1.45 μm, the absorption intensities of other gases and water vapor in the flue gas are significantly different from those of CO2. The infrared absorption of other gases has almost no impact on the CO2 detection in this study. In summary, CO2 exhibits strong absorption at 1.45 μm with virtually no interference from other gases; therefore, 1.45 μm was chosen as the frequency band for CO2 detection.

[0078] 1.5 Lambert-Beer Law

[0079] The Lambert-Beer law is a fundamental law used primarily to study the light absorption of gases. It forms the theoretical basis for the quantitative analysis of absorption spectra. (See...) Figure 4dA uniform gaseous medium is transmitted through a spectrum I0. After uniform absorption by the gaseous medium, a transmission spectrum is obtained, with the transmitted light intensity denoted as I. The absorbance A of the substance is defined as A = lgI0 / I. The magnitude of absorbance A is linearly related to the substance concentration c and the optical path length b, i.e., A = Kcb + ε. Here, K represents the extinction (or absorption) coefficient, the value of which is related to the properties of the substance and the wavelength of the incident light. When the substance being measured is a gas, the absorption cross section δ of the gas molecules can be used.

[0080] Let (λ) represent K, which indicates the extinction ability of gas molecules under certain temperature and pressure. In this case, the Lambert-Beer law can be expressed using the mathematical model shown in Equation 1.

[0081] I(λ)=I0(λ)×exp[-δ(λ)×C×L] (Formula 1);

[0082] In the formula, I0(λ) is the incident spectrum of the light source, I(λ) is the transmission spectrum after passing through the gas, C represents the gas concentration, L is the optical path length of the absorbing gas, and δ(λ) represents the absorption cross section of the gas. The optical path length L and the gas absorption cross section δ(λ) are constant values ​​and can be determined in advance by certain experiments. In the actual measurement process, the gas concentration can be derived by measuring the original spectral intensity of the light source and the transmission spectral intensity after passing through the gas using a spectrometer. This is the theoretical basis for measuring gas concentration using absorption spectroscopy. The concentration C of the gas to be measured is shown in Equation 2.

[0083]

[0084] Define the relative light intensity (I) of the gas being measured. r As shown in Equation 3.

[0085]

[0086] Substituting Equation 3 into Equation 2, the concentration of the analyte is obtained as shown in Equation 4:

[0087]

[0088] This allows us to obtain the concentration of the substance being measured.

[0089] Part Two: Construction of a Model for Detecting CO2 Concentration in Boiler Flue Gas Using Infrared Absorption

[0090] 2.1 Algorithm for CO2 Concentration Detection in Sample Gas

[0091] 2.1.1 CO2 concentration calibration

[0092] The entire experiment requires detecting the CO2 concentration, applying the Lambert-Beer law, and recording the incident spectrum I0(λ) and transmission spectrum I(λ) of the light source. In this embodiment, a gas chamber of a certain length is used for the experiment, so the optical path length L is known. The cross-sectional area of ​​the entire gas chamber, also called the gas absorption cross-section δ(λ), is a constant. The CO2 concentration is then calculated using the Lambert-Beer law.

[0093] During the experiment, five standard gases with CO2-free concentrations and concentrations close to 10-50% of the CO2 concentration in flue gas were used for multiple experiments. Multiple sets of relative light intensity data were collected, and the data were processed to calculate the average value. Experimental result curves were plotted to attempt to locate the absorption peak position of CO2 in the near-infrared band. Simultaneously, modeling was completed to predict CO2 concentration.

[0094] 2.1.2 CO2 Concentration Modeling Based on Least Squares Method

[0095] This embodiment requires infrared detection experiments on mixed gas samples of different concentrations. The experiment will acquire multiple sets of data, which need to be processed before CO2 concentration modeling is performed.

[0096] Based on the fundamental principles of the Lambert-Beer Law, concentration calculations are performed, and a least squares mathematical model is established to predict CO2 concentration. The least squares method finds the best function fit for the data by minimizing the sum of squared errors, and can reflect the basic trends in experimental data. The curve fitted using the least squares method does not necessarily pass through all experimental data points; it is only a method of calculating unknown quantities from a set of experimental data. Each curve has a corresponding prediction formula, which is helpful for subsequent CO2 concentration prediction experiments.

[0097] This embodiment utilizes the least squares method to fit the curve relationship between relative light intensity and concentration. The specific principle is as follows: Let a linear combination of a certain function be:

[0098] g(x)=c1 f1(x)+c2 f2(x)+c3 f3(x)+…+c n f n (x);

[0099] Where f1(x), f2(x), f3(x)...f n Let (x) be a known function, and c1, c2, c3…c n These are coefficients to be determined.

[0100] Assume that the measured data (x1, y1), (x2, y2), (x3, y3)...(x n ,y n Then, the following linear equation can be established: A × c = y; where,

[0101] c = [c1, c2, ..., c n ] T ;

[0102] By obtaining the least squares solution from the equation, the required fitting equation can be derived.

[0103] 2.1.3 CO2 Concentration Modeling Based on Backpropagation Neural Network

[0104] In experiments detecting CO2 concentration in exhaust gas, environmental factors such as temperature and humidity are involved. A neural network approach can be considered for CO2 concentration modeling. The experiment can acquire temperature, humidity, and light intensity data, which can be used to build a three-input, one-output neural network model to calculate the CO2 concentration in the flue gas.

[0105] Neural networks possess non-linear mapping capabilities, and their modeling process does not require a precise mathematical model of the independent variables, making them suitable for building multi-input, multi-output models. This embodiment establishes a three-input, one-output neural network model, facilitating computation and making it easier to implement model calculations using software. First, the input sample data is trained to establish the hidden layer patterns of the neural network. Then, the CO2 concentration is calculated using the patterns established by the hidden layers, yielding the concentration value. The neuron model of the artificial neural network is as follows: Figure 5 The CO2 concentration is calculated using a neural network model. The calculation process is as follows: Figure 6 .

[0106] This embodiment utilizes MATLAB to establish a BP mathematical model, using temperature, relative humidity, and light transmittance as inputs, and concentration as the output, constructing a three-input, one-output BP neural network. Twenty experiments were conducted using gases with CO2 volume fractions ranging from 0% to 50%, obtaining a total of 120 sets of data. 108 sets were selected as test data to train the model, and the remaining 12 sets were used for testing to verify the accuracy of the established mathematical model.

[0107] 2.2 CO2 Concentration Conversion

[0108] A CO2 concentration detection experiment was conducted by simulating the humidity of a flue gas environment, yielding new data. Based on the Lambert-Beer law, the CO2 concentration in the dried sample gas was re-modeled and calculated. Since most of the water vapor was filtered out during the pretreatment stage, the total volume changed, necessitating volume conversion.

[0109] 2.2.1 CO2 Concentration Conversion Method

[0110] Volume fraction conversion mainly includes the volumes of three parts: water volume, dry sample gas volume, and total volume. The volume fraction of a gas refers to the percentage of a specific component gas in the total volume of a gas mixture. CO2 is an important component of flue gas, and the volume fraction of CO2 in coal-fired boiler flue gas can be as high as 15%. The experiment uses a CO2 volume fraction of 10%-50%, which is relatively close to the CO2 volume fraction in the flue gas, facilitating the location of the CO2 absorption peak and completing the calibration experiment, laying the foundation for the detection of CO2 concentration in flue gas. Based on the relative light intensity data obtained from infrared detection, a mathematical model is established using relevant mathematical modeling methods. This embodiment compares the least squares method and the BP neural network modeling method, selecting the method with smaller experimental error and higher accuracy for modeling. Finally, the concentration of CO2 in the flue gas is obtained through volume fraction conversion.

[0111] 2.2.2 Factors Affecting CO2 Concentration Conversion

[0112] The concentration of CO2 in flue gas is primarily affected by factors such as temperature, flow rate, and humidity. First, calibration experiments are conducted at ambient temperature and pressure to measure different CO2 concentrations, providing a concentration reference for flue gas CO2 concentration detection. Then, the influence of environmental factors on the experiment is considered.

[0113] Before designing the experimental system, the entire experimental process needs to be analyzed to achieve online detection of CO2 concentration and demonstrate its feasibility from a theoretical perspective. Based on this theoretical foundation, the effects of temperature, flow rate, humidity, and similar gases on CO2 concentration detection must be considered. Therefore, when designing the detection system, variables need to be controlled and influencing factors reduced to accurately measure the CO2 concentration in the exhaust gas as much as possible.

[0114] The laboratory uses standard gas for experiments, eliminating interference from other gases. The biggest influencing factor in CO2 concentration detection is water vapor. The design process focused on minimizing the impact of water vapor, i.e., humidity, on concentration detection. Furthermore, temperature also affects the overall experimental data and the calculation of CO2 concentration in the mixed gas; therefore, temperature was controlled in the experiment. A flow meter was used to maintain a consistent flow rate for the mixed sample gas, reducing the impact of flow rate on the experiment. In summary, the CO2 concentration detection experiment controls temperature and flow rate to reduce interference before conducting the humidity experiment.

[0115] 2.2.3 CO2 Concentration Conversion Model

[0116] The mixed gas contains the sum of the volumes of water vapor and CO2. During the experiment, the water vapor needs to be condensed, which will separate the water vapor from the mixed gas before subsequent detection. The relationship between the sample gas volume, the mixed gas volume, and the water vapor volume is: V 混 =V 水 +V 样 ;

[0117] The volume of the mixed gas is calculated based on the flow rate detected by the flow meter: V 混 =S×t; where S is the flow rate of the mixed gas and t is time.

[0118] The volume of water vapor needs to be calculated based on the relative humidity of the gas mixture. First, calculate the mass of the water vapor at the absolute humidity, then calculate the volume V of water in the water vapor based on the relative volume fraction of water. 水 Finally, the CO2 concentration of the flue gas can be calculated.

[0119] V 水 =n×V m ;

[0120] In the above formula, m is the mass of water vapor at absolute humidity, M is the molar mass of water, and n is the amount of substance of water; Vm is the molar volume of water under standard conditions, Vm = 22.4 L / mol; C0 is the concentration of the sample gas after drying, which can be predicted based on calibration modeling; V in the formula 总 equals V 混 .

[0121] Based on the above derivation process, the CO2 concentration of the mixed gas in the flue can be obtained using MATLAB programming based on the data obtained from the experimental system.

[0122] 2.2.4 CO2 Conversion Program Design

[0123] The collected flue gas flow rate is first detected using a flow meter to calculate the mixed gas volume. Temperature and humidity sensors upload these values ​​to a PLC, which performs normalization processing to obtain the temperature and humidity data. A spectrometer collects relative light intensity data, and Lambertian-Beer spectrophotometry is used for data preprocessing to obtain the dry sample gas concentration. Finally, the CO2 concentration in the mixed sample gas is calculated using an established mathematical model and volume conversion. The overall design is as follows: Figure 7 As shown.

[0124] Part Three: Verification Experiments and Results Analysis

[0125] 3.1 Experimental Scheme for Boiler Flue Parameter Testing

[0126] In the overall experimental process of this embodiment, sensors are first used to monitor experimental environmental parameters to eliminate environmental factors that may affect the experimental results. The main monitoring parameters are temperature and humidity inside the flue, temperature and humidity after condensation, and gas flow rate. The temperature, flow rate, and gas sampling time are kept consistent before and after condensation in each experiment. Humidity is the primary parameter monitored, and its volume is finally converted to water volume to determine the true CO2 concentration inside the flue.

[0127] Temperature and humidity sensors and flow meters inside and outside the flue gas were used to collect parameters, and environmental parameters were monitored at multiple points multiple times. Calibration experiments were conducted at normal temperature and pressure to verify the feasibility of the system. A simulated CO2 detection experiment was performed in the flue gas exhaust, focusing on the effect of humidity on concentration. A humidification device was designed, and the mixed sample gas was humidified before being introduced into the detection chamber. Each experiment was timed for 5 minutes, and changes in relative light intensity were monitored in real time on a computer. Particular attention was paid to recording the relative light intensity at the absorption peak position for subsequent calculations.

[0128] 3.2 Sample Gas Detection and Calibration Experiment

[0129] 3.2.1 Testing Plan

[0130] When calibrating the CO2 gas concentration of this system, calibration experiments were conducted at room temperature and pressure using standard gases with CO2 concentrations ranging from 0% to 50%. For each concentration, the relative light intensity at different wavelengths was obtained for each experiment. Taking the CO2 concentration closest to the flue gas exhaust gas of 20% as an example, the relative light intensity data at the absorption peak from 0% to 50% are shown in Table 1.

[0131] Table 1

[0132]

[0133] The test indicators include temperature and humidity, repeatability testing, and stability testing. Multiple experiments were conducted on gases of different mixed concentrations using the experimental system. Analysis of the experimental data clearly identified the location of the CO2 absorption peak, and data processing was performed to plot the experimental results.

[0134] like Figure 8 As shown, a distinct absorption peak is clearly visible around 1.45 μm. Furthermore, the absorption peak at different concentrations shows a linear change in energy level, with more light energy being absorbed and the energy value decreasing as the CO2 concentration increases. Consulting the HITR database confirms the presence of an absorption peak for CO2 around the 1.45 μm wavelength in the near-infrared region, proving that the data detected by the constructed experimental system is indeed CO2 absorption data. Concentration modeling using the experimental data can provide a basis for the subsequent design of an online system.

[0135] 3.2.2 Least Squares Method Experiment and Result Analysis

[0136] Based on the experimental data, a model was built using the Lambert-Beer law. The least squares method was used to derive the concentration prediction formula. The resulting curve after data processing is shown below. Figure 9 As shown. The prediction model formula is:

[0137] C0 = -0.00031385I + 5.8546;

[0138] In the formula: C0 is the gas concentration; I is the experimentally measured relative light intensity.

[0139] Using MATLAB programming, the correlation coefficient of the fitted curve was calculated, yielding R = 0.9987, which is close to 1, indicating a good fit and suitable for subsequent CO2 concentration prediction. Substituting the relative light intensities detected in experiments at different concentrations into the prediction model formula, the predicted CO2 concentration values ​​are shown in Table 2. Using the least squares method, tests were conducted on the system with standard gases of different concentrations, and the relative error values ​​of the predicted concentrations under different concentration experiments were calculated, with a maximum of 8.7% and a minimum of 2.1%.

[0140] Table 2

[0141]

[0142] 3.2.3 Neural Network Experiment and Result Analysis

[0143] Standard gases containing 0%-50% CO2 by volume were used for detection. Twenty experiments were conducted for each concentration, yielding 120 sets of experimental data. A three-input, one-output model was built using a neural network. Temperature and humidity data from each experiment were collected, and combined with the final transmitted light intensity as the three inputs to build the model and calculate the final sample gas concentration. Eighteen sets of data for each concentration, totaling 102 sets, were used to train the model. The remaining two sets of data for each concentration were used to validate the detection model. The neural network model was used to calculate the concentrations for six different concentrations, and the prediction results are as follows: Figure 10 As shown, the relative error values ​​of the predicted concentrations under different concentration experiments were calculated using the BP neural network algorithm, with a maximum of 9% and a minimum of 5.52%.

[0144] Table 3

[0145]

[0146] Both CO2 concentration detection algorithms tested showed errors within 10%, indicating small experimental errors. This verifies the feasibility of the system and is beneficial for the subsequent design of an online CO2 monitoring system for flue gas. A comparison of the least squares method and the BP neural network algorithm, combined with experimental results from the prediction model, revealed that the relative error of the neural network is slightly larger than the minimum error of the least squares method. Therefore, the least squares method will be used for subsequent experimental modeling.

[0147] 3.3 Simulation Detection Experiment of CO2 Concentration in Boiler Flue Gas

[0148] The previous experiment involved testing the standard gas and calibrating the sample gas. The entire experiment proceeded under ideal conditions, free from water vapor interference. This section's experiment requires conducting simulated flue gas testing, building upon the standard gas testing.

[0149] 3.3.1 Experimental Scheme

[0150] To simulate flue gas emissions, a mixture of standard gas and water vapor was used for testing. In this experiment, a container filled with a certain amount of water was mixed with the standard gas and transferred, thus increasing the relative humidity. The relative humidity was controlled at RH (%) = 50% before condensation and at RH (%) = 23% after condensation. The entire system was operated automatically by a PLC.

[0151] First, the flow rate of the sample gas at the outlet of the gas tank was adjusted and monitored using a flow meter. The transfer time was set to 5 minutes, and the gas transfer rate was 200 mL / min. A stopwatch was used to time the process, and the total volume of the mixed gas could be calculated using the formula mentioned earlier. Then, the temperature and humidity of the gas before and after condensation were recorded. The CO2 concentrations of the experimental mixed sample gas were 10%, 20%, 30%, 40%, and 50%. A PLC controller was used to control the heating rod to heat the dried sample gas after condensation, raising the temperature to the level before condensation to reduce the influence of temperature on the experiment. Finally, the dried and heated sample gas entered the gas chamber, and infrared detection was performed to obtain relative light intensity data. A new model was then created and used for CO2 concentration calculation. Figure 11 The diagram shows the experimental setup for the effect of humidity. Standard gas is passed into a container filled with water. The sample gas is then humidified and passed into the gas chamber, thus increasing the humidity of the gas and simulating the humidity environment of flue gas. The gas flow rate is measured as follows: Figure 12 As shown. The gas transmission flow rate was between 198 and 202 mL / min, and the error was within the allowable range. The temperature and humidity curves detected by the temperature and humidity sensor are shown below. Figure 13 , Figure 14 As shown. The sensor detected the gas temperature before condensation as 19.8℃, which differs from the experimental ambient temperature of 20℃ by 0.2℃. The sensor detected the relative humidity of the gas before condensation as RH(%) = 49.8%, which differs from the set experimental ambient relative humidity as RH(%) = 50% by 0.2%. The temperature and humidity errors are within the allowable range.

[0152] 3.3.2 Experimental Research and Results Analysis

[0153] Based on the experimental data after humidification and the plotted experimental result curve Figure 15 It is evident that within the same wavelength range, the absorption peak position remains unchanged due to the influence of humidity, while the relative light intensity decreases. Table 4 shows the relative light intensity data at the absorption peak after adding humidity, with values ​​ranging from 10% to 50%.

[0154] The table shows that, compared with previous results, the relative light intensity decreased due to the absorption of light by humidity. Based on the data obtained from the flue gas simulation experiment, and using the Lambert-Beer law combined with the least squares method, the concentration prediction formula after humidification was derived. The predicted binary linear model formula is: C0 = 11.0178RH - 0.0331I, where: C0 is the concentration of the dry sample gas; RH is the experimentally measured relative humidity; and I is the experimentally measured relative light intensity.

[0155] Table 4

[0156]

[0157] The experiment yields data such as relative humidity, flow rate, and relative light intensity. Using this data and a binary linear prediction formula, the CO2 concentration C0 of the dried sample gas can be directly calculated. Furthermore, the volume of each component can be derived from the volume formula, as stated in the previous formula C = (V... 样 ×C0) / V 总 The CO2 concentration C in the mixed gas sample can be calculated. The experimentally measured CO2 concentrations are shown in Table 5.

[0158] Table 5

[0159]

[0160] Experiments were conducted using mixed gas samples of different concentrations to simulate flue gas. The relative light intensities obtained are shown in the table above. Calculations show the predicted concentration, calculated concentration, and relative error. Due to the influence of water vapor, the maximum relative error was 9%, and the minimum was 1.16%, which accurately verifies the feasibility of the experimental system.

Claims

1. A method for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas, characterized in that, Includes the following steps: Step (1): Calibration experiments were conducted on standard gases with different CO2 concentrations at room temperature and pressure to obtain the relative light intensity at different wavelengths; the detection frequency band for infrared detection of CO2 concentration in boiler flue gas was 1.45 μm. Step (2): The experimental data obtained in the calibration experiment in step (1) are processed using different analysis and processing methods to obtain standard gas CO2 concentration prediction models. The prediction results of different standard gas CO2 concentration prediction models are compared to select the analysis and processing methods of the experimental data. Step (3): Mix the standard gas with water vapor to prepare mixed sample gases with different CO2 concentrations; Step (4): After the mixed sample gas is dehumidified by condensation, its relative humidity is controlled to reach the target value and then infrared detection is performed to obtain the relative light intensity of the mixed sample gas with different CO2 concentrations after dehumidification; the relative humidity of the mixed sample gas with different CO2 concentrations after dehumidification is controlled at 22-24%; and the temperature of the mixed sample gas is controlled to be the same as the temperature before dehumidification when infrared detection is performed. Step (5): The experimental data obtained in step (4) is analyzed and processed using the experimental data analysis and processing method selected in step (2) to obtain the CO2 concentration prediction model of the mixed sample gas after dehumidification; the CO2 concentration prediction model of the mixed sample gas after dehumidification is: C0=11.0178 RH -0.0331 I C0 represents the CO2 concentration in the dehumidified mixed sample gas. RH The relative humidity is measured by the sensor before the mixed gas sample is dehumidified; I To measure the relative light intensity in the experiment; Step (6): Calculate the CO2 concentration of the boiler flue gas tail gas according to the CO2 concentration prediction model of the mixed sample gas after dehumidification; The formula for calculating the CO2 concentration in boiler flue gas is: C=( V 样 ×C0) / V 总 ; C represents the CO2 concentration in the boiler flue gas. V 总 The total volume of the mixed sample gas. V 样 C0 represents the remaining volume of the mixed sample gas after dehumidification; C0 represents the CO2 concentration of the mixed sample gas after dehumidification.

2. The method for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas according to claim 1, characterized in that, In step (1), the CO2 concentration in the standard gas is 0, 10%, 20%, 30%, 40% and 50%; the water vapor content in the standard gas is 0.

3. The method for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas according to claim 2, characterized in that, In step (2), the analysis and processing methods are least squares method and neural network model.

4. The method for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas according to claim 3, characterized in that, In step (2), when constructing the standard gas CO2 concentration prediction model using a neural network model, the number of experiments for each concentration of standard gas is 20. The temperature, humidity and light intensity transmitted by light in each experiment are used as three input variables, and the CO2 concentration is used as the output variable to build the model and calculate the final sample gas concentration. 18 sets of data are selected for each concentration, for a total of 102 sets of data for model training. The remaining 2 sets of data for each concentration are used to verify the detection model. When constructing the standard gas CO2 concentration prediction model using the least squares method, the number of experiments for each concentration of standard gas is greater than or equal to 5.

5. The method for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas according to claim 3, characterized in that, In step (2), when constructing the standard gas CO2 concentration prediction model using the least squares method, the fitting equation of the prediction model is assumed to be: g ( x )= c 1 f 1( x )+ c 2 f 2( x )+ c 3 f 3( x )+…+ c n f n ( x )； in, f 1( x ), f 2( x ), f 3( x … f n ( x () is a known function. c 1, c 2, c 3… c n Undetermined coefficients: ( x 1 , y 1 ), ( x 2 , y 2 ), ( x 3 , y 3 )…( x n , y n If ) represents the data measured in the experiment, then: A× c = y ； ； ； ； Thus, a prediction model for standard gas CO2 concentration was calculated.

6. The method for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas according to claim 5, characterized in that, In step (3), the CO2 concentration in the mixed sample gas is 0, 10%, 20%, 30%, 40% and 50%; the relative humidity of the mixed sample gas with different CO2 concentrations is 50%.

7. An infrared absorption detection model for CO2 concentration in boiler flue gas, characterized in that, The method described in claim 1 for constructing an infrared absorption detection model for CO2 concentration in boiler flue gas is used; that is, the formula for calculating CO2 concentration in boiler flue gas is: C = ( V 样 ×C0) / V 总 ; C represents the CO2 concentration in the boiler flue gas. V 总 The total volume of the mixed sample gas. V 样 C0 represents the remaining volume of the mixed sample gas after dehumidification; C0 is the CO2 concentration prediction model of the mixed sample gas after dehumidification obtained in step (5). The CO2 concentration prediction model of the mixed sample gas after dehumidification is as follows: C0=11.0178 RH -0.0331 I C0 represents the CO2 concentration in the dehumidified mixed sample gas. RH The relative humidity is measured by the sensor before the mixed gas sample is dehumidified; I The relative light intensity was measured in the experiment.

Images