System and method for detecting gas concentration by adopting dual-wavelength laser under influence of particulate matters
Through the dual-wavelength laser detection system and convolutional neural network algorithm, the inaccuracy problem of gas concentration detection under particulate matter interference in the prior art is solved, more accurate gas concentration measurement is achieved, and better environmental monitoring data support is provided.
Patent Information
- Application Number
- CN202411105188.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-22
AI Technical Summary
Existing gas monitoring technologies have problems such as inaccurate detection, difficult to operate and low flexibility in detecting waste gases generated by coal combustion, especially in the case of particulate matter interference.
The dual-wavelength laser detection system is used to measure the particle concentration and the tunable gas laser to measure the gas concentration through TOF lidar. The convolutional neural network algorithm is used to correct the gas concentration at different particle concentrations, and multiple convolutional neural network models are constructed to realize the Accurate measurement.
It improves the accuracy and reliability of gas concentration detection, provides better support for environmental pollution control data, and can accurately detect gas concentration under multi-factor interference, especially under particulate interference.
Smart Images

Figure CN119000604A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of environmental monitoring, and in particular, relates to a system and method for detecting gas concentration using dual-wavelength lasers under the influence of particulate matter. Background Art
[0002] Gas detection from coal combustion is a vital part of environmental protection and industrial regulation. Modern coal combustion produces a variety of pollutants, such as sulfur dioxide, nitrogen oxides, particulate matter, etc. These pollutants are potentially harmful to air quality and human health. Therefore, it is particularly important to accurately and timely monitor harmful gases emitted by coal combustion.
[0003] Among the existing gas monitoring technologies, commonly used methods include online monitoring and intermittent monitoring. Online monitoring is to monitor the concentration and emission of pollutants in the exhaust gas in real time through sensors and monitoring equipment installed at the exhaust port. These devices can measure the concentration of different pollutants, such as sulfur dioxide, nitrogen oxides and particulate matter, and transmit the data to the central monitoring system for real-time analysis and recording. Intermittent monitoring is to evaluate the pollutant content of exhaust gas emissions by regularly taking samples and sending them to the laboratory for chemical analysis.
[0004] However, the above methods are almost all detected by electrochemical methods, and all have problems such as inaccurate detection, difficult operation, and low flexibility. In view of the above situation, the present invention discloses a gas concentration detection system under the interference of particulate matter. The present invention corrects the gas concentration under different particulate matter concentrations through a neural network algorithm, thereby obtaining more accurate and reliable gas concentration data, and ultimately achieving the purpose of providing better data support for environmental pollution control. Summary of the invention
[0005] The present invention needs to measure the particle concentration and gas concentration at the same time, which is achieved by emitting lasers of two different wavelengths. Since the laser wavelength for measuring particle concentration and the laser wavelength for measuring gas are quite different, emitting these two wavelengths of lasers at the same time will not affect the measurement of particles and gases. The wavelength for measuring particle concentration is the fixed wavelength of the TOF laser radar, and the wavelength for measuring gas concentration is the wavelength near the gas absorption peak of the relevant gas. Since particulate matter has a great influence on gas measurement, we classify the particle concentration and build different convolutional neural network models for different particle concentration ranges to achieve accurate measurement of gas concentration under different particle concentrations.
[0006] To achieve the above object, the technical solution of the present invention is as follows:
[0007] The gas concentration prediction system, method and model training method are mainly composed of a gas concentration training method, a test method and a gas concentration detection system under the influence of particulate matter. The present invention aims to detect gas concentration under the interference of multiple factors. The interference mentioned above mainly refers to the concentration of particulate matter, and the detected gas is mainly SO2;
[0008] Furthermore, a total of K particle concentrations are obtained, and the K particle concentrations are divided into J categories from small to large, so as to obtain J types of particle concentration information; J particle concentration sets are constructed, marked as k 1 , k 2 , ..., k J , where k 1 , k 2 , ..., k J Each contains m j particle concentration information, j∈[1,J] and m 1 +m 2 +...+m J =K;
[0009] Furthermore, the particle concentration set (k j ), for each particle concentration set m j The particle concentration information is at the mth j The air chamber containing the particle concentration information is filled with a concentration of c i The gas, where the subscript i represents the gas concentration, is obtained by tunable gas laser at the mth gas concentration. j The i-th gas concentration under the interference of particle concentration
[0010] Furthermore, for each category m j The particle concentration is set to m, and the different gas concentration data under each particle concentration are repeatedly collected to obtain the j The particle concentration is c i Repeat the experiment N times in the air chamber and obtain j Input matrix for particle conditions It is expressed as:
[0011]
[0012] in, Medium j Indicates the mth j particle concentration condition, c i Indicates that in the mth ji different concentrations of gas are filled under the conditions of i particle concentrations, N represents the number of repeated data collection under the same gas concentration and particle concentration, Indicates that the Nth collection is performed at a gas concentration of c I , the particle concentration is m j The intensity value at the gas absorption wavelength at ;
[0013] For J particle concentration categories, corresponding to J gas concentration input matrices, the gas concentration matrices of each category are Input the corresponding neural network model for training;
[0014] Furthermore, for different categories, a convolutional neural network model is trained, which includes an input layer, a convolutional layer, an activation layer, a pooling layer, and a fully connected layer;
[0015] Furthermore, the input layer is used to extract the gas concentration characteristics under different particle concentrations, and input different gas concentration data under different particle concentrations. And the corresponding gas concentration true value data c i , convert the input data into feature matrices and feature vectors that can be recognized by the network;
[0016] Furthermore, the convolutional layer operation formula is:
[0017]
[0018] Among them, i, j represent the spatial position of the output, s represents the sth convolution kernel, m, n represent the spatial position of the convolution kernel, and c represents the number of input channels;
[0019] Further use the ReLU activation function, the formula is as follows:
[0020] f(x) = max(0, x)
[0021] Where x is the input value and f(x) is the output of the function;
[0022] Furthermore, the pooling layer uses the maximum pooling to reduce the dimension, and the output of the pooling layer is:
[0023]
[0024] Furthermore, the output of the fully connected layer is:
[0025]
[0026] Among them, i represents the input node of the pooling window, and k represents the output node.
[0027] Furthermore, the trained neural network model is integrated into the real-time gas concentration detection system to predict the actual concentration data through the real-time collected particle concentration data and gas concentration data;
[0028] Furthermore, the gas concentration detection system is a dual-wavelength gas concentration detection system structure based on TOF laser radar and tunable gas laser, which mainly includes a particle concentration detection module and a gas concentration detection module;
[0029] Furthermore, by emitting a wavelength of λ 1 The laser, based on the TOF ranging principle, obtains the relationship between distance and concentration, thereby obtaining the particle concentration.
[0030] Furthermore, by emitting a wavelength of λ 2 The laser intensity at the gas absorption wavelength is obtained
[0031] Furthermore, the particle concentration data of the area to be tested is obtained through the above module. and the intensity value of the gas absorption wavelength under the influence of particles
[0032] Further, the particle concentration is determined According to the concentration category, select the corresponding gas concentration prediction model;
[0033] Furthermore, the gas concentration data of the area to be measured is Input the gas concentration prediction model described above;
[0034] Further, the gas concentration prediction value in the test area output by the gas concentration prediction model is obtained.
[0035] Furthermore, the TOF laser radar particle concentration detection system includes the following components:
[0036] Transmitter module and receiver module;
[0037] Furthermore, the ToF laser emission module includes the following components:
[0038] Light source module: includes laser emitting tube, laser emitting board and radar main control board, providing the basic light source for smoke aerosol detection system. The laser emitting tube emits light source signal, and the laser emitting board diffuses and propagates the light;
[0039] Laser driving circuit: used to provide a stable driving voltage suitable for the TOF laser radar to transmit optical signals;
[0040] Sensor interface: provides power to the sensor, required bias voltage / current signals, digital control phase, and reads data stream from the sensor;
[0041] Furthermore, the ToF laser receiving module includes the following components:
[0042] Laser receiving lens: collects the light reflected from the object being measured;
[0043] Detection system photosensitive element: receives reflected light;
[0044] Sensor interface: used for communication with the outside (to PC or processing unit), transmitting the read distance information and concentration information to the storage and analysis module for processing;
[0045] Furthermore, the difference between the measured distance and the true distance is established as follows and particle concentration Relationship:
[0046] a) The fixed wavelength signal with a modulated signal emitted by the laser transmitting module passes through the standard air chamber, and the receiving module obtains the corresponding distance information d 0 ;
[0047] b) A fixed wavelength signal with a modulated signal emitted by the laser emission module is passed through the detection chamber, and the receiving module obtains the corresponding distance information d;
[0048] c) Get the difference between the measured distance and the actual distance
[0049] d) obtaining the particle concentration f at that time through the particle concentration sensor in the air chamber; wherein different concentrations of particles are generated by burning cigarette cakes of different weights;
[0050] e) Through polynomial fitting, establish The relationship between the particle concentration and the distance difference is obtained. Thus, the particle concentration data is calculated through the measurement distance of TOF;
[0051] Furthermore, the gas concentration detection module is implemented by a tunable semiconductor laser and includes the following components:
[0052] Main control module: controls the normal operation of the entire system and is responsible for calculating chimney exhaust emission information;
[0053] Semiconductor laser: using integrated semiconductor laser and laser absorption spectroscopy to measure the amount of chimney exhaust gas;
[0054] Detection device: used to receive the optical power absorbed by the gas to be measured;
[0055] Lock-in amplifier: completes the extraction of the second harmonic signal of the detection signal;
[0056] Further, the exhaust gas concentration is detected according to the following method. The present invention uses a semiconductor laser and laser absorption spectroscopy technology to measure the gas amount (SO2);
[0057] Furthermore, the semiconductor laser uses the laser absorption spectrum of sulfur dioxide near 7.4 μm as the central wavelength to detect the concentration of the gas, and the laser output port includes a focusing lens, which focuses the laser power after full absorption on the detection device to ensure the signal quality;
[0058] Furthermore, the semiconductor laser includes a signal generator that can generate a high-frequency sinusoidal signal and a low-frequency scanning signal. The sinusoidal signal realizes high-frequency modulation of the laser signal, reduces the interference of low-frequency noise and provides a reference signal for the phase-locked amplifier. The phase-locked amplifier module obtains information about the change of gas concentration over time by inverting the ratio of the second harmonic and the first harmonic intensity of the corresponding gas absorption peak;
[0059] Furthermore, the trained convolutional neural network model is integrated into the analysis module, and the gas concentration prediction results are saved in the storage module. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] Figure 1 It is a schematic diagram of the structure of a system for detecting gas concentration using dual-wavelength laser under the influence of particulate matter of the present invention;
[0061] Figure 2 It is a schematic diagram of the structure of the gas concentration detection system under a certain particle concentration;
[0062] Figure 3 Construct a flow chart for the distance versus particle concentration relationship;
[0063] Figure 4 A flow chart of the training method of the gas concentration prediction model under the influence of particulate matter; DETAILED DESCRIPTION
[0064] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0065] like Figure 1 As shown, a dual-wavelength laser detection system for gas concentration under the influence of particulate matter. The particulate matter concentration detection module emits a wavelength of λ 1 The laser, based on the TOF ranging principle, obtains the relationship between distance and concentration, thereby obtaining the particle concentration. d represents the detected distance information; the gas concentration detection module emits a wavelength of λ2 The laser intensity at the gas absorption wavelength is obtained
[0066] More specifically, the analysis module determines the particle concentration The concentration category is selected, and the corresponding gas concentration prediction model is selected to calculate the intensity value of the gas absorption wavelength in the measured area. Input the corresponding gas concentration prediction model and output the gas concentration prediction value in the test area output by the gas concentration prediction model. Store in the storage module for subsequent research;
[0067] More specifically, ToF laser radar is used to detect particle concentration. In environments with different particle dispersions, ToF laser radar is used to detect distance information by emitting a wavelength of λ 1 The particle concentration is detected by laser. By measuring the difference between the distance and the actual distance, a particle concentration model is established and the concentration information of the particle is obtained through analysis.
[0068] More specifically, if Figure 2 As shown, the gas concentration information under different particle concentrations is collected according to the following method: a certain concentration of gas is filled into a gas chamber with a certain particle concentration, and then a semiconductor laser of a certain wavelength is used in combination with laser absorption spectroscopy technology to detect the gas concentration, and the collected gas concentration data and the corresponding particle concentration data are stored in the system for subsequent analysis;
[0069] More specifically, a gas concentration detection system based on an adjustable laser includes a main control module, a semiconductor laser, a detection device, and a lock-in amplifier;
[0070] More specifically, the semiconductor laser uses the laser absorption spectrum of sulfur dioxide near 7.4 μm as the central wavelength to detect the concentration of gas. The laser output port includes a focusing lens, which focuses the laser power after full absorption on the detection device to ensure signal quality;
[0071] More specifically, the semiconductor laser includes a signal generator that can generate a high-frequency sinusoidal wave signal and a low-frequency scanning signal. The sinusoidal wave signal realizes high-frequency modulation of the laser signal, reduces the interference of low-frequency noise and provides a reference signal for the phase-locked amplifier. The phase-locked amplifier module obtains information on the change of gas concentration over time by inverting the ratio of the second harmonic and the first harmonic intensity of the corresponding gas absorption peak, thereby obtaining the gas concentration.
[0072] More specifically, if Figure 3 As shown, the difference between the measured distance and the actual distance is established as follows and particle concentration Relationship:
[0073] a) The fixed wavelength signal with a modulated signal emitted by the laser transmitting module passes through the standard air chamber, and the receiving module obtains the corresponding distance information d 0 ;
[0074] b) A fixed wavelength signal with a modulated signal emitted by the laser emission module is passed through the detection chamber, and the receiving module obtains the corresponding distance information d;
[0075] c) Get the difference between the measured distance and the actual distance
[0076] d) obtaining the particle concentration f at that time through the particle concentration sensor in the air chamber; wherein different concentrations of particles are generated by burning cigarette cakes of different weights;
[0077] e) Through polynomial fitting, establish The relationship between the particle concentration and the distance difference is obtained. Thus, the particle concentration data is calculated through the measurement distance of TOF;
[0078] More specifically, if Figure 4 As shown, the gas concentration prediction model is trained according to the following method.
[0079] The preferred embodiment of the present invention has been shown in detail above, but the specific description of the embodiment is only used to assist in understanding the method of the present invention and its core idea, rather than to limit it. For those skilled in the art, it is still possible to modify or change the aforementioned embodiments, or to replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for training a gas concentration prediction model under the influence of particulate matter, characterized in that: include: (1) A total of K particle concentrations are obtained, and the K particle concentrations are divided into J categories from small to large, thereby obtaining J types of particle concentration information; J particle concentration sets are constructed and marked as k1, k2, ..., k J , where k1, k2, ..., k J Each contains m j particle concentration information, j∈[1,J] and m1+m2+...+m J =K; (2) The particle concentration set (k j ), for each particle concentration set m j Information on particle concentration, in m j The air chamber containing the particle concentration information is filled with a concentration of c i The gas of the mth gas is obtained by tunable gas laser. j The i-th gas concentration under the interference of particle concentration (3) For each category m j The mth particle concentration is obtained by repeatedly collecting different gas concentration data at each particle concentration. j At the particle concentration of m, the experiment is repeated N times for the gas chamber with a gas concentration of ci. j Input matrix under particle conditions It is expressed as: in, Medium j Indicates the mth j particle concentration condition, c i Indicates that in the mth j i different concentrations of gas are filled under the conditions of i particle concentrations, N represents the number of repeated data collection under the same gas concentration and particle concentration, Indicates that the Nth collection is performed at a gas concentration of c I , the particle concentration is m j The intensity value at the gas absorption wavelength at ; For p particle concentration categories, corresponding to p gas concentration input matrices, the gas concentration matrices of each category are Input the corresponding neural network model for training; (4) Integrate the trained neural network model into the real-time gas concentration detection system and predict the actual concentration data through the real-time collected particle concentration data and gas concentration data.
2. A method for predicting gas concentration under the influence of particulate matter, characterized in that: include: (1) Structure of dual-wavelength gas concentration detection system based on TOF laser radar and tunable gas laser: ① Particle concentration detection module: By emitting a laser with a wavelength of λ1, based on the TOF ranging principle, the relationship between distance and concentration is obtained, thereby obtaining the particle concentration. ② Gas concentration detection module: By emitting a laser with a wavelength of λ2, the intensity value at the gas absorption wavelength is obtained (2) The concentration prediction method is as follows: ① Obtain the particle concentration data of the area to be tested and the intensity value of the gas absorption wavelength under the influence of particles ② Determine the particle concentration According to the concentration category, select the corresponding gas concentration prediction model; ③ The gas concentration data of the area to be measured Inputting the gas concentration prediction model described in claim 1; ④ Obtain the predicted value of gas concentration in the test area output by the gas concentration prediction model 3. The m-th method according to claim 1 j Input matrix under particle conditions Its characteristics are: By applying voltage to both ends of the laser fiber, after the WMS (wavelength modulation spectrum) module, the photodetector output signal q(l) is obtained after spatial optical path attenuation and gas absorption, where q(l)<q(l+1) and q(l)<q(l-1), thereby obtaining the signal intensity output vector of different gas concentrations And let p = q(l).
4. The convolutional neural network model according to claim 1, comprising an input layer, a convolutional layer, an activation layer, a pooling layer, and a fully connected layer, specifically as follows: (1) Input layer: used to extract gas concentration characteristics under different particle concentrations and input different gas concentration data under different particle concentrations And the corresponding gas concentration true value data c i , convert the input data into a feature matrix that the network can recognize and the eigenvectors [c1, c2, ..., c i ] T ; (2) Convolution layer, the convolution operation formula is: in, i, j represent the spatial position of the output, s represents the sth convolution kernel, m, n represent the spatial position of the convolution kernel, and c represents the number of input channels; (3) Activation layer, using the ReLU activation function, the formula is as follows: f(x) = max(0, x) Where x is the input value and f(x) is the output of the function; (4) Pooling layer: the present invention uses maximum pooling to perform dimensionality reduction processing, and the output of the pooling layer is: (5) Fully connected layer: Among them, i represents the input node of the pooling window, and k represents the output node.
5. The method for obtaining particle concentration data according to claim 2, wherein: The particle concentration data is obtained based on TOF laser radar, by establishing the difference between the measured distance and the actual distance and particle concentration The relationship between the particle concentration and the particle concentration is obtained, which specifically includes the following steps: a) A fixed wavelength signal with a modulated signal emitted by the laser transmitting module passes through a standard air chamber, and the receiving module obtains the corresponding distance information d0; b) A fixed wavelength signal with a modulated signal emitted by the laser emission module is passed through the detection chamber, and the receiving module obtains the corresponding distance information d; c) Get the difference between the measured distance and the actual distance d) obtaining the particle concentration f at that time through the particle concentration sensor in the air chamber; wherein different concentrations of particles are generated by burning cigarette cakes of different weights; e) Through polynomial fitting, establish The relationship between the particle concentration and the distance difference is obtained. The particle concentration data is then calculated using the TOF measurement distance.
6. The method for obtaining and training concentration detection data of gases of different concentrations under different particle concentrations as claimed in claim 2, characterized in that: Based on the acquisition of gas concentration data by tunable semiconductor lasers and the use of laser absorption spectroscopy technology to measure gas, the present invention mainly detects SO2 and is characterized by: The semiconductor laser uses the laser absorption spectrum of sulfur dioxide near 7.4 μm as the central wavelength for gas detection. The laser output port includes a focusing lens, which focuses the laser power after full absorption on the detection device to ensure signal quality.
7. The method for obtaining particle concentration data based on TOF laser radar according to claim 4, characterized in that: The ToF laser emission module comprises the following components: (1) Light source module: includes laser emitting tube, laser emitting board and radar main control board, providing the basic light source for the smoke aerosol detection system; the laser emitting tube emits the light source signal, and the laser emitting board diffuses and propagates the light; (2) Laser driving circuit: used to provide a stable driving voltage suitable for the TOF laser radar to emit light signals; (3) Sensor interface: provides power supply, required bias voltage / current signal, digital control phase, and reads data stream from the sensor; The ToF laser receiving module comprises the following components: (1) Laser receiving lens: collects light reflected from the object being measured; (2) Detection system photosensitive element: receives reflected light; (3) Sensor interface: used for communication with the outside (to a PC or processing unit) to transmit the read distance information and concentration information to the storage and analysis module for processing.
8. The tunable diode laser absorption spectroscopy (TDLAS) technique according to claim 5, characterized in that: As a laser light source, the diode laser can output an adjustable laser wavelength. In TDLAS technology, the wavelength of the laser will be adjusted to a wavelength that can be absorbed by the gas for measurement according to the absorption characteristics of the gas being measured. The gas being measured absorbs specific laser energy at this wavelength, and the amount absorbed is proportional to the concentration of the gas. By measuring the intensity of the absorbed light, the gas concentration can be calculated.
9. As claimed in claim 6, the gas concentration data is obtained based on a tunable semiconductor laser, characterized in that: The semiconductor laser includes a signal generator that can generate a high-frequency sinusoidal wave signal and a low-frequency scanning signal. The sinusoidal wave signal realizes high-frequency modulation of the laser signal, reduces the interference of low-frequency noise and provides a reference signal for the phase-locked amplifier; the phase-locked amplifier module obtains information on the change of gas concentration over time by inverting the ratio of the second harmonic and the first harmonic intensity of the corresponding gas absorption peak, thereby obtaining the gas concentration.
