Methane concentration measuring method and measuring system based on direct absorption spectrum

Abstract

The invention relates to the technical field of gas measurement, and discloses a methane concentration measurement method and system based on a direct absorption spectrum, and the method comprises the following steps: collecting N groups of methane samples with different concentrations to obtain corresponding real methane transmission spectrum data; adding Gaussian noise into the simulated methane transmission spectrum to obtain simulated methane transmission spectrum data; improving the deep residual shrinkage network model to obtain an improved deep residual shrinkage network model, and constructing a methane concentration measurement model; training the methane concentration measurement model by using the training set data and the verification set data to obtain a trained methane concentration measurement model; acquiring actual methane transmission spectrum data in the target scene, and inputting the actual methane transmission spectrum data into the trained methane concentration measurement model to obtain methane concentration data corresponding to the target scene; according to the invention, the problems of large noise interference and low measurement precision of the existing methane concentration measurement method based on the direct absorption spectrum are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of gas measurement, and in particular to a methane concentration measurement method and a measurement system based on direct absorption spectroscopy. Background Art

[0002] Methane gas is the second most important greenhouse gas after carbon dioxide, which can seriously affect the global climate and environment. It is colorless, odorless, highly flammable and explosive. It is the main component of natural gas and is widely used as a fuel in industry and society. Nowadays, the continuous use and emission of methane gas has become an important factor leading to global warming and climate change. In addition, in the industrial field, mine methane explosions and natural gas system leaks also pose a threat to human safety. Therefore, conducting high-sensitivity, high-precision and robust methane concentration research and monitoring is crucial to the field of climate change research and the industrial field. Optical sensors based on absorption spectroscopy are widely used in methane detection due to their advantages such as high sensitivity, non-contact and fast response.

[0003] Direct absorption spectroscopy is a gas detection technology of tunable diode laser absorption spectroscopy. It is widely used in environmental detection and industrial production due to its simple system installation, high resolution, strong real-time performance, and no need for calibration. However, the detection sensitivity and stability of direct absorption spectroscopy are easily affected by different types of noise, which can be mainly divided into two categories: electronic noise and optical noise. Electronic noise usually comes from electronic components in the system, usually manifested as high-frequency white noise signals; optical noise mainly comes from interference and multi-surface reflection effects in the optical path, manifested as low-frequency noise. At present, common noise suppression methods are all based on software programs for processing. Common traditional noise suppression methods include Savitzky-Golay filtering, wavelet threshold denoising, Kalman filtering, etc., but these methods require the selection of appropriate parameters and can only process specific noise. It can be seen that the existing methane concentration measurement method based on direct absorption spectroscopy has the problems of large noise interference and low measurement accuracy. Summary of the invention

[0004] The present invention provides a methane concentration measurement method and a measurement system based on direct absorption spectroscopy, so as to solve the problems of large noise interference and low measurement accuracy in the existing methane concentration measurement method based on direct absorption spectroscopy.

[0005] In order to achieve the above object, the present invention is implemented by the following technical solutions: In a first aspect, the present invention provides a methane concentration measurement method based on direct absorption spectroscopy, comprising: Collecting N groups of methane samples of different concentrations to obtain corresponding real methane absorption spectra, and performing baseline removal on the real methane absorption spectra to obtain corresponding real methane transmission spectrum data, wherein N is a positive integer greater than 1; Generate M simulated methane transmission spectra of different concentrations by simulation, and add Gaussian noise to the simulated methane transmission spectra to obtain simulated methane transmission spectrum data, wherein M is a positive integer greater than 1; A deep residual shrinkage network model is constructed, a residual shrinkage construction unit in the deep residual shrinkage network model is improved to obtain an improved deep residual shrinkage network model, and a methane concentration measurement model is constructed based on the improved deep residual shrinkage network model; The simulated methane transmission spectrum data is used as training set data, the real methane transmission spectrum data is used as verification set data, and the training set data and the verification set data are used to train a methane concentration measurement model to obtain a trained methane concentration measurement model; The actual methane transmission spectrum data corresponding to the methane sample in the target scene is obtained, and the actual methane transmission spectrum data is input into the trained methane concentration measurement model to obtain the methane concentration data corresponding to the target scene.

[0006] Optionally, collecting N groups of methane samples of different concentrations to obtain corresponding real methane absorption spectra includes: A direct absorption spectroscopy experimental device was built. N groups of methane samples with different concentrations were configured by a methane gas distributor in the direct absorption spectroscopy experimental device. The preset concentration of the methane samples configured by the methane gas distributor was used as a label to collect the methane absorption spectra corresponding to each group of methane samples as real methane absorption spectra, where N is a positive integer greater than 1.

[0007] Optionally, generating M simulated methane transmission spectra of different concentrations by simulation includes: The methane transmission spectrum under the experimental environment is simulated through the spectral database to generate M methane transmission spectrum simulation data with different concentrations, and the methane transmission spectrum simulation data is used as the simulated methane transmission spectrum, wherein M is a positive integer greater than 1.

[0008] Optionally, the adding Gaussian noise to the simulated methane transmission spectrum to obtain simulated methane transmission spectrum data includes: The noise parameters are obtained by performing noise analysis on the direct absorption spectrum of methane in real environment, and the mean, variance and standard deviation of the noise parameters are calculated; With the mean and variance as the first label, Gaussian white noise corresponding to the first label is added to the simulated methane transmission spectrum. With the standard deviation as the second label, Gaussian noise corresponding to the second label is added to the simulated methane transmission spectrum. The simulated methane transmission spectrum with Gaussian white noise and high-speed noise added is used as simulated methane transmission spectrum data.

[0009] Optionally, the improved deep residual shrinkage network model obtained by improving the residual shrinkage construction unit in the deep residual shrinkage network model includes: A new threshold function is constructed, and the new threshold function is used to replace the original soft threshold function in the residual shrinkage construction unit in the deep residual shrinkage network model to obtain a new residual shrinkage construction unit, wherein the new threshold function satisfies the following relationship: ; In the formula, represents the input features, represents the characteristics of the output, is a symbolic formula function, Represents the threshold value obtained by the above network threshold calculation, represents an arbitrary constant; The deep residual shrinkage network model containing the new residual shrinkage building unit is taken as the improved deep residual shrinkage network model.

[0010] Optionally, constructing a methane concentration measurement model based on the improved deep residual shrinkage network model includes: The gas spectrum filter is constructed using the improved deep residual shrinkage network model and multi-layer perceptron, and the gas concentration predictor is constructed using three fully connected networks; A methane concentration measurement model is constructed using the gas spectrum filter and the gas concentration predictor.

[0011] Optionally, the training process of the methane concentration measurement model includes: S1, frozen gas concentration predictor, uses the training set data to train other spectral filters to obtain the filter model of methane transmission spectrum; S2. Load the pre-trained weights of the gas spectrum filter, and use the training data set to train the gas spectrum filter and the gas concentration predictor at the same time.

[0012] In a second aspect, an embodiment of the present application provides a methane concentration measurement system based on direct absorption spectroscopy, including a processor and a memory; Memory, used to store computer programs; The processor is used to implement any method step described in the first aspect when executing a program stored in the memory.

[0013] Beneficial effects: The methane concentration measurement method based on direct absorption spectroscopy provided by the present invention can directly process the spectrum to obtain concentration data, which not only simplifies data processing and improves the efficiency of training and deployment, but also enhances the generalization ability of the model to new data. This entire model is based on the DRSN backbone network and is suitable for processing situations interfered by strong noise. This is due to the fact that the threshold function in DRSN can achieve adaptive denoising to better mine the information in the noise spectrum. Comparative analysis shows that the proposed GSF has better denoising performance than other denoising methods. In terms of concentration prediction, the present invention not only accelerates the training convergence speed through transfer learning, but also improves the accuracy and reliability of retrieving methane concentration from the absorption spectrum. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of a methane concentration measurement method based on direct absorption spectroscopy according to a preferred embodiment of the present invention; Figure 2 A schematic diagram of the structural framework of a residual shrinkage construction unit according to a preferred embodiment of the present invention; Figure 3 Schematic diagram of the structure of a methane concentration measurement model according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0015] The technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the usual meanings understood by persons with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate quantity restrictions, but indicate the existence of at least one. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0017] See also Figure 1 The present application embodiment provides a methane concentration measurement method based on direct absorption spectroscopy, comprising: Collecting N groups of methane samples of different concentrations to obtain corresponding real methane absorption spectra, and performing baseline removal on the real methane absorption spectra to obtain corresponding real methane transmission spectrum data, wherein N is a positive integer greater than 1; Generate M simulated methane transmission spectra of different concentrations by simulation, and add Gaussian noise to the simulated methane transmission spectra to obtain simulated methane transmission spectrum data, wherein M is a positive integer greater than 1; A deep residual shrinkage network model is constructed, a residual shrinkage construction unit in the deep residual shrinkage network model is improved to obtain an improved deep residual shrinkage network model, and a methane concentration measurement model is constructed based on the improved deep residual shrinkage network model; The simulated methane transmission spectrum data is used as training set data, the real methane transmission spectrum data is used as verification set data, and the training set data and the verification set data are used to train a methane concentration measurement model to obtain a trained methane concentration measurement model; The actual methane transmission spectrum data corresponding to the methane sample in the target scene is obtained, and the actual methane transmission spectrum data is input into the trained methane concentration measurement model to obtain the methane concentration data corresponding to the target scene.

[0018] In the above embodiment, the method specifically includes: S1, acquiring methane direct absorption spectrum (DAS); S2, establishing a simulation data set based on the acquired DAS signal; S3, constructing an end-to-end methane concentration measurement model based on a deep residual shrinkage network (DRSN); S4, training and optimizing the methane concentration measurement model; S5, evaluating the performance of the model.

[0019] The establishment of the simulation data set includes: analyzing the real DAS signal to obtain the noise parameters; simulating the spectral database (HITRAN) to obtain the noise-free spectral data in the training set and the concentration levels corresponding to the spectra; adding multiple noise levels to the noise-free signal to construct the model training set.

[0020] The establishment process of the end-to-end methane concentration measurement model based on DRSN: improve the modules in DRSN; the basic structure and module group of the model. The performance evaluation of the model includes: filtering performance analysis of the test set absorption spectrum data; concentration prediction performance analysis of the test set absorption spectrum data; and stability analysis of the model.

[0021] Compared with other methods, the end-to-end methane concentration measurement model proposed in the present invention can directly process the spectrum to obtain concentration data, which not only simplifies data processing and improves the efficiency of training and deployment, but also enhances the model's generalization ability for new data. This entire model is based on the DRSN backbone network and is suitable for dealing with situations with strong noise interference. This is due to the fact that the threshold function in DRSN can achieve adaptive denoising to better mine the information in the noise spectrum. Comparative analysis shows that the proposed GSF has better denoising performance than other denoising methods. In terms of concentration prediction, the present invention not only accelerates the training convergence speed through transfer learning, but also improves the accuracy and reliability of retrieving methane concentration from the absorption spectrum.

[0022] In this embodiment, the real methane absorption spectrum is the absorption spectrum data obtained based on the real collected methane samples, and therefore it is the real methane absorption spectrum. The simulated methane transmission spectrum is the transmission spectrum data obtained by database simulation, and therefore it is the simulated methane transmission spectrum. The simulated methane transmission spectrum data is to add Gaussian noise to the simulation data to simulate the real collected transmission spectrum data, and therefore it is the simulated methane transmission spectrum data. The actual methane transmission spectrum data is the data obtained through the methane samples in the actual measurement scene when the methane concentration measurement model is trained and applied to the actual measurement scene, and therefore it is the actual methane transmission spectrum data.

[0023] Optionally, collecting N groups of methane samples of different concentrations to obtain corresponding real methane absorption spectra includes: A direct absorption spectroscopy experimental device was built. N groups of methane samples with different concentrations were configured by a methane gas distributor in the direct absorption spectroscopy experimental device. The preset concentration of the methane samples configured by the methane gas distributor was used as a label to collect the methane absorption spectra corresponding to each group of methane samples as real methane absorption spectra, where N is a positive integer greater than 1.

[0024] In the above embodiment, the DAS experimental device is constructed at room temperature (25°C) and normal pressure (1atm), and 40 groups of 0~100ppm methane concentrations are configured by the methane gas distributor in the device. The preset concentration of the gas distributor is used as a label, and 40 groups of methane absorption spectra are collected respectively. The collected spectra are baseline-removed to obtain the corresponding transmission spectrum data, and these 40 groups of data are used as the test set of the model. Using a neural network model to predict the concentration of spectral data requires a large amount of data for training, and using experimental data for training requires a lot of manpower and material resources, so it is necessary to build a simulation data set.

[0025] The specific data in the embodiment are only used as examples here and are not limited. Any data size that meets the requirements of the method flow is applicable to the method steps.

[0026] Optionally, generating M simulated methane transmission spectra of different concentrations by simulation includes: The methane transmission spectrum under the experimental environment is simulated through the spectral database to generate M methane transmission spectrum simulation data with different concentrations, and the methane transmission spectrum simulation data is used as the simulated methane transmission spectrum, wherein M is a positive integer greater than 1.

[0027] Optionally, the adding Gaussian noise to the simulated methane transmission spectrum to obtain simulated methane transmission spectrum data includes: The noise parameters are obtained by performing noise analysis on the direct absorption spectrum of methane in real environment, and the mean, variance and standard deviation of the noise parameters are calculated; With the mean and variance as the first label, Gaussian white noise corresponding to the first label is added to the simulated methane transmission spectrum. With the standard deviation as the second label, Gaussian noise corresponding to the second label is added to the simulated methane transmission spectrum. The simulated methane transmission spectrum with Gaussian white noise and high-speed noise added is used as simulated methane transmission spectrum data.

[0028] In the above embodiment, the noise parameters (mean and variance) of the DAS spectrum in the real experimental environment are analyzed; the methane transmission spectrum in the experimental environment is simulated through the HITRAN database, and 2000 methane transmission spectrum simulation data with a concentration range of 0-100ppm are first generated, and each data consists of 1130 sampling points; Gaussian white noise with a mean of 0 and a variance of 0.0028 is added to each simulated spectrum to simulate the absorption spectrum collected in the experimental environment, and in order to enhance the generalization ability of the model, Gaussian noise with standard deviations of 0.0008, 0.0018, 0.0038 and 0.0048 are added to the methane transmission spectrum at each concentration level for data enhancement. Therefore, a total of 10,000 sets of methane transmission spectrum data are generated as simulation data sets. At the same time, the concentration value corresponding to each transmission spectrum constitutes a concentration set as a training label for the model.

[0029] The specific data in the embodiment are only used as examples here and are not limited. Any data size that meets the requirements of the method flow is applicable to the method steps.

[0030] Optionally, the improved deep residual shrinkage network model obtained by improving the residual shrinkage construction unit in the deep residual shrinkage network model includes: A new threshold function is constructed, and the new threshold function is used to replace the original soft threshold function in the residual shrinkage construction unit in the deep residual shrinkage network model to obtain a new residual shrinkage construction unit, wherein the new threshold function satisfies the following relationship: ; In the formula, represents the input features, represents the characteristics of the output, is a symbolic formula function, Represents the threshold value obtained by the above network threshold calculation, represents an arbitrary constant; The deep residual shrinkage network model containing the new residual shrinkage building unit is taken as the improved deep residual shrinkage network model.

[0031] In the above embodiment, DRSN is composed of a plurality of residual shrinkage building units (RSBU) of different scales, so that deep information can be mined block by block for noisy transmission spectra; DRSN is composed of a plurality of RSBUs, and the structure of RSBU is as follows: Figure 2 As shown in the figure, RSBU uses a feature map of size C×W×1 as the input of the unit; C and W represent the number of channels and width of the feature map respectively; the height of the feature map is always 1, and this embodiment uses a one-dimensional transmission spectrum signal as the input data; then the input data first enters two convolutional layers to obtain the corresponding feature map, which is sent to two routes, one is directly linked to the threshold function, denoted as X, and the other will enter the special module, which can be considered as the attention mechanism module of RSBU, which is also one of the advantages of DRSN; in the special module, there are also two routes, one is to calculate the absolute value first, and then obtain a feature through global average pooling (GAP), denoted as A; the other feature map after GAP is input into a small fully connected network, and the Sigmoid function is used as the last layer to normalize the output to between 0 and 1 to obtain a coefficient, denoted as ; The final threshold can be expressed as *A, this method not only ensures that the threshold is positive and not too large, but also makes different samples have different thresholds. This is why this special module can be regarded as a special attention mechanism; finally, the threshold *A and X are thresholded; the present invention uses a new threshold function to replace the soft threshold function of the original RSBU; RSBU contains a special module for estimating the threshold required to implement the new threshold function; its threshold calculation process is as follows: first, a feature map with a size of C×W×1 is used as the input of the unit; wherein C and W represent the number of channels and width of the feature map respectively; the height of the feature map is always 1, because this paper uses a one-dimensional vibration signal as input data; second, the input feature map is trained with two convolutional layers (Conv); wherein BN represents batch normalization and ReLU represents an activation function; third, in the special module, absolute value processing is performed first, and then global average pooling (GAP) is used to obtain a one-dimensional vector with a length equal to the number of channels ; Fourth, this one-dimensional vector After passing through two fully connected layers (FC), the scaled parameters of the one-dimensional vector are obtained. ; Fifth, use the sigmoid function to implement the scaling parameters Mapping in the range 0-1; Sixth, the threshold of RSBU Equal to the scaling parameter and the one-dimensional vector After the threshold calculation is completed, the new threshold function will be used to filter the information of the multi-channel feature map.

[0032] The new threshold function overcomes the shortcomings of the soft and hard threshold functions. First, the new threshold function has the same continuity as the soft threshold function, and there is no jump point at the threshold of the hard threshold function; secondly, after applying the soft and hard threshold functions, the overall level of the signal will be offset or adjusted, that is, a constant deviation, while the new threshold function will be constant when the appropriate N value is obtained. As asymptote, the influence of constant deviation can be eliminated.

[0033] Optionally, constructing a methane concentration measurement model based on the improved deep residual shrinkage network model includes: The gas spectrum filter is constructed using the improved deep residual shrinkage network model and multi-layer perceptron, and the gas concentration predictor is constructed using three fully connected networks; A methane concentration measurement model is constructed using the gas spectrum filter and the gas concentration predictor.

[0034] In the above embodiment, an end-to-end methane concentration measurement model is constructed based on the improved DRSN and multi-layer perceptron. The model framework is as follows: Figure 3 As shown in the figure, the whole model includes two modules: gas spectrum filter (GSF) and gas concentration predictor (GCP). GSF is composed of an improved DRSN and a multi-layer perceptron, and DRSN contains 6 RSBU sub-modules; GCP is constructed by three fully connected layers. For GSF, three fully connected layers are added after GSF, and the deep features extracted by GSF are upsampled layer by layer to obtain a pure transmission spectrum.

[0035] Optionally, the training process of the methane concentration measurement model includes: S1, frozen gas concentration predictor, uses the training set data to train other spectral filters to obtain the filter model of methane transmission spectrum; S2. Load the pre-trained weights of the gas spectrum filter, and use the training data set to train the gas spectrum filter and the gas concentration predictor at the same time.

[0036] In the above embodiment, the specific training parameters are as follows: Batchsize=16; Optimizer=Adam; Rate=10−4, the learning rate decays 10 times every 100 epochs; Epochs=300. The training steps of the model are mainly divided into two parts: the first step is to freeze the GCP part, pre-train the GSF, and obtain the filtering model of the methane transmission spectrum; the second step is to load the pre-trained weights of the GSF, and train the two modules of GSF and GCP at the same time. The model is verified by loading the test set. The denoising performance of the model is verified by the pre-trained weights of the GSF obtained in the first step, and the concentration prediction performance of the model is verified by the GSF and GCP training parameters obtained in the second step.

[0037] In addition, this embodiment also includes: performing performance evaluation on the trained methane concentration measurement model, and the specific implementation steps include: Perform filtering performance analysis on the test set spectral data. Calculate the signal-to-noise ratio of the spectrum containing noise and the spectrum obtained by different filtering methods. The filtering performance method can be SG filtering, wavelet threshold denoising, Kalman filtering, etc. Perform concentration prediction analysis on the test set spectral data. The correlation coefficient between the predicted concentration and the label concentration can be calculated to evaluate the performance of concentration prediction. The traditional Vogit nonlinear fitting algorithm is selected for comparison. The larger the correlation coefficient, the more accurate the measurement result of the method. The stability of the model was analyzed. In order to evaluate the impact of the end-to-end methane concentration measurement model on the stability of DAS, two MFCs were used to configure a 12ppm methane gas flow into the absorption cell and perform continuous methane measurement for one hour. The half-width at half maximum (HWHM) can be calculated by measuring the statistical histogram of the concentration value. The size of the HWHM can reflect the measurement accuracy of the methane concentration, and the smaller the HWHM, the better.

[0038] The embodiment of the present application also provides a methane concentration measurement system based on direct absorption spectroscopy, including a processor and a memory; Memory, used to store computer programs; The processor is used to implement any method step in the methane concentration measurement method based on direct absorption spectroscopy when executing the program stored in the memory.

[0039] The above-mentioned methane concentration measurement system based on direct absorption spectroscopy can implement various embodiments of the above-mentioned methane concentration measurement method based on direct absorption spectroscopy, and can achieve the same beneficial effects, which will not be described in detail here.

[0040] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A methane concentration measurement method based on direct absorption spectroscopy, characterized in that: include: Collecting N groups of methane samples of different concentrations to obtain corresponding real methane absorption spectra, and performing baseline removal on the real methane absorption spectra to obtain corresponding real methane transmission spectrum data, wherein N is a positive integer greater than 1; Generate M simulated methane transmission spectra of different concentrations by simulation, and add Gaussian noise to the simulated methane transmission spectra to obtain simulated methane transmission spectrum data, wherein M is a positive integer greater than 1; A deep residual shrinkage network model is constructed, a residual shrinkage construction unit in the deep residual shrinkage network model is improved to obtain an improved deep residual shrinkage network model, and a methane concentration measurement model is constructed based on the improved deep residual shrinkage network model; The simulated methane transmission spectrum data is used as training set data, the real methane transmission spectrum data is used as verification set data, and the training set data and the verification set data are used to train a methane concentration measurement model to obtain a trained methane concentration measurement model; The actual methane transmission spectrum data corresponding to the methane sample in the target scene is obtained, and the actual methane transmission spectrum data is input into the trained methane concentration measurement model to obtain the methane concentration data corresponding to the target scene.

2. The methane concentration measurement method based on direct absorption spectroscopy according to claim 1, characterized in that: The collecting of N groups of methane samples with different concentrations to obtain corresponding real methane absorption spectra includes: A direct absorption spectroscopy experimental device was built. N groups of methane samples with different concentrations were configured by a methane gas distributor in the direct absorption spectroscopy experimental device. The preset concentration of the methane samples configured by the methane gas distributor was used as a label to collect the methane absorption spectra corresponding to each group of methane samples as real methane absorption spectra, where N is a positive integer greater than 1.

3. The methane concentration measurement method based on direct absorption spectroscopy according to claim 1, characterized in that: The method of generating M simulated methane transmission spectra of different concentrations by simulation includes: The methane transmission spectrum under the experimental environment is simulated through the spectral database to generate M methane transmission spectrum simulation data with different concentrations, and the methane transmission spectrum simulation data is used as the simulated methane transmission spectrum, wherein M is a positive integer greater than 1.

4. The methane concentration measurement method based on direct absorption spectroscopy according to claim 1, characterized in that: The adding Gaussian noise to the simulated methane transmission spectrum to obtain simulated methane transmission spectrum data comprises: The noise parameters are obtained by performing noise analysis on the direct absorption spectrum of methane in real environment, and the mean, variance and standard deviation of the noise parameters are calculated; With the mean and variance as the first label, Gaussian white noise corresponding to the first label is added to the simulated methane transmission spectrum. With the standard deviation as the second label, Gaussian noise corresponding to the second label is added to the simulated methane transmission spectrum. The simulated methane transmission spectrum with Gaussian white noise and high-speed noise added is used as simulated methane transmission spectrum data.

5. The methane concentration measurement method based on direct absorption spectroscopy according to claim 1, characterized in that: The improved deep residual shrinkage network model obtained by improving the residual shrinkage construction unit in the deep residual shrinkage network model includes: A new threshold function is constructed, and the new threshold function is used to replace the original soft threshold function in the residual shrinkage construction unit in the deep residual shrinkage network model to obtain a new residual shrinkage construction unit, wherein the new threshold function satisfies the following relationship: ； In the formula, represents the input features, represents the output characteristics, is a symbolic formula function, Represents the threshold value obtained by the above network threshold calculation, represents an arbitrary constant; The deep residual shrinkage network model containing the new residual shrinkage building unit is taken as the improved deep residual shrinkage network model.

6. The methane concentration measurement method based on direct absorption spectroscopy according to claim 1, characterized in that: The methane concentration measurement model is constructed based on the improved deep residual shrinkage network model, including: The gas spectrum filter is constructed using the improved deep residual shrinkage network model and multi-layer perceptron, and the gas concentration predictor is constructed using three fully connected networks; A methane concentration measurement model is constructed using the gas spectrum filter and the gas concentration predictor.

7. The methane concentration measurement method based on direct absorption spectroscopy according to claim 6, characterized in that: The training process of the methane concentration measurement model includes: S1, frozen gas concentration predictor, uses the training set data to train other spectral filters to obtain the filter model of methane transmission spectrum; S2. Load the pre-trained weights of the gas spectrum filter, and use the training data set to train the gas spectrum filter and the gas concentration predictor at the same time.

8. A methane concentration measurement system based on direct absorption spectroscopy, characterized in that: Including processor and memory; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 7 when executing a program stored in a memory.

Images