Gas detection device and method

Abstract

The invention discloses a gas detection device, comprising: a gas cell, the gas cell comprising a main cavity extending along a first direction and a side arm communicated with the middle of the main cavity and extending along a second direction; a first laser of the TDLAS module couples a TDLAS laser beam into a main cavity through a transmitting optical fiber, the TDLAS laser beam is propagated in the first direction to form an absorption light path, and the laser beam absorbed by gas is collected to a first detector through a receiving optical fiber; a second laser of the Raman module introduces and focuses a Raman laser beam to a detection point in the main cavity through a side arm in a second direction by exciting an optical fiber, and the detection point is located on a TDLAS laser beam light path; raman scattering light is collected to a second detector by a collecting optical fiber; the temperature and pressure acquisition module is used for acquiring the temperature and the pressure of gas in the gas pool respectively; according to the device, concurrent detection and data complementation of TDLAS detection and Raman detection can be realized, and the gas detection precision and reliability are improved.

Description

Technical Field

[0001] This application relates to the field of gas detection technology, and in particular to a gas detection device and method. Background Technology

[0002] Currently, while tunable semiconductor laser absorption spectroscopy (TDLAS) can achieve highly sensitive detection of specific gases, it has a limited range of monitored species and is easily affected by interference under complex background gases, pressure, and temperature changes, resulting in significant measurement errors. Raman spectroscopy can simultaneously identify multiple components, but its weak signal and low quantitative accuracy make it difficult to use independently for high-precision monitoring.

[0003] Existing technologies that combine the two often employ separate gas chambers or optical paths, which result in problems such as system complexity, large size, and signal crosstalk. Furthermore, they lack an effective mechanism for real-time and dynamic environmental compensation of TDLAS measurements using Raman information, thus limiting their application in variable industrial environments. Summary of the Invention

[0004] This application provides a gas detection device and method. The device improves gas detection accuracy through dual-modal common-point detection.

[0005] In a first aspect, this application provides a gas detection device, comprising: The gas pool includes a main cavity extending along a first direction and a side arm communicating with the middle of the main cavity and extending along a second direction orthogonal to the first direction. Temperature acquisition module, used to acquire the gas temperature in the gas pool; The pressure acquisition module is used to acquire the gas pressure in the gas pool; The TDLAS module includes a first laser and a first detector. The first laser couples the TDLAS laser beam into the main cavity through the transmitting fiber, so that the TDLAS laser beam propagates in the main cavity along the first direction to form an absorption optical path. The TDLAS laser beam after gas absorption is collected by the receiving fiber and transmitted to the first detector. The transmitting fiber and the receiving fiber are located on the same side to form a transceiver integrated structure. The Raman module includes a second laser and a second detector. The second laser guides the Raman laser beam through the excitation fiber into the main cavity via the side arm along the second direction and focuses it on the detection point in the main cavity. The detection point is located on the optical path of the TDLAS laser beam. The Raman scattered light is collected and converged by the lens group and then transmitted to the second detector by the collection fiber. The excitation fiber and the collection fiber point towards the detection point at a 180° back-to-back angle.

[0006] Secondly, this application also provides a gas detection method, which applies the gas detection device as described in the first aspect, and the gas detection method includes: Align the TDLAS measurement data obtained by the TDLAS module with the Raman measurement data obtained by the Raman module; The Raman and TDLAS measurement data are cross-validated based on data alignment, and the concentration is inverted based on the validated TDLAS measurement data to obtain the concentration of the target gas in the gas to be measured.

[0007] In one embodiment, cross-verification of Raman measurement data and TDLAS measurement data based on data alignment includes background subtraction of interfering gas spectral lines from the TDLAS measurement data based on the Raman measurement data, comprising the following steps: Acquire a synchronized time series dataset, which includes multiple data points. Each data point includes the concentration of interfering gas, the original mixed signal, temperature data, and pressure data corresponding to the same time stamp. The concentration of interfering gas is obtained through a Raman module, and the original mixed signal is obtained through a TDLAS module. The original mixed signal includes the absorption spectrum signals generated by the target gas and the interfering gas. A regression vector is constructed based on the concentration, temperature, and pressure data of interfering gases in the data points. The predicted interference signal is obtained by performing an inner product between the regression vector and the preset parameter vector. The parameter vector includes the system's response coefficients to each interference element in the regression vector. The predicted interference signal is subtracted from the original mixed signal of the data points to obtain the corrected effective signal of the target gas.

[0008] In one embodiment, updating the parameter vector includes the following steps: The prediction error is calculated based on the original mixed signal and the predicted interference signal. Based on the old parameter vector, the prediction error is weighted by gain, thereby updating the old parameter vector into a new parameter vector, and the predicted interference signal is calculated based on the new parameter vector.

[0009] In one embodiment, the gain is related to the current regression vector, the covariance matrix, and a preset forgetting factor. The gain vector is constructed using recursive least squares and is adaptively updated based on the current regression vector. The covariance matrix is ​​updated online according to the updated gain vector.

[0010] In one embodiment, the parameter vector is updated when the concentration change exceeds a preset concentration threshold, based on the concentration of the interfering gas measured by the Raman module.

[0011] In one embodiment, cross-verification of Raman measurement data and TDLAS measurement data based on data alignment further includes weighted calibration of the Raman quantitative model based on the TDLAS measurement data, including the following steps: Obtain a synchronized time series dataset, which includes multiple calibration data points. Each calibration data point includes Raman measurement data and TDLAS measurement data corresponding to the same timestamp. Based on the TDLAS measurement data in the calibration data points, the concentration of the first number of TDLAS measured gas components is obtained by inversion. Based on the Raman measurement data in the calibration data points, the concentration of the second number of Raman measured gas components is calculated using a preset Raman quantitative model. The concentrations of the TDLAS measured gas components and the Raman measured gas components are concatenated into a label vector. The types of TDLAS measured gas components and Raman measured gas components are different from each other. Substitute the label vector into the preset weighted objective function. The weighted objective function reflects the error between the concentration of each component in the label vector and the predicted component concentration. The predicted component concentration is calculated based on the updated Raman quantification model. Furthermore, when calculating the error, the weight of the gas component measured by TDLAS is greater than the weight of the gas component measured by Raman. The Raman quantitative model is updated according to a preset update law with the goal of minimizing the weighted objective function.

[0012] In one embodiment, the update law is expressed by the following formula: ; ; In the formula, This represents the predicted component concentration vector from the Raman quantification model; This represents the transpose of the regression coefficient matrix; Represents the Raman spectral eigenvectors of Raman measurement data; Represents the bias vector; This represents the regression coefficient matrix before the update; This represents the updated regression coefficient matrix; Indicates the learning rate; Represents the weight matrix; This represents the label vector.

[0013] In one embodiment, the concentration of gas components measured by TDLAS is used for transfer calibration of the concentration of Raman gas components not measured by TDLAS: ; In the formula, and The coefficient representing the linear relationship between the concentration of the gas component measured by the j-th Raman method and the concentration of the gas component measured by the ith TDLAS method. This represents the correction factor for the Raman residuals; This represents the concentration of the j-th gas component predicted by the Raman module; This represents the concentration of the i-th gas component measured by the TDLAS module; q This indicates the number of gaseous components measured using the TDLAS module.

[0014] In one embodiment, the background gas composition and proportion are analyzed in real time using Raman measurement data. Combined with real-time measured pressure and temperature, the line intensity and linewidth parameters of the TDLAS absorption spectrum are dynamically calculated. Then, the proportioning coefficient in the concentration inversion algorithm is corrected based on the line intensity and linewidth parameters.

[0015] The aforementioned gas detection device integrates an orthogonal sidewall gas cell with a dual-mode optical path, achieving the fusion of TDLAS detection and Raman spectroscopy detection at the same detection point, thus improving gas detection accuracy. The long optical path design of the TDLAS module ensures highly sensitive quantitative detection of specific gases, while the Raman module enables broad-spectrum qualitative identification of multi-component gases. The data from both modules can be cross-checked and compensated, reducing interference errors caused by complex background gases. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a gas detection device in one embodiment; Figure 2 This is a schematic diagram of the gas detection device in one embodiment; Figure 3 Here is a flowchart of a gas detection method in one embodiment; Figure 4 This is a flowchart illustrating background subtraction of interfering gas spectral lines from TDLAS measurement data based on Raman measurement data in one embodiment. Figure 5 This is a flowchart of a weighted calibration of a Raman quantitative model based on TDLAS measurement data in one embodiment. Detailed Implementation

[0017] The present application will be described in detail below with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present application. Any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the protection scope of the present application.

[0018] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0019] In one embodiment, such as Figure 1 As shown, a gas detection device is provided, which includes: a gas cell 101, a temperature acquisition module 102, a pressure acquisition module 103, a TDLAS module 104, and a Raman module 105.

[0020] The gas pool 101 includes a main cavity extending along a first direction and a side arm communicating with the middle of the main cavity and extending along a second direction orthogonal to the first direction. The temperature acquisition module 102 is used to acquire the gas temperature in the gas pool. The pressure acquisition module 103 is used to acquire the gas pressure in the gas pool. The TDLAS module 104 includes a first laser and a first detector. The first laser couples a TDLAS laser beam into the main cavity through a transmitting fiber, so that the TDLAS laser beam propagates in the main cavity along the first direction to form an absorption optical path. The TDLAS laser beam absorbed by the gas is collected by a receiving fiber and transmitted to the first detector. The transmitting fiber and the receiving fiber are located on the same side to form a transceiver integrated structure. The Raman module 105 includes a second laser and a second detector. The second laser guides a Raman laser beam into the main cavity through the side arm along the second direction through an excitation fiber and focuses it on a detection point in the main cavity. The detection point is located on the TDLAS laser beam optical path. The Raman scattered light is collected by a collecting fiber and transmitted to the second detector. The excitation fiber and the collecting fiber point to the detection point at a 180° back-to-back angle.

[0021] Specifically, the gas cell 101 adopts an orthogonal sidewall design. The gas cell 101 includes a main cavity extending along a first direction (e.g., horizontal direction) and a side arm extending along a second direction (e.g., vertical direction) orthogonal to the first direction, communicating with the center of the main cavity. This forms an orthogonal sidewall structure, enabling a single sampling gas chamber to simultaneously serve both TDLAS and Raman detection technologies. The gas cell 101 can be made of stainless steel or aluminum alloy to withstand corrosion from various industrial gases, including acids and alkalis. The inner wall of the gas cell 101 is finished with a deep black matte finish to maximize the absorption of stray light. The optical window is double-polished and coated with a dual-band anti-reflection film.

[0022] The temperature acquisition module 102 can be used to acquire the gas temperature in the gas pool 101 in real time. The gas temperature can be obtained by the Raman module 105 through analysis of the intensity ratio of the rotating Raman band of, for example, N2. The pressure acquisition module 103 can be used to acquire the gas pressure value in the gas pool in real time. The pressure acquisition module 103 can be a high-precision pressure sensor.

[0023] The TDLAS module 104 includes a first laser and a first detector. The optical path of the TDLAS module 104 adopts an integrated transmit / receive design, with the transmitting and receiving optical fibers located on the same side of the probe. The TDLAS laser beam emitted by the first laser is coupled into one end of the main cavity through the transmitting fiber and propagates within the main cavity along a first direction. By placing highly reflective mirrors at both ends of the main cavity, the TDLAS laser beam forms a long optical path absorption path with multiple reflections. The transmitted light, after being absorbed by the gas, is collected by the receiving fiber located on the same side and transmitted to the first detector (e.g., a photodetector) for signal conversion, thereby enabling the acquisition of gas concentration information.

[0024] The Raman module 105 includes a second laser (e.g., a solid-state laser outputting 532nm continuous laser light) and a second detector (e.g., a CCD spectrometer). The second laser, via an excitation fiber, guides the Raman laser beam through an orthogonal side arm into the main cavity along a second direction. After passing through optical components within the side arm (e.g., a collimating mirror, a dichroic mirror, and a focusing objective), the Raman laser beam is focused onto a detection point within the main cavity. This detection point is located on the optical path of the TDLAS laser beam, achieving a high degree of spatial overlap and co-point detection between the TDLAS and Raman laser beams. The excitation and collection fibers point towards this detection point at a 180° back-to-back angle. The generated Raman backscattered light is collected by multiple lens groups, filtered by a long-pass filter, and converged by a collimating lens before being transmitted via the collection fiber to the second detector for Raman spectral acquisition. Through this orthogonal sidewall gas cell design, the Raman analysis region highly overlaps spatially with the main absorption region of the TDLAS. This optical path design is relatively independent, resulting in low crosstalk risk and good signal isolation.

[0025] This gas detection device integrates an orthogonal sidewall gas cell 101 with a dual-mode optical path, achieving the fusion of TDLAS detection and Raman spectroscopy detection at the same detection point, thus improving the accuracy of gas detection. The long optical path design of the TDLAS module 104 ensures high-sensitivity quantitative detection of specific gases, while the Raman module 105 enables broad-spectrum qualitative identification of multi-component gases. The data from both modules can be cross-checked and compensated to reduce interference errors caused by complex background gases.

[0026] In one embodiment, a schematic diagram of the gas detection device is shown below. Figure 2 As shown, the entire gas detection device synchronously drives a Raman laser (outputting a 532nm continuous laser) and a TDLAS laser (outputting a specific wavelength and modulating it) according to a preset mode or real-time operating conditions. The TDLAS laser DL1, after passing through the collimating lens AL and the optical isolator OI, is incident on a long optical path multi-pass absorption cell composed of mirrors (M1, M2) and multiple sets of confocal cavity mirrors (CM1-CM4, etc.), forming an effective absorption optical path of tens of meters, which improves the detection sensitivity. The transmitted light is received by the photodetector In1, obtaining the original TDLAS signal containing gas absorption information (the inversion results have shown a high linearity of R²=0.999).

[0027] The Raman laser DL2, after being collimated by the collimating lens AL, optical isolator OI, and aperture AP, is focused vertically into a long-path multi-pass absorption cell, where it intersects with the TDLAS optical path, exciting the Raman scattering signal of the gas. The scattered light is filtered by the long-pass filter LF to remove Rayleigh stray light and is then received by a spectrometer and CCD detector to obtain the Raman spectrum. The gas being measured continuously flows through the absorption cell under the control of a pressure controller and a flow meter. Both signals are transmitted to the data processing unit, which performs data alignment, physical model-based cross-verification (e.g., using Raman data to compensate for TDLAS spectral line interference), and a concentration inversion algorithm, outputting and displaying a high-precision concentration result (R²=0.999) after fusion correction.

[0028] Based on the same concept, this application also provides a gas detection method, which uses the gas detection device as described above, including: Step S301: Align the TDLAS measurement data obtained by the TDLAS module with the Raman measurement data obtained by the Raman module; Step S302: Perform cross-verification between the data-aligned Raman measurement data and TDLAS measurement data, and perform concentration inversion based on the verified TDLAS measurement data to obtain the concentration of the target gas in the gas to be measured.

[0029] Specifically, data alignment refers to unifying the temporal and spatial correspondence between TDLAS measurement data and Raman measurement data. Since both TDLAS and Raman measurement data originate from the same gas sample detected in the same gas cell, data alignment ensures consistency between the two sets of data in terms of acquisition time and spatial location of the measured gas.

[0030] Mutual verification refers to dynamically correcting TDLAS measurement data using gas types, pressures, temperatures, and background gas composition (e.g., gas composition ratios) provided by Raman measurement data. Real-time Raman spectroscopy allows for the identification of background gas types and their relative concentrations. Combined with temperature and pressure data, this dynamically updates the absorption line intensity and line shape parameters in the TDLAS concentration retrieval algorithm, effectively compensating for measurement deviations caused by changes in background gases.

[0031] Similarly, the Raman quantitative model was reverse-calibrated using TDLAS measurement data. Using the high-precision concentration data provided by TDLAS as a benchmark, the parameters of the Raman quantitative model were dynamically updated through a weighted calibration algorithm, improving the accuracy of quantitative analysis using Raman spectroscopy.

[0032] In this embodiment, the method ensures the temporal consistency of TDLAS detection and Raman detection through data alignment, and improves the accuracy and anti-interference capability of gas concentration measurement under complex working conditions by utilizing a two-way verification mechanism.

[0033] In one embodiment, cross-validation is performed between data-aligned Raman measurement data and TDLAS measurement data, and concentration inversion is performed based on the validated TDLAS measurement data to obtain the concentration of the target gas in the gas to be measured, including: By analyzing the background gas composition and proportion in real time using Raman measurement data, and combining it with real-time measured pressure and temperature, the line intensity and linewidth parameters of the TDLAS absorption spectrum are dynamically calculated. Then, the proportion coefficient in the concentration inversion algorithm is corrected based on the line intensity and linewidth parameters.

[0034] Specifically, TDLAS measures gas concentration based on Beer-Lambert's law: ; in, I 0( v () represents the intensity of the incident light; C represents the concentration of the absorbing substance; L represents the absorption path length; absorption coefficient. It includes the type of gas, pressure P, temperature T, and background gas composition (mole fraction x). i The function of ) Line strength S(T) As temperature changes, the linear function g(v-v0, P, T, x) iThe main focus is on expanding pressure.

[0035] For gases exhibiting vibrational-rotational Raman bands (such as N2 and O2), the bands consist of a series of rotational peaks. The intensity distribution of these rotational peaks follows a Boltzmann distribution and is closely related to temperature. By selecting two rotational peaks (such as J and J') and measuring the intensity ratio R between the two peaks, the gas temperature T can be directly calculated. ; in, It is the rotational constant. It is the nuclear spin degeneracy, h is Planck's constant, c is the speed of light, and k is the nuclear spin degeneracy. B It is Boltzmann's constant. and R represents the intensity of the two rotating peaks, and R represents the intensity ratio of the two rotating peaks.

[0036] Pressure data can be obtained directly by measuring a high-precision pressure sensor integrated into the gas pool.

[0037] After denoising and baseline correction of the real-time acquired Raman spectra, the Raman spectra are used to identify gas components and quantify their concentrations by analyzing the positions and intensities of characteristic peaks. For an ideal gas mixture, the Raman signal intensity of gas i is... Its mole fraction x i Raman scattering cross section Laser power and instrument response function The Raman signal intensity of each gas is directly proportional to the signal intensity of the gas. : .

[0038] Furthermore, based on the formula Using a pre-calibrated instrument response function Raman scattering cross section Laser power and total molecular number density N total Calculate the proportion x of each interfering gas component in the current airflow using parameters such as [parameter 'x']. i Furthermore, based on the proportion of each interfering gas component x in the current airflow... i The concentrations of each interfering gas were calculated. .

[0039] Furthermore, the real-time acquired gas composition ratio x i Substitute the temperature T and pressure P parameters into an absorption spectral database (e.g., the HITRAN database) to calculate the absorption line intensity S(T) and line shape function g(v-v0, P, T, x) of the target gas under the current operating conditions. iThis updates the absorption coefficient α(v).

[0040] The updated absorption coefficient Substituting the data into the TDLAS concentration inversion algorithm, taking wavelength modulation spectrum as an example, the relationship between signal amplitude and concentration is as follows: ; Where the proportionality constant k is P, T, x i The function, R 2f / 1f This is the signal amplitude of 2f / 1f. By dynamically updating the scaling factor k using the real-time parameters provided by Raman, the corrected concentration C can be obtained.

[0041] In this embodiment, the scaling factor k(P, T, x) is dynamically updated. i This enables real-time correction of the target gas concentration inversion results, effectively eliminating measurement deviations caused by environmental factors such as background gas changes, temperature, and pressure fluctuations, and improving the measurement accuracy of TDLAS in complex industrial environments.

[0042] In one embodiment, such as Figure 4 As shown, cross-verification of Raman and TDLAS measurement data based on data alignment includes background subtraction of interfering gas spectral lines from the TDLAS measurement data based on the Raman measurement data, comprising the following steps: Step 401: Obtain the synchronous time series dataset. The synchronous time series dataset includes multiple data points. Each data point includes the concentration of interfering gas, the original mixed signal, temperature data, and pressure data corresponding to the same time stamp. The concentration of interfering gas is obtained through the Raman module, and the original mixed signal is obtained through the TDLAS module. The original mixed signal includes the absorption spectrum signals generated by the target gas and the interfering gas. Specifically, the concentration of interfering gases can be obtained by analyzing the scattering spectra acquired in real time by the Raman module. The original mixed signal can be obtained through the TDLAS module. It can be an absorption spectrum signal that is acquired and demodulated at the same time, reflecting the superposition effect of the absorption spectra of the target gas and the interfering gas.

[0043] For gases exhibiting vibrational-rotational Raman bands (such as N2 and O2), the bands consist of a series of rotational peaks. The intensity distribution of these rotational peaks follows a Boltzmann distribution and is closely related to temperature. By selecting two rotational peaks (such as J and J') and measuring the intensity ratio R between them, the gas temperature T can be directly calculated. Pressure data can be obtained directly through a high-precision pressure sensor integrated into the gas cell.

[0044] Assign a unified timestamp to the four types of data obtained at the same time to form a complete data point, and multiple consecutive data points form a synchronous time series dataset.

[0045] Step 402: Construct a regression vector based on the concentration, temperature, and pressure data of interfering gases in the data points; Specifically, for one data point obtained from a synchronized time series dataset, based on the concentration of interfering gas in that data point... ,temperature and pressure Based on the formula Constructing regression vectors φ k Where T0 and P0 are preset reference temperature and reference pressure.

[0046] Step 403: Perform the inner product of the regression vector and the preset parameter vector to obtain the predicted interference signal. The parameter vector includes the system's response coefficients to each interference element in the regression vector. Specifically, parameter vector This includes the dynamic response coefficients to each interfering element, i.e., the reference response coefficients. α 0. Temperature response coefficient α 1. Pressure response coefficient α 2. Nonlinear coefficients β and time drift items γ Among them, the reference response coefficient α 0 can represent the basic response intensity of the interfering gas under standard conditions; temperature response coefficient. α 1 and pressure response coefficient α 2 can represent the physical effects of temperature changes on detector response and spectral line intensity, and pressure changes on spectral line broadening and collision cross section, respectively; the nonlinear coefficient β represents the absorption saturation or detector nonlinearity effect that may occur at high concentrations; the time drift term γ represents the slow signal drift caused by factors such as window contamination and laser aging.

[0047] For the regression vector With the preset parameter vector Perform inner product dot multiplication to obtain the predicted interference signal. : .

[0048] Here, random noise ε represents the white noise inherent in the measurement.

[0049] Step 404: Subtract the predicted interference signal from the original mixed signal of the data points to obtain the corrected effective signal of the target gas.

[0050] Specifically, from the raw mixed signal of the data points Subtracting predicted interference signals Obtain the corrected effective signal of the target gas. : .

[0051] Original mixed signal This includes the combined absorption contributions of target gas A and interfering gas B. The purpose of this calculation is to extract the signal component of interfering gas B, which is accurately predicted by the model, from the superimposed mixed signal. Separate and subtract to obtain the corrected effective signal of the target gas generated only by target gas A. The effective signal of the target gas eliminated the deviation caused by the overlap of the spectral lines of interfering gas B.

[0052] In one embodiment, updating the parameter vector includes the following steps: The prediction error is calculated based on the original mixed signal and the predicted interference signal. Based on the old parameter vector, the prediction error is weighted by gain, thereby updating the old parameter vector into a new parameter vector, and the predicted interference signal is calculated based on the new parameter vector.

[0053] Specifically, for the newly acquired data point k, based on the old parameter vector θ k The predicted interference signal is 1. The predicted interference signal is then mixed with the original measured signal. Compare and calculate the prediction error e k The calculation formula is: .

[0054] Furthermore, in the old parameter vector θ k Based on 1, through the gain vector K k For prediction error e k A weighted adjustment is performed, and then the new parameter vector is obtained. .

[0055] Wherein, gain vector K k This determines the magnitude of the correction, the gain vector. K k The calculation formula is as follows: .

[0056] After the parameter vector is updated, based on the new parameter vector θ k Calculate the predicted interference signal, that is, by The updated predicted interference signal value is obtained.

[0057] In one embodiment, the gain is related to the current regression vector, the covariance matrix, and a preset forgetting factor. The gain vector is constructed using recursive least squares and is adaptively updated based on the current regression vector. The covariance matrix is ​​updated online according to the updated gain vector.

[0058] Specifically, the gain vector K k Based on formula The calculation yielded, where k Let this be the regression vector at the current moment. M k 1 represents the covariance matrix of the previous time step. λ This is the preset forgetting factor (usually set to 0.99).

[0059] Each time a new data point k is processed, it is based on the current regression vector. k Covariance Matrix M k 1. Adaptive calculation of updated gain K k Therefore, the gain is adaptively updated based on the current system state.

[0060] Using this gain vector to complete the parameter vector θ k After the update, the covariance matrix M It also needs to be updated online to prepare for processing the next data point k+1. Based on the formula... The covariance matrix is ​​updated online, thus completing one full iteration of the recursive least squares method.

[0061] In one embodiment, based on the concentration of interfering gas measured by the Raman module, an update of the parameter vector is triggered when the concentration change exceeds a preset concentration threshold.

[0062] Specifically, based on the concentration values ​​of interfering gases measured by the Raman module x B，k When the absolute value of the difference between the concentrations of interfering gases at two consecutive sampling times satisfies At that time, it is determined whether the current background gas environment has changed significantly.

[0063] If this condition is met, the complete parameter vector update process is immediately triggered and executed. This includes calculating the prediction error.e k Gain vector K k And based on the formula Update parameter vector θ k and based on the formula Update covariance matrix M k .

[0064] If the concentration change does not reach this threshold, the parameter vector is maintained. θ k No change, skip this update.

[0065] In this embodiment, the concentration change triggering mechanism ensures that parameter vector updates are only performed when necessary. This ensures timely tracking of the system's real dynamic changes while avoiding invalid calculations caused by minor fluctuations or noise when the operating conditions are stable, thus achieving a balance between accuracy and efficiency.

[0066] In one embodiment, such as Figure 5 As shown, the mutual verification between Raman measurement data and TDLAS measurement data based on data alignment also includes weighted calibration of the Raman quantitative model based on TDLAS measurement data, including the following steps: Step 501: Obtain the synchronized time series dataset. The synchronized time series dataset includes multiple calibration data points. Each calibration data point includes Raman measurement data and TDLAS measurement data corresponding to the same timestamp. Specifically, each calibration data point k includes Raman measurement data and TDLAS measurement data corresponding to the same time stamp: Raman measurement data consists of raw spectral data acquired by a Raman module. After preprocessing (e.g., noise reduction and baseline correction), the raw spectral data can be extracted into d-dimensional spectral feature vectors. .

[0067] TDLAS measurement data provides high-precision concentration measurements of q key gas components (e.g., CO, HF) by the TDLAS module. TDLAS measurement data can be used as a calibration reference.

[0068] Step 502: Based on the TDLAS measurement data in the calibration data points, the first number of TDLAS measured gas component concentrations are obtained by inversion. Based on the Raman measurement data in the calibration data points, the second number of Raman measured gas component concentrations are calculated using a preset Raman quantitative model. The TDLAS measured gas component concentrations and Raman measured gas component concentrations are concatenated into a label vector. The types of TDLAS measured gas components and Raman measured gas components are different. Specifically, assuming the total number of components in the gas to be measured is m, the TDLAS module can measure q key gas components (such as CO, HF, etc.) with high precision, while the Raman module is used to measure all m components, but its quantitative accuracy is weaker.

[0069] For each calibration data point k, the precise concentration values ​​of q key gases are obtained through a concentration inversion algorithm in the TDLAS measurement data (e.g., harmonic analysis based on wavelength modulation spectrum), denoted as the TDLAS measured gas component concentration vector. .

[0070] Using a pre-defined Raman quantitative model Enter the Raman spectrum data of the current calibration data point. The predicted concentrations of all m gases were calculated. This represents the predicted component concentration vector from the Raman quantification model; This represents the transpose of the regression coefficient matrix; Represents the Raman spectral eigenvectors of Raman measurement data; This represents the bias vector.

[0071] For mq gases measured without TDLAS, the predicted values ​​from the Raman quantitative model are used directly. For q gases measured with TDLAS, only the predicted values ​​from the Raman quantitative model are used as initial estimates, but these will be replaced in subsequent calibrations. The concentrations of the gas components measured by Raman are denoted as vectors. .

[0072] The concentration vector of q gas components measured by TDLAS and Raman-predicted concentrations of mq gas components The vectors are concatenated to form an m-dimensional label vector. Tag vector In this study, the components measured by TDLAS and Raman spectroscopy are different from each other, covering all the gases to be measured.

[0073] Step 503: Substitute the label vector into the preset weighted objective function. The weighted objective function reflects the error between the concentration of each component in the label vector and the predicted component concentration. The predicted component concentration is calculated based on the updated Raman quantitative model. Furthermore, when calculating the error, the weight of the gas component measured by TDLAS is greater than the weight of the gas component measured by Raman. Specifically, in order to calibrate the Raman model, a weighted objective function needs to be constructed, which is as follows: .

[0074] Where W is an m×m diagonal weight matrix. The diagonal elements of the weight matrix are as follows: .

[0075] Gas components measured by the TDLAS module are assigned a higher weight (1.0) and have higher reliability; gas components measured by non-TDLA modules are assigned a lower weight (0.2) and have relatively lower accuracy.

[0076] Step 504: Update the Raman quantitative model according to the preset update law with the goal of minimizing the weighted objective function.

[0077] Specifically, the regression coefficient matrix of the Raman quantitative model is initialized with the original parameters of the system. B = B old .

[0078] Calculate the predicted concentration: And calculate the predicted value. With label vector Error between .

[0079] The Raman quantitative model is updated according to the preset update law: ; Among them, the learning rate η =0.01, This represents the predicted component concentration vector from the Raman quantification model; This represents the transpose of the regression coefficient matrix; Represents the Raman spectral eigenvectors of Raman measurement data; Represents the bias vector; This represents the regression coefficient matrix before the update; This represents the updated regression coefficient matrix; Indicates the learning rate; Represents the weight matrix; This represents the label vector.

[0080] The iterative process continues until a preset convergence condition is met: the root mean square error (RMSE) between the concentrations of the q gas components measured by TDLAS and the true values ​​in the label vector is less than a set threshold (when the prediction error RMSE of the TDLAS components < 1%). The algorithm stops iterating when the convergence condition is met. The final updated regression coefficient matrix is ​​output, and the Raman quantitative model is calibrated based on this updated regression coefficient matrix.

[0081] In one embodiment, the gas detection method further includes performing a transfer calibration of non-TDLAS-measured Raman gas component concentrations using a TDLAS-measured gas component concentration as follows: ; In the formula, and The coefficient representing the linear relationship between the concentration of the gas component measured by the j-th Raman method and the concentration of the gas component measured by the ith TDLAS method. This represents the correction factor for the Raman residuals; This represents the concentration of the j-th gas component predicted by the Raman module; This represents the concentration of the i-th gas component measured by the TDLAS module; q This indicates the number of gaseous components measured using the TDLAS module.

[0082] Specifically, the precise concentrations of q key gas components are obtained through the TDLAS module. Simultaneously, the initial predicted concentrations of all gas components were obtained using a Raman quantitative model. and .

[0083] Furthermore, leveraging the absolute accuracy of TDLAS measurement data, through the established linear relationship ( + This provides a reliable anchor point for Raman measurements, and fine-tunes the results using the prediction residuals of the Raman model itself, ultimately outputting calibrated and more accurate concentration values. .

[0084] This application can be applied to early warning of thermal runaway in lithium-ion battery storage compartments and electric vehicle battery packs. This scenario requires rapid, accurate, and reliable detection of characteristic gases such as CO and HF before smoke or fire occurs. The background gases are complex, including air (N2, O2) and electrolyte solvent vapors (such as DMC, EC). A single TDLAS system may generate false alarms for HF due to interference from water vapor (H2O) spectral lines; a single Raman system may misinterpret water vapor peaks as CO peaks. When the TDLAS module in this gas detection device detects a suspected HF signal, Raman spectral analysis is simultaneously retrieved. If the Raman module clearly identifies only H2O changes, and the abnormal HF signal from the TDLAS disappears after Raman background compensation, the overall judgment is "interference, no alarm," thereby improving reliability.

[0085] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0086] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The above are merely preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the scope of protection of this application.

Claims

1. A gas detection device, characterized in that, The gas detection device includes: A gas pool, the gas pool including a main cavity extending along a first direction, and a side arm communicating with the middle of the main cavity and extending along a second direction orthogonal to the first direction. A temperature acquisition module is used to acquire the gas temperature in the gas pool; The pressure acquisition module is used to acquire the gas pressure in the gas pool; The TDLAS module includes a first laser and a first detector. The first laser couples a TDLAS laser beam into the main cavity through a transmitting fiber, so that the TDLAS laser beam propagates in the main cavity along the first direction to form an absorption optical path. The TDLAS laser beam after being absorbed by the gas is collected by a receiving fiber and transmitted to the first detector. The transmitting fiber and the receiving fiber are located on the same side to form a transceiver integrated structure. The Raman module includes a second laser and a second detector. The second laser guides a Raman laser beam through an excitation fiber into the main cavity via the side arm along a second direction and focuses it onto a detection point within the main cavity. The detection point is located on the optical path of the TDLAS laser beam. The Raman scattered light is collected and converged by a lens group and then transmitted to the second detector via a collection fiber. The excitation fiber and the collection fiber point towards the detection point at a 180° angle.

2. A gas detection method, characterized in that, The gas detection method uses the gas detection device as described in claim 1, and the gas detection method includes: The TDLAS measurement data obtained by the TDLAS module and the Raman measurement data obtained by the Raman module are aligned. The Raman measurement data and the TDLAS measurement data are cross-verified based on data alignment, and the concentration is inverted based on the verified TDLAS measurement data to obtain the concentration of the target gas in the gas to be measured.

3. The gas detection method according to claim 2, characterized in that, The mutual verification of the Raman measurement data and the TDLAS measurement data based on data alignment includes background subtraction of interfering gas spectral lines from the TDLAS measurement data based on the Raman measurement data, including the following steps: Acquire a synchronized time series dataset, which includes multiple data points. Each data point includes the concentration of interfering gas, the original mixed signal, temperature data, and pressure data corresponding to the same time stamp. The concentration of interfering gas is obtained through the Raman module, and the original mixed signal is obtained through the TDLAS module. The original mixed signal includes the absorption spectrum signals generated by the target gas and the interfering gas. Based on the concentration, temperature, and pressure data of the interfering gas in the data points, a regression vector is constructed. The predicted interference signal is obtained by performing an inner product between the regression vector and a preset parameter vector, wherein the parameter vector includes the system's response coefficients to each interference element in the regression vector. The predicted interference signal is subtracted from the original mixed signal of the data points to obtain the corrected effective signal of the target gas.

4. The gas detection method according to claim 3, characterized in that, The method further includes updating the parameter vector, including the following steps: Calculate the prediction error based on the original mixed signal and the predicted interference signal; Based on the old parameter vector, the prediction error is weighted by gain, thereby updating the old parameter vector into a new parameter vector, and the predicted interference signal is calculated based on the new parameter vector.

5. The gas detection method according to claim 4, characterized in that, The gain is a gain vector constructed using recursive least squares based on the current regression vector, covariance matrix, and preset forgetting factor. The gain is adaptively updated based on the current regression vector, wherein the covariance matrix is ​​updated online based on the updated gain vector.

6. The gas detection method according to claim 4 or 5, characterized in that, Based on the concentration of interfering gas measured by the Raman module, the parameter vector is updated when the concentration change exceeds a preset concentration threshold.

7. The gas detection method according to claim 2, characterized in that, The mutual verification of the Raman measurement data and the TDLAS measurement data based on data alignment also includes weighted calibration of the Raman quantitative model based on the TDLAS measurement data, including the following steps: Obtain a synchronized time series dataset, which includes multiple calibration data points, and each calibration data point includes Raman measurement data and TDLAS measurement data corresponding to the same timestamp; Based on the TDLAS measurement data in the calibration data points, the concentration of a first number of TDLAS measured gas components is obtained by inversion. Based on the Raman measurement data in the calibration data points, the concentration of a second number of Raman measured gas components is calculated using a preset Raman quantitative model. The concentrations of the TDLAS measured gas components and the concentrations of the Raman measured gas components are concatenated into a label vector. The types of TDLAS measured gas components and Raman measured gas components are different from each other. Substitute the label vector into the preset weighted objective function. The weighted objective function reflects the error between the concentration of each component in the label vector and the predicted component concentration, which is calculated based on the updated Raman quantitative model. Furthermore, when calculating the error, the weight of the gas component measured by TDLAS is greater than the weight of the gas component measured by Raman. The Raman quantitative model is updated according to a preset update law with the goal of minimizing the weighted objective function.

8. The gas detection method according to claim 7, characterized in that, The update law is expressed by the following formula: ； ； In the formula, This represents the predicted component concentration vector from the Raman quantification model; This represents the transpose of the regression coefficient matrix; Represents the Raman spectral eigenvectors of Raman measurement data; Represents the bias vector; This represents the regression coefficient matrix before the update; This represents the updated regression coefficient matrix; Indicates the learning rate; Represents the weight matrix; This represents the label vector.

9. The gas detection method according to claim 7, characterized in that, The gas detection method further includes performing a transfer calibration of the non-TDLAS-measured Raman gas component concentration using the TDLAS-measured gas component concentration, according to the following formula: ； In the formula, and The coefficient representing the linear relationship between the concentration of the gas component measured by the j-th Raman method and the concentration of the gas component measured by the ith TDLAS method. This represents the correction factor for the Raman residuals; This represents the concentration of the j-th gas component predicted by the Raman module; This represents the concentration of the i-th gas component measured by the TDLAS module; q This indicates the number of gaseous components measured using the TDLAS module.

10. The gas detection method according to claim 2, characterized in that, The gas detection method further includes: The background gas composition and proportion are analyzed in real time using the Raman measurement data. Combined with the real-time measured pressure and temperature, the line intensity and linewidth parameters of the TDLAS absorption spectrum are dynamically calculated. Then, the proportion coefficient in the concentration inversion algorithm is corrected based on the line intensity and linewidth parameters.

Images