A method of estimating indoor pollutant emission concentration

Abstract

The present application provides a kind of indoor pollutant emission concentration estimation method, it is related to environmental monitoring and spectral analysis technical field, the present application is laid out monitoring point at different height in the region to be detected, and the data of temperature and humidity and multispectral absorption curve are synchronously collected and preset rule screening data;Reference industry list and selected initial candidate pollutant in combination with expert experience, merge the pollutant similar in spectral characteristics, form candidate set;The peak intensity of candidate pollutant is corrected in temperature and humidity direction, adopts non-negative sparse dictionary matching method, solves weight vector with residual minimization as target, introduces time smoothing constraint and humidity and weight change rate consistency constraint;Weight vector is converted into concentration value, and concentration interval is generated by remaining wave band method quantization method;Construct vertical box model, and the matching of simulation result and actual diffusion characteristics is verified by vertical gradient consistency index, when it does not reach the standard, adjust candidate pollutant, until output concentration estimation interval.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of environmental monitoring and spectral analysis, in particular to an indoor pollutant emission concentration estimation method. BACKGROUND

[0002] In the existing pollutant concentration detection method based on spectral technology, the mainstream way is to use differential optical absorption spectrum and other technologies, to collect gas absorption spectrum by laying high-resolution spectral equipment in the monitoring area, to combine trace gas standard differential absorption cross section, to use least square fitting algorithm to inverse the measured spectrum, so as to obtain the concentration of pollutants. This kind of method needs to lay a large number of spectral monitoring equipment meeting the resolution requirement at different positions to ensure the accuracy of regional monitoring.

[0003] In the prior art, the publication number CN109975230A discloses an atmospheric pollutant concentration online detection method, which divides the grid unit according to geographical longitude and latitude and environmental factors, collects spectrum by using low-resolution first-type spectral detector at the preset monitoring point, collects data by using high-resolution second-type spectral detector at the specific monitoring point in the grid center, and trains and periodically calibrates the pollutant gas inversion model based on the two types of spectrum at the local server, so as to realize the prediction of pollutant concentration at low cost. Although the environmental parameter monitoring is introduced, the correction mechanism for the slight shift of spectral characteristics caused by temperature and humidity factors is relatively macroscopic, and lacks a fine correction process for the peak intensity of spectral characteristics. In addition, the model prediction result mainly depends on the spectral data and the inversion algorithm, and there is no direct verification link of pollutant diffusion characteristics and measured concentration data, so it is difficult to fully guarantee the reliability of the concentration estimation in complex environments such as closed small spaces.

[0004] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0005] The present application aims to provide an indoor pollutant emission concentration estimation method to solve the problems raised in the background.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0007] An indoor pollutant emission concentration estimation method, the specific steps include:

[0008] Select an initial candidate pollutant, for any two pollutants, first determine the number of overlapping characteristic absorption bands according to the room temperature spectral cross section library, only for two pollutants with overlapping number not less than a preset minimum threshold, extract the standard peak intensity of the overlapping characteristic absorption band to construct a vector, and merge the same items by calculating the cosine similarity of the vector to form a set of candidate pollutants.

[0009] The method comprises the following steps: arranging monitoring points in a detection area and synchronously collecting multi-spectral absorption curves, temperature and humidity; screening the multi-spectral absorption curves whose temperature deviates from a standard room temperature by a preset threshold; correcting the peak intensity of a characteristic absorption band of a candidate pollutant according to a real-time temperature based on the Beer-Lambert law; obtaining the water vapor absorption value under the corresponding humidity of the candidate pollutant whose characteristic absorption band overlaps with water vapor according to a water vapor spectrum library, and further correcting the peak intensity of the characteristic absorption band, so as to complete the construction of the spectral characteristics of the candidate pollutant.

[0010] The method comprises the following steps: adopting a non-negative and sparse dictionary matching method, minimizing the residual of the weighted combination of the multi-spectral absorption curve and the spectral characteristics of the candidate pollutant for each monitoring point, obtaining the weight vector of each candidate pollutant, and simultaneously introducing a time smoothing constraint to limit the mutation of the weight vector at adjacent time points, and increasing the consistency constraint of the humidity change rate and the weight vector change rate for the humidity-sensitive candidate pollutant.

[0011] The method comprises the following steps: converting the weight vector into a pollutant concentration value according to the corresponding relationship between the spectral response and the concentration in the reference standard, quantifying the uncertainty by using the leave-band method, and converting the concentration value into a corresponding concentration interval based on the uncertainty, constructing a box model to verify whether it is consistent with the diffusion characteristics, modifying the candidate pollutant if it is not consistent, and outputting the pollutant concentration estimation result until it is consistent with the diffusion characteristics.

[0012] Further, the method for merging the same items to form the set of candidate pollutants by calculating the cosine similarity of the vectors is as follows:

[0013] The peak intensities of the two initial candidate pollutants in the overlapping characteristic absorption bands are integrated into a feature vector, wherein the vector dimension corresponds to the discrete wavelength points in the overlapping characteristic absorption bands, and the vector elements are the peak intensities of the wavelength points; the normalized cosine similarity of the two feature vectors is calculated; a similarity threshold is set , and When the cosine similarity of the two pollutants is greater than or equal to the threshold, it is determined that the spectral characteristics of the two pollutants are highly similar, and they are merged into the same category to form the set of candidate pollutants.

[0014] Further, the method for arranging monitoring points in a detection area and synchronously collecting multi-spectral absorption curves, temperature and humidity, and screening the multi-spectral absorption curves whose temperature deviates from a standard room temperature by a preset threshold is as follows:

[0015] For the to-be-detected area, a monitoring level is planned in the vertical direction according to the net height of the room, and the vertical height difference of adjacent levels needs to be not less than 0.8 m. A multispectral spectrometer and a temperature and humidity sensor are arranged at each monitoring point. The resolution of the multispectral spectrometer is adjusted according to the characteristic absorption band of the pollutant. The sampling frequency is consistent with the temperature and humidity sensor. The multispectral absorption curve and the temperature and humidity data are time-aligned at the millisecond level. Finally, a raw data set containing time, space, spectral signal and temperature and humidity is formed. From the raw data set, the multispectral absorption curve data of the complete period within the preset threshold deviation of the temperature from the standard room temperature 25℃ is selected, and the time and temperature are recorded synchronously to form the multispectral absorption curve data.

[0016] Further, the method for correcting the peak intensity of the candidate pollutant characteristic absorption band according to the real-time temperature is:

[0017] Firstly, the reference spectral line intensity of the candidate pollutant at the standard room temperature, the molecular partition function parameter, the low-state energy level energy, the spectral line center wave number and the Lorentz half-width under the reference temperature and pressure are retrieved from the spectral database. Secondly, based on the principle of molecular spectroscopy, combined with the real-time temperature, the reference spectral line integral intensity is temperature-compensated by introducing the molecular partition function ratio, the Boltzmann distribution factor and the temperature correction term of the oscillator intensity, to obtain the spectral line integral intensity at the real-time temperature, so as to correct the influence of temperature on the molecular energy level population and the total absorption capacity. Then, the Doppler full width and the corresponding Gaussian standard deviation are calculated according to the real-time temperature, and the reference Lorentz half-width is temperature and pressure corrected by combining the indoor real-time atmospheric pressure and the temperature index, to quantify the influence of the temperature and pressure environment on the line width. Further, the value of the normalized Voigt profile at the center of the spectral line is calculated by the complex error function, that is, the Voigt profile center value, which represents the concentration degree of the spectral line absorption intensity at the peak position under the real-time temperature and pressure. Finally, the spectral line integral intensity at the real-time temperature is multiplied by the Voigt profile center value to obtain the line center absorption cross section of the candidate pollutant characteristic absorption band at the real-time temperature. The cross section is the peak intensity of the characteristic absorption band after temperature correction, which realizes the correction of the peak intensity according to the temperature.

[0018] Further, the method for further correcting the peak intensity of the characteristic absorption band of water vapor according to the water vapor spectrum library to obtain the water vapor absorption value under the corresponding humidity is:

[0019] The water vapor spectral line parameters overlapping with the characteristic absorption band of the candidate pollutant are retrieved from the water vapor spectral library, including the reference spectral line intensity of water vapor at standard room temperature, molecular partition function parameters, low state energy level energy, spectral line center wave number, Lorentz half-width at reference temperature and pressure, and temperature index; secondly, based on the principles of meteorology and molecular physics, combined with the real-time monitored indoor environmental temperature and relative humidity, the water vapor saturation vapor pressure at real-time temperature is calculated through the Magnus empirical formula, and then the water vapor partial pressure is obtained according to the product of relative humidity and saturation vapor pressure, and then the water vapor molecular number density at real-time humidity is converted combined with the Boltzmann constant and real-time temperature;

[0020] Then, the molecular partition function ratio of water vapor, the Boltzmann distribution factor and the temperature correction term of the oscillator strength are introduced to obtain the spectral line integral intensity of water vapor at real-time temperature, and the reference Lorentz half-width is corrected in terms of temperature and pressure by combining the real-time atmospheric pressure in the room with the temperature index of water vapor, and the Doppler full width of the water vapor spectral line is calculated according to the real-time temperature, and the corresponding Gaussian standard deviation is calculated, so as to quantify the influence of temperature and pressure environment on the width of water vapor spectral line; the normalized Voigt profile center value of water vapor spectral line at real-time temperature and pressure is calculated through the complex error function, the water vapor spectral line integral intensity at real-time temperature is multiplied by the Voigt profile center value, and the line center absorption cross section of each overlapping water vapor spectral line at the center wave number is obtained; then, combined with the water vapor molecular number density calculated before and the optical path length of the monitoring system, the absorbance contribution of a single water vapor spectral line at the corresponding wave number of the candidate pollutant characteristic absorption band is calculated through Beer-Lambert law, and finally the absorbance contributions of all water vapor spectral lines overlapping with the candidate pollutant characteristic absorption band are accumulated to obtain the total absorbance interference value of water vapor in the characteristic absorption band; the absorbance corresponding to the line center absorption cross section of the candidate pollutant characteristic absorption band after temperature correction is subtracted by the total absorbance interference value of water vapor, and the non-negative result is taken, and finally the peak intensity of the candidate pollutant characteristic absorption band after water vapor interference correction is obtained.

[0021] Further, the weight vector of each candidate pollutant is obtained, and the time smoothing constraint is introduced to limit the mutation of the weight vector at adjacent time, and the consistency constraint of humidity change rate and weight vector change rate is added for the humidity sensitive candidate pollutant.

[0022] The spectral feature matrix of the candidate pollutant is taken as a dictionary library, and each column of the dictionary library, i.e. the spectral feature vector corresponding to one kind of candidate pollutant, is subjected to L2 normalization processing,

[0023]

[0024] In the formula, The dictionary library before normalization processing has a dimension of , denotes the total number of rows of the dictionary library matrix, denotes the number of candidate pollutant species, denotes the normalized dictionary library, denotes the column index of the dictionary library, where each row corresponds to a peak intensity vector of a candidate pollutant, wherein,

[0025] denotes the index of the discrete wavelength point in the dictionary library, i.e. the total number of discrete wavelength points in the characteristic absorption band of a single candidate pollutant, taking the multispectral absorption curve of the monitoring point as a vector to be decomposed, and constructing an objective function with the minimum residual sum of squares of the measured multispectral absorption curve and the characteristic weighted combination curve in the dictionary library as the core:

[0026]

[0027] wherein, denotes the vector of the multispectral absorption curve, denotes the dictionary library, denotes the weight vector of the candidate pollutant, denotes the sparse regularization parameter; denotes the L1 norm, and a non-negative constraint is imposed on the weight vector, i.e. ;

[0028] A time smoothing constraint is introduced, and a difference penalty term of the weight vectors of adjacent time points is added in the objective function to achieve:

[0029]

[0030] wherein, denotes the weight vector at the time point, denotes the weight vector at the previous time point, denotes the vector of the multispectral absorption curve at the time point, denotes the time smoothing coefficient, denotes the square of the L2 norm, denotes the time, and limits the mutation of the weights of adjacent time points. For the candidate pollutants sensitive to humidity, a consistency constraint penalty term of the humidity change rate and the corresponding weight change rate is added in the objective function:

[0031]

[0032] wherein, denotes the weight change amount of the candidate pollutant sensitive to humidity at the time point, denotes the humidity change amount at the moment, represents a consistency coefficient, which is set according to the humidity sensitivity of the humidity-sensitive pollutant, represents a humidity corresponding weight correlation constraint coefficient.

[0033] Further, the method for converting the weight vector into the pollutant concentration value according to the correspondence between the spectral response and the concentration in the reference standard is as follows:

[0034] Each component of the weight vector is defined as the spectral signal contribution degree of each candidate pollutant. By referring to the corresponding data of spectral response and concentration specified in the reference standard, the response factor of each candidate pollutant is extracted, which reflects the actual concentration value corresponding to the unit signal contribution degree. For each component in the weight vector, the response factor of the corresponding pollutant is multiplied to obtain the concentration value of the pollutant. After the above calculation is completed for all candidate pollutants corresponding to the non-zero components in the weight vector, a set of concentration values of the candidate pollutants is formed.

[0035] Further, the method for quantifying the uncertainty by using the leave-one-out method and converting the concentration value into the corresponding concentration interval based on the uncertainty is as follows:

[0036] The characteristic absorption waveband of the candidate pollutant is divided into several sub-wavebands according to the wavelength points. The leave-one-out method is used for iterative verification: each time, one sub-waveband is selected as the verification waveband, and the remaining sub-wavebands are used as the training waveband. The weight vector is recalculated based on the multi-spectral absorption curve of the training waveband and the concentration estimate value is converted. The estimate value is compared with the reference concentration value calculated based on all wavebands, and the error value of each verification is calculated. After the iterative verification of all sub-wavebands is completed, the standard deviation of all error values is calculated as the quantified uncertainty, and the concentration value of each candidate pollutant is converted into the corresponding concentration interval: taking the concentration value as the center, the uncertainty value is floated up and down, i.e. the concentration interval is the lower limit of the concentration value minus the uncertainty, and the upper limit of the concentration value plus the uncertainty.

[0037] Further, the method for constructing a box model to verify whether it is consistent with the diffusion characteristics is as follows:

[0038] The to-be-detected area is divided into several homogeneous boxes according to the vertical height level, i.e., the height corresponding to the monitoring point, each box corresponds to the spatial range of a monitoring point, a box diffusion model containing the box height, volume and horizontal cross-sectional area is constructed, the box height corresponds to the vertical distance of the height interval of the monitoring point, the horizontal cross-sectional area is consistent with the plane area of the enclosed area, the volume is the product of the box height and the horizontal cross-sectional area, adjacent boxes are connected through horizontal interfaces, and the pollutants are exchanged vertically only through internal diffusion and natural convection, wherein the diffusion coefficient adopts the empirical value corresponding to the standard room temperature of 25 DEG C in a closed environment, and the mixing rate adopts the benchmark value of natural convection in a closed environment, based on the law of conservation of mass in a closed environment, the kinetic relationship of the concentration of each box changing with time is established: the concentration change caused by molecular diffusion in the box is calculated through the diffusion coefficient, the pollutant exchange amount between adjacent boxes caused by natural convection is calculated in combination with the mixing rate, and the concentration change law of each box at different times is comprehensively obtained, the simulation concentration interval of each box at each time is obtained by solving the above relationship, the simulation concentration interval calculated by the model is compared with the actual concentration interval of the corresponding monitoring point, i.e., the set of concentration values of the candidate pollutants, one is to calculate the interval overlap degree, i.e., the proportion of the overlapping part of the simulated concentration interval and the actual concentration interval in the actual concentration interval, and the preset qualified threshold is not less than 90%; two is to evaluate the vertical gradient consistency, i.e., to calculate the ratio of the simulated concentration difference and the actual concentration difference of adjacent boxes, and the preset qualified threshold is not more than 20%, both indicators reach the preset threshold, it is judged that the diffusion characteristics are consistent, and the concentration interval corresponding to the candidate pollutant set is taken as the final estimation result, when any one indicator does not reach the preset threshold, it is judged that the diffusion characteristics are not consistent, the initial candidate pollutant is reselected and subsequent calculation is carried out, until the model simulation result is consistent with the diffusion characteristics, and the candidate pollutant set and the corresponding concentration interval are output as the final estimation result.

[0039] Compared with the prior art, the beneficial effects of the present application are:

[0040] The present application collects the multi-spectral absorption curve and the temperature and humidity data by arranging monitoring points at different heights in the enclosed space, corrects the temperature and water vapor absorption direction of the spectral feature peak value intensity, solves the weight vector by combining the non-negative sparse dictionary matching method, obtains the pollutant concentration value, improves the accuracy of the pollutant concentration estimation, at the same time, quantifies the uncertainty by the wave band method to form a concentration interval, and then verifies the diffusion characteristics by the box model, and dynamically adjusts the candidate pollutants based on the result of whether the diffusion characteristics are consistent, and further improves the accuracy and reliability of the indoor pollutant concentration type and estimation. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 It is the whole method flowchart of the present application;

[0042] Figure 2The benzene series combined schematic diagram of the present application;

[0043] Figure 3 The temperature and humidity change influence diagram of the present application;

[0044] Figure 4 The humidity sensitive pollutant verification diagram of the present application. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical scheme and advantages of the present application more clear and understandable, the present application is further described in detail below in combination with specific embodiments.

[0046] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present application should be the commonly understood meanings by those skilled in the art to which the present application belongs. The "first", "second" and similar words used in the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connected" or "connected" and similar words are not limited to physical or mechanical connection, but can include electrical connection, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to represent relative positional relationship, when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0047] EMBODIMENT

[0048] Please refer to Figure 1 The present application provides a technical scheme:

[0049] An indoor pollutant emission concentration estimation method, the specific steps include:

[0050] Step 1: Select the initial candidate pollutants, for any two pollutants, first determine the number of overlapping characteristic absorption bands according to the room temperature spectrum cross section library, only for two pollutants with overlapping number not less than the preset minimum threshold, extract the standard peak intensity in the overlapping characteristic absorption band to construct a vector, and combine similar items by calculating the cosine similarity of the vector to form a set of candidate pollutants;

[0051] According to the indoor environmental pollution control standard of civil building engineering, indoor air quality standard and other large engineering measurement verification, covering the common pollution characteristics of the pollutant list of the decorated house, the possible existing pollutants are selected, which have the common pollution characteristics of the decorated house, such as formaldehyde, toluene, benzene and ammonia, etc. Based on the traceability list of decoration materials in the area to be detected, such as according to the material type of decoration parts such as board, paint and wallpaper, modify the possible existing pollutants;

[0052] For the possible pollutants, according to the room temperature spectrum cross section library, such as HITRAN database, the characteristic absorption band of the pollutant is obtained through the spectral characteristics of typical indoor pollutants. The molecular structure of different pollutants is significantly different, and the energy level difference is unique, so the characteristic absorption band has exclusive identification, and the pollutant can be distinguished by the band. In the indoor spectral detection, when the spectral curves of two candidate pollutants in the selected band range are very similar, or because of instrument noise interference, it is difficult to distinguish the pollutant species and their respective concentrations, so for the pollutants with 3 or more overlapping characteristic absorption bands, the peak intensity of the two initial candidate pollutants in the overlapping characteristic absorption band is integrated into a feature vector, and the vector dimension of 3 bands is sufficient, and the cosine similarity calculation is more reliable. 1-2 bands may have false high similarity and accidental interference. The vector dimension corresponds to the discrete wavelength points in the overlapping characteristic absorption band, and the vector elements are the peak intensities of each wavelength point. The normalized cosine similarity of the two feature vectors is calculated. The more similar the spectral characteristics of the two pollutants are, the closer the cosine similarity is to 1. Set the similarity threshold , and , which can ensure that the vector direction is close when the cosine similarity of the two pollutants is greater than or equal to the threshold value. When the cosine similarity of the two pollutants is greater than or equal to the threshold value, it is determined that the spectral characteristics of the two pollutants are highly similar, and they are combined into the same category to form a candidate pollutant set. This reduces the number of candidate pollutants and does not cause the detection accuracy to decrease due to missing features. The cosine similarity of toluene and xylene, benzene and toluene is calculated, as shown in Figure 2 , the cosine similarity of toluene, xylene and benzene is 0.89, 0.89 and 0.82 respectively, which can be combined into a benzene series after merging.

[0053] Step 2: Set up monitoring points in the detection area and synchronously collect multi-spectral absorption curves, temperature and humidity. The multi-spectral absorption curve is filtered when the temperature deviation from the standard room temperature is within the preset threshold. Based on the Beer-Lambert law, the peak intensity of the characteristic absorption band of the candidate pollutant is corrected according to the real-time temperature. For the candidate pollutants whose characteristic absorption band overlaps with water vapor, the water vapor absorption value at the corresponding humidity is obtained from the water vapor spectrum library to further correct the peak intensity of the characteristic absorption band. The construction of the spectral characteristics of the candidate pollutants is completed.

[0054] According to the size of the indoor space, the grouping of the to-be-detected area is carried out, for example, a single bedroom, a study room, no complex shielding, furniture, and decoration material distribution is concentrated, and the whole space can be regarded as one to-be-detected area. For large-area spaces such as a living room and an open kitchen, the to-be-detected area can be divided into multiple to-be-detected areas according to the distribution characteristics of the pollution source, for example, according to the position concentration, and the distance between pollution sources is greater than 3 m. The monitoring levels are planned in the vertical direction according to the net height of the room. The vertical height difference between adjacent levels is not less than 0.8 m. A multispectral spectrometer and a temperature and humidity sensor are arranged at each monitoring point. For example, at least three height levels of 0.5-1 m from the ground, 1.5-2 m from the ground, and 0.5-1 m from the ceiling are covered in the vertical direction. The height of 0.5-1 m from the ground is because this height is in the daily sitting and lying height and the height of children's activity zone. The height of 1.5-2 m from the ground is directly related to the daily activities of adults. The height of 0.5-1 m from the ceiling is because some volatile pollutants, such as benzene and toluene, have a density slightly smaller than air and are easy to diffuse upward. The adjacent height difference of not less than 0.8 m provides reliable gradient data support for subsequent diffusion characteristic verification. If the net height of the room of the to-be-detected area cannot support three vertical monitoring points, the number of monitoring points can be reduced to two, but the vertical height difference between adjacent levels still needs to be not less than 0.8 m. The resolution of the multispectral spectrometer is adjusted according to the characteristic absorption wavelength band of the pollutant, for example, the resolution of the multispectral spectrometer is set to -1 for a pollutant with a characteristic absorption wavelength band of greater than or equal to 2 cm . The resolution of the multispectral spectrometer is set to 1 cm -1 for a pollutant with a characteristic absorption wavelength band higher than . The characteristic absorption wavelength band of the pollutant is usually in the range of . If the resolution is too low, the peak intensity of the characteristic absorption wavelength band of each pollutant cannot be accurately extracted. The sampling frequency of the temperature and humidity sensor is consistent with that of the multispectral spectrometer. The multispectral absorption curve and the temperature and humidity data are time-aligned at the millisecond level. For example, the multispectral absorption curve data is collected once every 5 minutes. Finally, a raw data set containing time, space, spectral signal, and temperature and humidity is formed. From the raw data set, the multispectral absorption curve data of the complete period in which the temperature deviation from the standard room temperature 25℃ is within a preset temperature deviation threshold is selected. This is formulated according to the indoor air quality standard process, and the doors and windows need to be closed for 12 hours before detection. The time and temperature are recorded simultaneously to form the multispectral absorption curve data. At the same time, the smaller the temperature deviation and the longer the complete period, the smaller the error. A balance needs to be found between the two, for example, when the temperature deviation threshold is ±2℃, the corresponding temperature range is 23-27℃, which meets the conventional indoor living temperature. The complete period is more than 2 hours, which can reduce the significant influence of temperature fluctuations on spectral absorption. At the same time, the release of pollutants from decoration materials is slow, and this period will not affect the subsequent diffusion characteristic verification.

[0055] The monitoring points are arranged in the to-be-detected area, and the multispectral absorption curve, temperature and humidity are synchronously collected. After the multispectral absorption curve whose temperature deviation from the standard room temperature is within a preset threshold is screened, the light intensity values of each waveband of the multispectral absorption curve are analyzed, and the candidate pollutant list is adjusted based on the spectral characteristics, such as adding a pollutant with a matching feature and removing a pollutant without a corresponding absorption peak. The temperature directly changes the spectral absorption essential characteristics of the pollutant molecules. If not corrected, the peak intensity will deviate from the true value, and then the accuracy of subsequent pollutant qualitative identification and quantitative analysis will be affected. For each pollutant in the candidate pollutant set, the initial peak intensity at the standard room temperature is extracted from the room temperature spectral cross-section library, and the parameters are called from the HITRAN spectral database. It is an internationally recognized high-precision molecular spectral parameter library. The molecular partition function parameters, low-state energy level energy, spectral line center wave number, Lorentz half-width and temperature index of the pollutant are called. In the Beer-Lambert law, the temperature changes the vibration and rotation energy level spacing of the molecules, and finally reflects the change of the peak intensity. Therefore, the corrected spectral line intensity is calculated based on the Beer-Lambert law:

[0056]

[0057] In the formula, the spectral line intensity after temperature correction is represented, the reference spectral line intensity of the spectral database is represented, the molecular partition function parameter is represented, the low-state energy level energy is represented, the spectral line center wave number is represented, the temperature is represented, the standard room temperature is represented, and , Kelvin is used for representation, the second radiation constant is represented;

[0058] Since the peak height also depends on the line width, the Doppler full width is calculated again:

[0059]

[0060] In the formula, the Doppler full width is represented, the mass of a single molecule is represented, the speed of light is represented, the spectral line center frequency is represented, , the Boltzmann constant is represented;

[0061] The Gaussian standard deviation is calculated:

[0062]

[0063] where, denotes the Gaussian standard deviation;

[0064] Calculate the Lorentzian half-width:

[0065]

[0066] where, denotes the Lorentzian half-width, denotes the indoor atmospheric pressure, denotes the standard atmospheric pressure, denotes the temperature exponent, denotes the Lorentzian half-width extracted from the spectral database at the reference temperature and pressure;

[0067] Calculate the normalized Voigt on-center value:

[0068]

[0069] where, denotes the normalized Voigt on-center value, denotes the real part, denotes the complex error function, denotes the imaginary unit, which can be directly given by numerical libraries such as Python, where

[0070] ;

[0071] The formula for calculating the absorption cross-section at the center of the spectral line is:

[0072]

[0073] where, denotes the absorption cross-section at the center of the spectral line;

[0074] Based on Beer-Lambert, calculate the absorbance at the center of the line after temperature correction:

[0075]

[0076] where, denotes the absorbance at the center of the line after temperature correction, i.e., the peak intensity after temperature correction, denotes the number density of molecules, where,

[0077]

[0078] For the set of candidate pollutants, the candidate pollutants with overlapping characteristic absorption bands and water vapor absorption bands are screened out, the characteristic absorption bands of each pollutant that need to be corrected are recorded, the water vapor spectrum library is called, based on the real-time monitored humidity value, the absorption value of water vapor at the characteristic absorption band of each pollutant that needs to be corrected under the humidity condition is matched, a water vapor absorption value vector is formed, the peak intensity after temperature correction and the water vapor absorption value vector are corrected, first, the saturated vapor pressure is calculated based on the Magnus empirical formula suitable for 0-50℃ conditions:

[0079]

[0080] In the formula, Saturation vapor pressure is represented by P, Real-time monitored indoor environmental temperature is represented by T;

[0081] The water vapor partial pressure is calculated:

[0082]

[0083] The water vapor molecular number density is calculated:

[0084]

[0085] For each relevant water vapor spectrum line, the absorbance at the line center after temperature correction is extracted, and the line center absorbance contribution of the spectrum line under the optical path L is calculated:

[0086]

[0087] In the formula, The optical path is represented by L, which is only determined by the hardware structure of the monitoring system, The line center absorbance contribution of the first Spectrum line under the optical path L is represented by A, The index of a single water vapor spectrum line overlapping with the characteristic absorption band of the candidate pollutant is represented by i, Is a positive integer greater than 0, The absorption cross section at the center of the first Spectrum line is represented by Q;

[0088] The contribution of all water vapor spectrum lines to the wave number is accumulated:

[0089]

[0090] In the formula, The contribution of all water vapor spectrum lines to the wave number is represented by A;

[0091] The absorbance at the line center is calculated:

[0092]

[0093] wherein, represents the absorbance at the line center, i.e. the corrected peak intensity, which can be greater than the original absorbance due to water vapor absorption, limiting the peak intensity of all characteristic absorption bands to form the candidate pollutant spectral features after processing;

[0094] Table 1 shows the concentration measurement results of 44 groups of benzene and methane standard gases with known concentrations without temperature and humidity correction, and the concentration error rate is calculated and summarized into a concentration error statistical table. The reference concentration is the actual gas concentration, and the humidity unit is percentage;

[0095]

[0096] As shown in Table 1, when the gas is benzene, every 6 serial numbers form a group. Under the condition that the temperature and reference concentration in each group are basically unchanged, the error of the concentration measurement result increases with the increase of humidity, and the error caused by humidity fluctuation needs to be reduced by algorithm. As shown in Figure 3 when the gas is methane, every 4 serial numbers form a group. When the temperature is fixed in all groups, the pollutant concentration does not show a consistent increasing trend with the increase of humidity, which verifies that methane is not a humidity-sensitive candidate pollutant, and only needs to correct the peak intensity of the characteristic absorption band by temperature. Figure 4

[0097] Step 3: Using the non-negative and sparse dictionary matching method, the residual minimization of the weighted combination of the multi-spectral absorption curve and the candidate pollutant spectral feature is taken as the target for each monitoring point to obtain the weight vector of each candidate pollutant, and at the same time, the time smoothing constraint is introduced to limit the mutation of the weight vector at adjacent time, and the consistency constraint of humidity change rate and weight vector change rate is added for humidity-sensitive candidate pollutants;

[0098] The candidate pollutant spectral feature matrix is used as a dictionary library, and each column of the dictionary library, i.e. the spectral feature vector of a corresponding candidate pollutant, is subjected to L2 normalization processing,

[0099]

[0100] wherein, represents the dictionary library before normalization, with a dimension of , represents the total number of rows of the dictionary library matrix, represents the number of candidate pollutant species, represents the normalized dictionary library, represents the column index of the dictionary library, wherein each row corresponds to a peak intensity vector of a candidate pollutant, and wherein,​

[0101] denotes the index of discrete wavelength point in the dictionary library, i.e. the total number of discrete wavelength points in the single candidate pollutant characteristic absorption band, taking the multi-spectral absorption curve of the monitoring point as the vector to be decomposed, a target function is constructed to minimize the residual square sum of the measured multi-spectral absorption curve and the characteristic weighted combination curve in the dictionary library:

[0102]

[0103] wherein, denotes the vector of multi-spectral absorption curve, denotes the dictionary library, denotes the weight vector of candidate pollutant, denotes the sparse regularization parameter, since the spectral data noise is usually small, it is directly set , denotes the L1 norm, the non-negative constraint is imposed on the weight vector, i.e. ;

[0104] The time smoothing constraint is introduced, and the difference penalty term of the weight vectors at adjacent time points is added in the target function to realize:

[0105]

[0106] wherein, denotes the weight vector at time point t, denotes the weight vector at time point t-1, denotes the vector of multi-spectral absorption curve at time point t, denotes the time smoothing coefficient, it is set , denotes the square of L2 norm, denotes the time, the mutation of the weight at adjacent time points is limited, for the humidity-sensitive candidate pollutant, wherein the humidity-sensitive candidate pollutant can be selected from the international standard and technical manual, and the pollutant whose release concentration is related to the relative humidity or absolute humidity of the environment has been proved by experiments, for example, formaldehyde, benzene VOC, etc., the consistency constraint penalty term of the humidity change rate and the corresponding weight change rate is added in the target function:

[0107]

[0108] wherein, denotes the weight change amount of humidity-sensitive candidate pollutant at time point t, i.e. the weight of the pollutant at time point t-1, ​​​​The weight value of the time step and the previous time step The difference in weight values ​​at different times express The change in humidity over time, i.e. Humidity at this moment compared to the previous moment The difference in humidity at different times, The consistency coefficient is determined as follows: For each pollutant, a gradient humidity is set in a sealed experimental chamber under constant temperature conditions, using a reference humidity level (e.g., 30%). The relative change rate of concentration and the relative change rate of humidity at each humidity point relative to the reference are calculated. The ratio of the relative change rate of concentration to the relative change rate of humidity at each humidity point is calculated, and the arithmetic mean is taken as the initial coefficient. To avoid humidity constraint excessively dominating the objective function, the above average value is scaled by a factor of 1 / 10. The final result is the consistency coefficient of the pollutant. A positive value is taken when the concentration increases with increasing humidity, and a negative value is taken otherwise. This represents the weighted constraint coefficient corresponding to humidity, and is set accordingly. .

[0109] Step 4: Based on the correspondence between spectral response and concentration in the reference standard, convert the pollutant concentration value into a weight vector, quantify the uncertainty using the band retention method, and convert the concentration value into the corresponding concentration range based on the uncertainty. Construct a box model to verify whether it matches the diffusion characteristics. If it does not match, modify the candidate pollutants until it matches the diffusion characteristics and output the pollutant concentration estimation result.

[0110] The weights of each pollutant are extracted from the weight vector. Since the characteristic spectral response intensity is linearly related to the mass concentration within a certain concentration range, a pollutant concentration formula is constructed based on the weight vector:

[0111]

[0112] In the formula, Indicates the first The concentration of each pollutant, Indicates the first Calibration coefficients for each pollutant, Indicates the first The weight of each pollutant, Indicates the first Instrumental background for each pollutant;

[0113] For the instrument background, the spectral response intensity (i.e., the peak intensity of the absorption peak) needs to be measured in zero air (i.e., air with a pollutant concentration of less than 0.001 mg / m³) under the same temperature and pressure conditions as the area to be detected. The weights in the blank state are then obtained using the dictionary normalization method. The formula for calculating the instrument background is: For the calibration coefficient calibration, 3 or more concentrations of standard gas are used, such as 0.1, 0.5, 1.0 mg / m³, and under the same temperature and pressure conditions as the detection area, the spectral response intensity of each standard gas, i.e. the peak intensity of the absorption peak, is measured, and a linear equation is fitted as , wherein, represents the concentration of the standard gas of the th pollutant, represents the weight in the quantitative model;

[0114] Then, the wave band method is used to quantify the uncertainty, and based on the uncertainty, the concentration value is converted into a corresponding concentration interval:

[0115] For the characteristic absorption wave band of the th pollutant, it is divided into equal-length sub-wave bands, denoted as , wherein, , to ensure statistical effectiveness, one sub-wave band is selected as the wave band to be verified each time, the absorbance data of all the remaining sub-wave bands are integrated into a training set, and the absorbance data of the selected sub-wave band to be verified are taken as a verification set. For each sub-wave band, the training set weight is solved by using the absorbance of the remaining sub-wave bands, and the verification set weight is solved by using only the absorbance of the sub-wave band. The absolute value of the difference between the training set weight and the verification set weight is taken as the weight error:

[0116]

[0117] , wherein, represents the weight error of the th wave band, represents the training set weight of the th wave band, represents the verification set weight of the th wave band, represents the index of the wave band, ; and

[0118] The weight error can reflect the influence of removing a single sub-wave band on the weight. The greater the error, the higher the uncertainty of the wave band selection. The weight error of rounds is taken as the standard uncertainty of the weight of the th pollutant:

[0119]

[0120] , wherein, represents the weight error of the ​The standard uncertainty of the weight of the pollutant, represents the average value of the M-wheel weight error, that is ;

[0121] Based on the error propagation law, the concentration interval is converted from the uncertainty. First, the standard uncertainty of the concentration is calculated:

[0122]

[0123] In the formula, represents the standard uncertainty of the concentration of the first pollutant;

[0124] The extended uncertainty is calculated at a confidence level of 95%, which is a common practice in engineering. Therefore, the calculation formula of the extended uncertainty is:

[0125]

[0126] In the formula, represents the extended uncertainty of the first pollutant, and the concentration interval of the pollutant obtained by the residual wave band method is calculated: , and is calibrated as , respectively represent the upper and lower limits of the observation concentration interval of the first pollutant in the first box;

[0127] Take a single bedroom after decoration as an example, with a length of 4 m, a width of 3 m, and a height of 2.8 m as the detection area. The three monitoring points are arranged at heights of 0.7 m, 1.7 m, and 2.3 m, respectively. A box diffusion model is constructed, and the bedroom is vertically divided into three homogeneous boxes, each corresponding to the spatial range of one monitoring point. Pollutants are only exchanged vertically through molecular diffusion and natural convection, without outdoor ventilation or horizontal diffusion.

[0128] The natural convection exchange frequency is determined by the engineering simplified method of indoor temperature difference correction:

[0129]

[0130] In the formula, represents the natural convection exchange frequency;

[0131] For a certain pollutant in the first box, the single-box concentration variation dynamics equation is constructed:

[0132]

[0133] In the formula, represents the index of the box, Corresponding to the first Each box. Indicates the first The pollutant in the first The concentration of each box Indicates the first The molecular diffusion coefficients of individual pollutants can be directly retrieved from standard manuals, authoritative databases, and academic literature, such as those related to indoor air quality and control. This indicates the horizontal cross-sectional area of ​​the box. This indicates the frequency of natural convection exchange between adjacent boxes. Indicates the first The volume of each box Indicates the first The pollutant in the first The concentration of each box Indicates the first The pollutant in the first The concentration of each box, among which... This indicates that there are no pollutants accumulated on the ground. The ceiling does not transmit power outwards;

[0134] Let the total exchange frequency of adjacent boxes be the sum of the diffusion and convection exchange frequencies:

[0135]

[0136] In the formula, Let represent the total exchange frequency between adjacent boxes. Then the dynamic equation simplifies to:

[0137]

[0138] In the formula, Indicates the first The pollutant in the first and The total switching frequency of each box Indicates the first and The total switching frequency of each box;

[0139] Monte Carlo sampling was used to sample a uniformly distributed ±10% population. ±15% uniformly distributed conduct Random sampling. For integers greater than 500, the dynamic equations are solved again after each sampling to obtain... The 2.5% to 97.5% percentiles were used as the simulated concentration range. , , They represent the first The pollutant in the first The upper and lower limits of the simulated concentration range for each box;

[0140] The overlap between the pollutant concentration range obtained by the lingering band method and the simulated concentration range is calculated:

[0141]

[0142] In the formula, Indicates the first The pollutant in the first The degree of overlap between the intervals of the boxes indicates that the model's prediction range of pollutant concentration in that box is closer to the actual detection range, and the stronger the model's reliability. Specifically, when... At that time, adopt Instead of dividing by zero;

[0143] Calculate the concentration difference ratio between adjacent boxes:

[0144]

[0145] In the formula, Indicates the first The pollutant in the first The average concentration observed in each box Indicates the first The pollutant in the first The average simulated concentration of each box Indicates the first The pollutant in the first Each box and The vertical gradient consistency error of each box, when all boxes If the pollutant is found to be consistent with the diffusion characteristics, the concentration range of the pollutant obtained by the wavelength band method is output. Otherwise, if the pollutant is found to be inconsistent with the diffusion characteristics, the candidate pollutant is modified. Table 2 shows the overlap rate and gradient error of the candidate pollutants calculated by the above formula for different detection areas. After comparison with the set threshold, the results show whether the pollutant is consistent with the diffusion characteristics. In some detection areas, toluene, xylene and benzene are combined into benzene series compounds. A total of 40 sets of data were collected and a diffusion characteristic verification table was generated after summarizing them.

[0146]

[0147] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0148] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. A person of skill in the art can be aware that units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0149] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, and can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiments according to actual needs.

[0150] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for estimating indoor pollutant emission concentrations, characterized in that, The specific steps include: Select initial candidate pollutants. For any two pollutants, first determine the number of overlaps in their characteristic absorption bands based on the room temperature spectral cross section library. Only for two pollutants with an overlap number not less than a preset minimum threshold, extract their standard peak intensities in the overlapping characteristic absorption bands to construct vectors. Then, combine like terms by calculating the cosine similarity of the vectors to form a set of candidate pollutants. Monitoring points were set up in the area to be tested and multispectral absorption curves, temperature and humidity were collected simultaneously. Multispectral absorption curves with temperature deviations from standard room temperature within a preset threshold were screened. Based on Beer-Lambert's law, the peak intensity of the characteristic absorption bands of candidate pollutants was corrected according to the real-time temperature. For candidate pollutants whose characteristic absorption bands overlapped with water vapor, the water vapor absorption values ​​at the corresponding humidity were obtained from the water vapor spectral library, and the peak intensity of their characteristic absorption bands was further corrected to complete the construction of the spectral characteristics of candidate pollutants. A non-negative and sparse dictionary matching method is adopted. For each monitoring point, the goal is to minimize the residual of the weighted combination of the multispectral absorption curve and the spectral features of the candidate pollutants. The weight vector of each candidate pollutant is obtained. At the same time, a time smoothing constraint is introduced to limit the abrupt change of the weight vector between adjacent time points. For humidity-sensitive candidate pollutants, a consistency constraint between the humidity change rate and the weight vector change rate is added. Based on the correspondence between spectral response and concentration in the reference standard, the pollutant concentration value is converted by combining the weight vector. The uncertainty is quantified by the band retention method, and the concentration value is converted into the corresponding concentration range based on the uncertainty. A box model is constructed to verify whether it is consistent with the diffusion characteristics. If it is inconsistent, the candidate pollutants are modified until it is consistent with the diffusion characteristics, and the pollutant concentration estimation result is output.

2. The method for estimating indoor pollutant emission concentration according to claim 1, characterized in that: The method for forming a set of candidate pollutants by merging like terms through the cosine similarity of vectors is as follows: The peak intensities of the two initial candidate pollutants in overlapping characteristic absorption bands are integrated into a feature vector, where the vector dimension corresponds to discrete wavelength points within the overlapping characteristic absorption bands, and the vector elements are the peak intensities at each wavelength point. The normalized cosine similarity between the two feature vectors is calculated, and a similarity threshold is set. ,and When the cosine similarity between two pollutants is greater than or equal to the threshold, they are determined to have highly similar spectral characteristics and are merged into the same category to form a set of candidate pollutants.

3. The method for estimating indoor pollutant emission concentration according to claim 1, characterized in that: The method for setting up monitoring points in the area to be tested and simultaneously collecting multispectral absorption curves, temperature, and humidity data, and then screening multispectral absorption curves whose temperature deviation from the standard room temperature is within a preset threshold is as follows: For the area to be monitored, monitoring levels are planned vertically based on the room's net height, with a minimum vertical height difference of 0.8m between adjacent levels. Each monitoring point is equipped with a multispectral spectrometer and a temperature and humidity sensor. The resolution of the multispectral spectrometer is adjusted according to the characteristic absorption bands of the pollutants, and the sampling frequency is consistent with that of the temperature and humidity sensor. The multispectral absorption curve is aligned with the temperature and humidity data at the millisecond level, ultimately forming a raw dataset containing time, space, spectral signals, and temperature and humidity. From the raw dataset, multispectral absorption curve data for complete time periods with a temperature deviation from the standard room temperature of 25℃ within a preset threshold are selected, and time and temperature are recorded simultaneously to form the multispectral absorption curve data.

4. The method for estimating indoor pollutant emission concentration according to claim 2, characterized in that: The method for correcting the peak intensity of the characteristic absorption band of candidate pollutants based on real-time temperature is as follows: First, the reference spectral line intensities, molecular partition function parameters, low-state energy levels, spectral line center wavenumbers, and Lorentz half-widths at reference temperature and pressure for candidate pollutants are retrieved from the spectral database at standard room temperature. Second, based on the principles of molecular spectroscopy and combined with real-time temperature, the integral intensity of the reference spectral lines is temperature-compensated by introducing the molecular partition function ratio, Boltzmann distribution factor, and oscillator intensity temperature correction term to obtain the integral intensity of the spectral lines at real-time temperature, thereby correcting the influence of temperature on molecular energy level population and total absorption capacity. Subsequently, the Doppler full width and corresponding Gaussian standard deviation are calculated based on real-time temperature. Combined with real-time indoor atmospheric pressure and temperature index, the reference Lorentz half width is corrected for temperature and pressure to quantify the impact of temperature and pressure environment on linewidth. Then, the value of the normalized Voigt profile at the center of the spectral line is calculated through a complex error function, i.e., the Voigt profile center value, which characterizes the concentration of spectral line absorption intensity at the peak position under real-time temperature and pressure. Finally, the spectral line integral intensity at real-time temperature is multiplied by the Voigt profile center value to obtain the line-center absorption cross section of the candidate pollutant characteristic absorption band at real-time temperature. This cross section is the peak intensity of the characteristic absorption band after temperature correction, realizing the correction of peak intensity according to temperature.

5. The method for estimating indoor pollutant emission concentration according to claim 4, characterized in that: The method for further correcting the peak intensity of the characteristic absorption bands by obtaining the water vapor absorbance values ​​at the corresponding humidity levels from the water vapor spectral library is as follows: Water vapor spectral parameters overlapping with the characteristic absorption bands of candidate pollutants were retrieved from the water vapor spectral library. These parameters included the reference spectral line intensity, molecular partition function parameters, low-state energy, spectral line center wavenumber, Lorentz half-width at reference temperature and pressure, and temperature index. Based on meteorological and molecular physics principles, and combined with real-time monitoring of indoor ambient temperature and relative humidity, the saturated vapor pressure of water vapor at the real-time temperature was calculated using the Magnus empirical formula. The partial pressure of water vapor was then obtained by multiplying the relative humidity and the saturated vapor pressure. Finally, the number density of water vapor molecules at the real-time humidity was calculated by combining the Boltzmann constant with the real-time temperature. Subsequently, the molecular partition function ratio of water vapor, Boltzmann distribution factor, and oscillator intensity temperature correction term were introduced to obtain the integrated intensity of water vapor spectral lines at real-time temperature. Simultaneously, the reference Lorentz half-width was corrected for temperature and pressure by combining the real-time indoor atmospheric pressure and the water vapor temperature index. The Doppler full width and corresponding Gaussian standard deviation of the water vapor spectral lines were calculated based on the real-time temperature to quantify the influence of temperature and pressure environment on the width of the water vapor spectral lines. The normalized Voigt profile center value of the water vapor spectral lines under real-time temperature and pressure was calculated using a complex error function. The integrated intensity of the water vapor spectral lines at real-time temperature was multiplied by the Voigt profile center value to obtain the center wavenumber of each overlapping water vapor spectral line. The absorbance cross section at the center of the line is calculated. Then, combining the previously calculated water vapor number density and the optical path length of the monitoring system, the absorbance contribution of a single water vapor spectral line at the corresponding wavenumber of the candidate pollutant's characteristic absorption band is calculated using the Beer-Lambert law. Finally, the absorbance contributions of all water vapor spectral lines overlapping with the candidate pollutant's characteristic absorption band are summed to obtain the total absorbance interference value of water vapor within that characteristic absorption band. The absorbance corresponding to the absorbance cross section at the center of the candidate pollutant's characteristic absorption band after temperature correction is subtracted from the total absorbance interference value of water vapor, and the non-negative result is taken to finally obtain the peak intensity of the candidate pollutant's characteristic absorption band after water vapor interference correction.

6. The method for estimating indoor pollutant emission concentration according to claim 2, characterized in that: The weight vectors of each candidate pollutant are obtained, and a time smoothing constraint is introduced to limit abrupt changes in the weight vectors between adjacent time steps. For humidity-sensitive candidate pollutants, a method is used to add a consistency constraint between the rate of change of humidity and the rate of change of the weight vector: The spectral features of candidate pollutants are used to construct a matrix as a dictionary, and each column of the dictionary, which corresponds to the spectral feature vector of a candidate pollutant, is subjected to L2 normalization. ； In the formula, This represents the dictionary base before normalization, with dimension 1. , This represents the total number of rows in the dictionary matrix. Indicates the number of candidate pollutant types. This represents the normalized dictionary. The column index representing the dictionary. Each row corresponds to a peak intensity vector of a candidate pollutant, where... ; This represents the index of discrete wavelength points in the dictionary. This refers to the total number of discrete wavelength points within the characteristic absorption bands of a single candidate pollutant. The multispectral absorption curves of the monitoring points are used as vectors to be decomposed. An objective function is constructed with the core objective of minimizing the sum of squared residuals between the measured multispectral absorption curves and the weighted combination curves of features in the dictionary. ； In the formula, A vector representing a multispectral absorption curve. Represents a dictionary. This represents the weight vector of the candidate pollutants. This represents the sparse regularization parameter; L1 norm represents the non-negativity constraint imposed on the weight vector, i.e. ; Introducing a time smoothing constraint is achieved by adding a penalty term for the difference between weight vectors at adjacent time steps to the objective function: ； In the formula, express The weight vector at time step, express The weight vector of the previous time step. express The vector of the multispectral absorption curve at time t. Indicates the time smoothing coefficient. Denotes the square of the L2 norm. To represent time and restrict abrupt changes in weights between adjacent time points, a consistency constraint penalty term between the rate of change of humidity and the corresponding rate of change of weight is added to the objective function for humidity-sensitive candidate pollutants. ； In the formula, express The change in weight of candidate pollutants that are sensitive to humidity at any given time. express Changes in humidity over time This represents the consistency coefficient, which is set based on the degree of humidity sensitivity of humidity-sensitive pollutants. This represents the weighted constraint coefficient corresponding to humidity.

7. The method for estimating indoor pollutant emission concentration according to claim 6, characterized in that: Based on the correspondence between spectral response and concentration in the reference standard, the method for converting pollutant concentration values ​​using a weighted vector is as follows: Each component of the weight vector is defined as the contribution of the spectral signal of each candidate pollutant. By using the corresponding data of spectral response and concentration specified in the reference standard, the response factor of each candidate pollutant is extracted. The response factor reflects the actual concentration value corresponding to the unit signal contribution. For each component in the weight vector, the response factor of the corresponding pollutant is multiplied to obtain the concentration value of the pollutant. After performing the above calculation for all candidate pollutants corresponding to non-zero components in the weight vector, a set of concentration values ​​of candidate pollutants is formed.

8. The method for estimating indoor pollutant emission concentration according to claim 7, characterized in that: The method of quantifying uncertainty using the spectral density method and converting concentration values ​​into corresponding concentration ranges based on the uncertainty is as follows: The characteristic absorption bands of candidate pollutants are divided into several sub-bands according to wavelength points, and the leave-one-out method is used for iterative verification: each time, one sub-band is selected as the verification band, and the remaining sub-bands are used as training bands. The weight vector is recalculated based on the multispectral absorption curves of the training bands and the concentration estimate is obtained. The estimated value is compared with the reference concentration value calculated based on all bands, and the error value of a single verification is calculated. After iterative verification of all sub-bands, the standard deviation of all error values ​​is calculated as the quantified uncertainty, and the concentration value of each candidate pollutant is converted into a corresponding concentration range: with the concentration value as the center, the uncertainty value fluctuates above and below, that is, the concentration range is the concentration value minus the uncertainty as the lower bound, and the concentration value plus the uncertainty as the upper bound.

9. The method for estimating indoor pollutant emission concentration according to claim 8, characterized in that: The method for constructing a box model to verify and confirm its consistency with diffusion characteristics is as follows: The area to be tested is divided into several homogeneous boxes according to vertical height levels, i.e., the height of the corresponding monitoring points. Each box corresponds to the spatial range of one monitoring point. A box-type diffusion model is constructed, including the box height, volume, and horizontal cross-sectional area. The box height corresponds to the vertical distance of the monitoring point height interval, the horizontal cross-sectional area is consistent with the planar area of ​​the closed area, and the volume is the product of the box height and the horizontal cross-sectional area. Adjacent boxes are connected through a horizontal interface. Pollutants are exchanged vertically only through internal diffusion and natural convection. The diffusion coefficient adopts the empirical value corresponding to the standard room temperature of 25°C in a closed environment, and the mixing rate adopts the benchmark value of natural convection in a closed environment. Based on the law of conservation of mass in a closed environment, the dynamic relationship of the concentration change of each box over time is established: the concentration change caused by molecular diffusion inside the box is calculated by the diffusion coefficient, and the amount of pollutant exchanged between adjacent boxes due to natural convection is calculated by combining the mixing rate. The concentration change of each box at different times is obtained by combining the results. The model works by solving the above relationships to obtain the simulated concentration range for each container at each time point. The simulated concentration range calculated by the model is then compared with the actual concentration range at the corresponding monitoring point, i.e., the set of candidate pollutant concentration values. First, the overlap between the simulated and actual concentration ranges is calculated, with a preset acceptable threshold of no less than 90%. Second, the vertical gradient consistency is evaluated, by calculating the ratio of the simulated concentration difference to the actual concentration difference between adjacent containers, with a preset acceptable threshold of no more than 20%. If both indicators reach the preset thresholds, the model is considered to be consistent with the diffusion characteristics, and the concentration range corresponding to the candidate pollutant set is used as the final estimation result. If either indicator fails to reach the preset threshold, the model is considered to be inconsistent with the diffusion characteristics, and the initial candidate pollutants are re-selected and subsequent calculations are performed until the model simulation results are consistent with the diffusion characteristics. Finally, the candidate pollutant set and its corresponding concentration range are output as the final estimation result.

Images