A gas sensor array, gas sensing system, and its application based on Fe-doped bicrystalline Nb₂O₅

Abstract

The application discloses a gas sensor array based on Fe-doped bimorphous phase Nb2O5, a gas sensing system and application thereof, and the gas sensor array comprises: a first sensing unit comprising a first gas-sensitive material (Fe-doped T-Nb2O5); and a second sensing unit comprising a second gas-sensitive material (Fe-doped H-Nb2O5); the working temperature of the first sensing unit and the second sensing unit is 300-400 DEG C. The gas sensing system comprises the gas sensor array, further comprises a constant-temperature chamber, a signal acquisition unit and a data processing unit, the data processing unit adopts a two-stage cascade machine learning model (XGBoost qualitative + random forest quantitative). Compared with the prior art, the gas sensor array and the gas sensing system break through the bottleneck that a single semiconductor sensor has poor selectivity and low stability in a complex flue gas (high temperature, high humidity and multi-component cross) environment, and provide a high-precision and low-cost in-situ sensing solution for fixed source carbon pollution collaborative stereoscopic online monitoring.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for carbon pollution in stationary source flue gas, and in particular to a gas sensor array, gas sensing system, and their applications based on Fe-doped bicrystalline Nb2O5. Background Technology

[0002] Flue gas emitted from stationary sources (such as coal-fired power plants, industrial boilers, and incinerators) contains various carbonaceous pollutants, including CO2, CO, volatile organic compounds (VOCs), and sulfur-containing pollutants. Real-time online monitoring of these gases is crucial for carbon emission reduction and pollution control. Acetone is a typical VOC component in flue gas, primarily originating from incomplete combustion of organic matter and photochemical reactions in the flue gas; hydrogen sulfide is a typical sulfur-containing odorous pollutant commonly found in industrial waste gases.

[0003] Current flue gas monitoring technologies primarily employ non-dispersive infrared (NDIR) spectroscopy. While this method offers high measurement accuracy, it suffers from drawbacks such as high equipment cost, complex maintenance, poor adaptability to high-humidity and high-dust environments, and difficulty in achieving simultaneous and rapid detection of multiple components. Metal oxide semiconductor gas sensors offer advantages such as low cost, small size, and fast response; however, traditional sensors face bottlenecks in complex flue gas environments, including poor selectivity, insufficient sensitivity, and susceptibility to moisture interference. Semiconductor gas sensors are gradually developing; for example, CN117554436A discloses a semiconductor micro gas sensor and its fabrication method, including a substrate and multiple sensing units disposed on the substrate. Each sensing unit includes an electrode assembly and a gas-sensitive membrane connected to the electrode assembly. The gas-sensitive membrane is made of metal-doped metal oxides, and at least two of the sensing units have different types and / or molar percentages of the metal in their gas-sensitive membranes. However, a more targeted sensor is still needed to address the challenge of distinguishing and detecting multi-component carbonaceous gases in complex flue gas environments.

[0004] Nb₂O₅, as a polycrystalline wide-bandgap semiconductor, exhibits gas-sensing properties closely related to its crystal structure. Studies have shown that Nb₂O₅ exists in multiple crystal phases, including T and H phases, with significant differences in interplanar spacing, oxygen vacancy concentration, and surface active sites among these phases, offering possibilities for selective gas identification. However, current research has not yet combined Fe doping with crystal phase modulation to address the challenge of distinguishing and detecting multi-component carbonaceous gases in complex flue gas environments.

[0005] Therefore, developing a gas sensor array and gas sensing system based on Fe-doped bicrystalline Nb2O5 to achieve high-precision in-situ measurement of carbon pollution characteristic gases (such as acetone and hydrogen sulfide) under high temperature, high humidity, and multi-component cross-interference conditions has important practical application value. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing flue gas monitoring technologies, such as poor selectivity, low stability, and difficulty in achieving simultaneous multi-component detection in complex environments (high temperature, high humidity, and cross-interference of multiple components). This invention provides a gas sensor array, gas sensing system, and its application based on Fe-doped bicrystalline Nb2O5. Specifically, it relates to a gas sensor array, gas sensing system, and multi-component detection method based on Fe-doped bicrystalline Nb2O5 for use in complex flue gas environments with fixed sources. This method is suitable for high-precision in-situ measurement of carbonaceous gases in complex flue gas environments with high temperature, high humidity, and cross-interference of multiple components.

[0007] The objective of this invention can be achieved through the following technical solutions: The first objective of this invention is to provide a gas sensor array based on Fe-doped bicrystalline Nb₂O₅, the gas sensor array comprising: The first sensing unit includes a first gas-sensitive material, which is Fe-doped T-Nb2O5; The second sensing unit includes a second gas-sensitive material, which is Fe-doped H-Nb2O5.

[0008] Furthermore, the operating temperature of the first sensing unit and the second sensing unit is 300-400 ℃.

[0009] Furthermore, the Fe-doped T-Nb2O5 and Fe-doped H-Nb2O5 were prepared using the Pechini gel method.

[0010] Furthermore, the preparation method of the Fe-doped T-Nb2O5 includes the following steps: Step A: Dissolve ammonium niobate in water, add ferric chloride hexahydrate (FeCl3·6H2O), and stir for 0.5-2 h until completely dissolved to obtain the precursor solution; Step B: Add citric acid (CA) to the precursor solution from Step A, and heat and stir at 50-75 °C for 0.5-2 h to allow the citric acid to react with Nb. 5+ The cations are completely complexed to obtain a complexed solution. Step C: Add ethylene glycol (EG) to the complexation solution from step B, heat to 90-110 °C and keep stirring for 0.5-2 h until the polymerization reaction is complete, and obtain a gel; Step D: The gel obtained in step C is pretreated by heating at 200-250 °C for 0.5-2 h to remove residual water and organic matter, and the pretreated product is obtained. Step E: Calcine the pretreated product obtained in step D at 600-800 °C for 1-4 h to obtain Fe-doped T-Nb2O5.

[0011] Furthermore, in the Fe-doped T-Nb2O5, the Fe doping molar ratio is 1-10%.

[0012] Furthermore, in step A, without adding ferric chloride hexahydrate, step E yields T-Nb2O5.

[0013] Further, in step A, the ratio of ammonium niobate oxalate to water is 0.1-0.5 g : 5-20 mL.

[0014] Further, in step A, the mass ratio of ferric chloride hexahydrate to ammonium niobate oxalate is 0.001-0.01 : 0.1-0.5.

[0015] Preferably, in step A, the mass-to-volume ratio of ammonium niobate oxalate to water is 0.3 g: 10 mL.

[0016] Preferably, in step A, the amount of ferric chloride hexahydrate added is 0.009 g (corresponding to 3% of the molar amount of Nb2O5 for Fe doping).

[0017] Preferably, in step A, the stirring time is 1 hour.

[0018] Preferably, in step B, the mixture is heated and stirred at 60 °C for 1 h.

[0019] Preferably, in step B, the molar ratio of citric acid to ammonium niobate oxalate is 3:1.

[0020] Preferably, in step C, the volume molar ratio of ethylene glycol to citric acid is 5 mL : 1 mol.

[0021] Preferably, in step C, the temperature is increased to 100 °C for 1 h.

[0022] Preferably, in step D, the sample is pretreated by heating at 220 °C for 1 h.

[0023] Preferably, in step E, the calcination is carried out at 700 °C for 2 h.

[0024] Furthermore, in step E, the heating rate of the calcination is 1-10 °C / min, and the calcination is carried out in an air atmosphere.

[0025] Furthermore, in step E, the heating rate of the calcination is 10 °C / min.

[0026] Furthermore, the preparation method of the Fe-doped H-Nb2O5 includes the following steps: Step A: Dissolve ammonium niobate in water, add ferric chloride hexahydrate (FeCl3·6H2O), and stir for 0.5-2 h until completely dissolved to obtain the precursor solution; Step B: Add citric acid (CA) to the precursor solution from Step A, and heat and stir at 50-75 °C for 0.5-2 h to allow the citric acid to react with Nb. 5+ The cations are completely complexed to obtain a complexed solution. Step C: Add ethylene glycol (EG) to the complexation solution from step B, heat to 90-110 °C and keep stirring for 0.5-2 h until the polymerization reaction is complete, and obtain a gel; Step D: The gel obtained in step C is pretreated by heating at 200-250 °C for 0.5-2 h to remove residual water and organic matter, and the pretreated product is obtained. Step E': Calcine the pretreated product obtained in step D at 900-1100 °C for 1-4 h to obtain Fe-doped H-Nb2O5.

[0027] Furthermore, in the Fe-doped H-Nb2O5, the Fe doping molar ratio is 1-10%.

[0028] Furthermore, in step A, without adding ferric chloride hexahydrate, step E' yields H-Nb2O5.

[0029] Further, in step A, the ratio of ammonium niobate oxalate to water is 0.1-0.5 g : 5-20 mL.

[0030] Further, in step A, the mass ratio of ferric chloride hexahydrate to ammonium niobate oxalate is 0.001-0.01 : 0.1-0.5.

[0031] Preferably, in step A, the mass-to-volume ratio of ammonium niobate oxalate to water is 0.3 g : 10 mL.

[0032] Preferably, in step A, the amount of ferric chloride hexahydrate added is 0.009 g (corresponding to 3% of the molar amount of Nb2O5 for Fe doping).

[0033] Preferably, in step A, the stirring time is 1 hour.

[0034] Preferably, in step B, the mixture is heated and stirred at 60 °C for 1 h.

[0035] Preferably, in step B, the molar ratio of citric acid to ammonium niobate oxalate is 3:1.

[0036] Preferably, in step C, the volume molar ratio of ethylene glycol to citric acid is 5 mL : 1 mol.

[0037] Preferably, in step C, the temperature is increased to 100 °C for 1 h.

[0038] Preferably, in step D, the pretreatment is carried out at 220 °C for 2 h.

[0039] Preferably, in step E, the calcination is carried out at 1050 °C for 2 h.

[0040] Furthermore, in step E, the heating rate of the calcination is 1-10 °C / min, and the calcination is carried out in an air atmosphere.

[0041] Furthermore, in step E, the heating rate of the calcination is 10 °C / min.

[0042] Furthermore, different crystalline phases were selectively obtained by controlling the calcination temperature: T-phase Nb2O5 was obtained by calcination at 600-800 ℃, and H-phase Nb2O5 was obtained by calcination at 900-1100 ℃.

[0043] Furthermore, the Fe-doped bicrystalline Nb₂O₅-based gas sensor array is used for the synergistic detection of carbon pollution in complex flue gas environments with fixed sources. The Fe-doped bicrystalline Nb₂O₅-based gas sensor array can achieve differentiated responses to typical characteristic gases of carbon pollution (such as acetone representing VOCs and hydrogen sulfide representing sulfur-containing pollutants). The second objective of this invention is to provide a gas sensing system for the synergistic detection of carbon pollution in complex flue gas environments with stationary sources. The gas sensing system includes the aforementioned gas sensor array based on Fe-doped bicrystalline Nb₂O₅, and further includes: The signal acquisition unit is used to acquire the response current signals of the first sensing unit and the second sensing unit under a simulated complex flue gas atmosphere (containing 5-20 vol.% moisture, with coexisting interfering gases such as CO2, CO, and SO2) and generate the It response curve. The data processing unit employs a two-level cascaded machine learning model to qualitatively identify and quantitatively predict characteristic gases of carbon pollution.

[0044] Preferably, the amount of Fe doping is 3% of the molar amount of Nb2O5.

[0045] Furthermore, in the two-level cascaded machine learning model, the first level uses a first machine learning algorithm to qualitatively identify gas types (distinguish carbon pollution characteristic components), and the second level calls the corresponding second machine learning algorithm to quantitatively predict concentration based on the qualitative identification results of gas types, correcting cross-interference.

[0046] Furthermore, the signal acquisition unit is used to acquire the response current signals of the first sensing unit and the second sensing unit under a complex flue gas atmosphere from a fixed source, and generate an It response curve.

[0047] Furthermore, the moisture content of the stationary source complex flue gas atmosphere is 5-20 vol.%.

[0048] Furthermore, the stationary source complex flue gas atmosphere includes acetone, hydrogen sulfide, and interfering gases.

[0049] Furthermore, the interfering gases in the complex flue gas atmosphere from the stationary source include CO2, CO, SO2, etc. That is, the complex flue gas atmosphere from the stationary source contains interfering gases such as CO2, CO, and SO2.

[0050] Furthermore, the first machine learning algorithm is the XGBoost model; the second machine learning algorithm is the Random Forest model. That is, the two-level cascaded machine learning model includes: using the first-level machine learning algorithm, the XGBoost model, for qualitative identification, and using the second-level machine learning algorithm, the Random Forest model, for quantitative prediction.

[0051] Furthermore, the first sensing unit and the second sensing unit are independently disposed on two heated interdigital electrodes, and the gas-sensitive material is coated on the surface of the interdigital electrodes, that is, the first gas-sensitive material is coated on the surface of one interdigital electrode and the second gas-sensitive material is coated on the surface of the other interdigital electrode.

[0052] Furthermore, the coating method is drop coating, and the dispersion is prepared as follows: 5-10 mg / L of the gas-sensitive composite material is dispersed in 1 mL of a water and ethanol mixed solvent with a volume ratio of 1:1 to obtain a suspension with a concentration of 5-10 mg / mL, and 10-30 µL of 5wt% Nafion solution is added to each mL of suspension to improve the viscosity.

[0053] Preferably, the coating method is drop coating, the concentration of the dispersion used is 10 mg / mL, the solvent is water and ethanol in a volume ratio of 1:1, and 20 µL of 5wt% Nafion solution is added.

[0054] Furthermore, the gas sensing system also includes: The constant temperature chamber is used to stably control the operating temperature of the first and second sensing units at 300-400℃.

[0055] Furthermore, the data processing unit is also configured to: extract the peak response ratio and recovery time ratio based on the It response curves of the first sensing unit and the second sensing unit; when the peak response ratio and / or recovery time ratio deviate from the preset proportion range corresponding to the known gas in the fixed source complex flue gas atmosphere, it is determined that there is an unknown interference component, and the peak response ratio, recovery time ratio and the time series features of the It response curve are jointly input into the first-level machine learning model to identify the type of unknown interference component.

[0056] The third objective of this invention is to provide an application of a gas sensing system for the synergistic detection of carbon pollution in complex flue gas environments with fixed sources.

[0057] Furthermore, the atmosphere of the stationary source complex flue gas environment is a stationary source complex flue gas atmosphere.

[0058] Furthermore, the moisture content of the stationary source complex flue gas atmosphere is 5-20 vol.%.

[0059] Furthermore, the stationary source complex flue gas atmosphere includes acetone, hydrogen sulfide, and interfering gases. Acetone represents VOCs, and hydrogen sulfide represents sulfur-containing pollutants.

[0060] Furthermore, the interfering gases in the complex flue gas atmosphere of the stationary source include CO2, CO, and SO2.

[0061] Furthermore, the first sensing unit and the second sensing unit can be integrated into an intelligent airborne "visual-olfactory" dual sensor platform for tracking the three-dimensional spatial gas concentration distribution of the plume from a fixed source exhaust.

[0062] The fourth objective of this invention is to provide a method for distinguishing and detecting acetone and hydrogen sulfide (the specific process of using the gas sensing system for the synergistic detection of carbon pollution in a complex flue gas environment with a stationary source), comprising the following steps: Step 1: The gas to be tested (complex flue gas atmosphere from a fixed source) is introduced into the gas sensor array based on Fe-doped bicrystalline Nb2O5 in the above gas sensing system, and the response signals (response current signals) of the first sensing unit and the second sensing unit are recorded by the signal acquisition unit respectively. Step 2: Input the two response signals into the trained first-level machine learning model, namely the XGBoost model, to perform qualitative analysis of the gas types and determine the gas type; Step 3: Based on the output of the XGBoost model, input the response signal into the second-level machine learning model, namely the random forest model, to predict the concentration of this type of gas.

[0063] Preferably, the training data for the XGBoost model and the random forest model are the response values ​​of acetone gas, hydrogen sulfide gas, and mixtures of acetone gas and hydrogen sulfide gas at different concentrations in the first and second sensing units.

[0064] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention provides a gas sensor array, a gas sensing system, and its applications based on Fe-doped biphase Nb2O5. By precisely controlling the calcination temperature, selective modulation of the Nb2O5 crystal phase is achieved. In the prior art, Fe-doped Nb2O5 typically does not distinguish between crystal phases or only obtains a single crystal phase. However, this invention prepares T / H phase Nb2O5 separately and combines them into a dual sensing unit, utilizing the difference in their crystal structures to achieve differentiated responses to different gases. This significantly improves the gas-sensing performance of Nb2O5. Compared to undoped H-phase Nb2O5, Fe-doped H-phase Nb2O5 shows a 4.9-fold increase in response value to acetone and a 5.1-fold increase in response value to hydrogen sulfide. Simultaneously, the gas sensing system of this invention has detection limits as low as 60 and 25 ppb for acetone and hydrogen sulfide, respectively, demonstrating excellent performance with high sensitivity and low detection limits.

[0065] (2) This invention provides a gas sensor array, a gas sensing system and its application based on Fe-doped bicrystalline Nb2O5. The gas-sensitive material is directly drop-coated onto the surface of the heated interdigitated electrode. A dispersion solvent of water and ethanol in a volume ratio of 1:1 is used and 5wt% Nafion is added as a binder, which improves the uniformity and adhesion of the coating. The gas sensor array can obtain a stable and repeatable response signal at an operating temperature of 300-400 ℃ (preferably 325 ℃), avoiding the energy consumption problem caused by excessively high temperature.

[0066] (3) This invention provides a gas sensor array, a gas sensing system and its application based on Fe-doped bicrystalline Nb2O5. Fe-doped Nb2O5 is prepared by Pechini gel method. The crystal phase can be precisely controlled by simply adjusting the calcination temperature (600-800 ℃ to obtain T phase, 900-1100 ℃ to obtain H phase). The preparation process is simple and has good repeatability, avoiding complex synthesis steps and is suitable for large-scale production.

[0067] (4) This invention provides a gas sensor array, a gas sensing system and its application based on Fe-doped bicrystalline Nb2O5. The constructed gas sensing system includes a constant temperature chamber, a signal acquisition unit and a data processing unit. The constant temperature chamber stably controls the working temperature of the dual sensing units at 300-400 ℃, effectively resisting temperature fluctuations in the fixed source flue. The signal acquisition unit acquires the It response curves of the dual sensing units in real time under complex flue gas atmosphere (moisture content 5-20 vol.%, coexisting interfering gases such as CO2, CO, SO2, etc.). The two-level cascaded machine learning model is used to identify gas types and predict concentrations online, providing a high-precision and low-cost in-situ sensing solution for the coordinated online monitoring of carbon pollution in fixed source flue gas.

[0068] (5) This invention provides a gas sensor array, a gas sensing system, and its application based on Fe-doped bicrystalline Nb2O5. The machine learning method of this invention differs from conventional single or parallel models; instead, it employs a two-stage cascaded architecture: the first-stage XGBoost classifier takes the humidity- and temperature-compensated response signal of the dual-sensor unit as input and outputs the types of carbon-polluting gases (such as acetone representing VOCs or hydrogen sulfide representing sulfur-containing pollutants); the second stage selects the corresponding random forest regressor for quantitative concentration prediction based on the output of the first stage. This qualitative-to-quantitative cascaded strategy is specifically designed for real-world scenarios where the types and concentrations of gases in stationary source flue gas are unknown and environmental interference is strong, resulting in higher recognition accuracy and prediction precision compared to a single model. Combining this architecture with Fe-doped T / H-Nb2O5 dual-sensor units (a gas sensor array based on Fe-doped bicrystalline Nb2O5) and a constant-temperature chamber significantly improves the accuracy and reliability of distinguishing and detecting carbon-polluting gases in complex flue gas environments. Attached Figure Description

[0069] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram illustrating the preparation process of Fe-doped T-phase Nb2O5 (Fe-doped T-Nb2O5) and Fe-doped H-phase Nb2O5 (Fe-doped H-Nb2O5) of the present invention.

[0070] Figure 2 This is a schematic diagram of a gas sensing system.

[0071] Figure 3The X-ray diffraction (XRD) comparison patterns of Fe-doped T-phase Nb2O5 (Fe-doped T-Nb2O5) and Fe-doped H-phase Nb2O5 (Fe-doped H-Nb2O5) prepared in Example 1 of this invention are shown; where (a) is the T phase and (b) is the H phase.

[0072] Figure 4 The image shows a comparison of the Raman spectra of Fe-doped T-phase Nb2O5 (Fe-doped T-Nb2O5) and Fe-doped H-phase Nb2O5 (Fe-doped H-Nb2O5) prepared in Example 1 of this invention; where (a) is the T phase and (b) is the H phase.

[0073] Figure 5 The images are scanning electron microscope (SEM) images of Fe-doped T-phase Nb2O5 (Fe-doped T-Nb2O5) and Fe-doped H-phase Nb2O5 (Fe-doped H-Nb2O5) prepared in Example 1 of this invention, where (a) is the T phase and (b) is the H phase.

[0074] Figure 6 The images show the dynamic response curves of the first sensing unit (Fe-T phase, Fe-doped T-Nb2O5) and the second sensing unit (Fe-H phase, Fe-doped H-Nb2O5) to different concentrations of acetone at an operating temperature of 325 °C for the gas sensing system of the present invention; where (a) is the T phase and (b) is the H phase.

[0075] Figure 7 The figures show the dynamic response curves of the first and second sensing units of the gas sensing system of the present invention to different concentrations of hydrogen sulfide at an operating temperature of 325 °C; where (a) represents the T phase and (b) represents the H phase.

[0076] Figure 8 This is a confusion matrix diagram of qualitative identification of acetone and hydrogen sulfide using the XGBoost model in Embodiment 1 of the present invention.

[0077] Figure 9 This is a performance comparison chart of Example 1 of the present invention (including Fe-doped T-Nb2O5 and Fe-doped H-Nb2O5) and the comparative examples (undoped T phase and undoped H phase) in terms of gas sensing response value and selectivity.

[0078] Figure 10 The sensing effect of Comparative Example 3 of the present invention (Fe-doped T-phase Nb2O5 and Fe-doped H-phase Nb2O5) without machine learning is shown in (a) and (b) without machine learning.

[0079] Figure 11This is a confusion matrix diagram of the qualitative identification of acetone and hydrogen sulfide using the XGBoost model in Comparative Example 4 of this invention. Detailed Implementation

[0080] The present invention will now be described in detail with reference to specific embodiments. These embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Component models, material names, connection structures, control methods, and other features not explicitly stated in this technical solution are considered to be common technical features disclosed in the prior art.

[0081] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may change. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0082] It should be noted that in this invention, 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 said element.

[0083] This invention belongs to the field of online monitoring technology for carbon pollution in stationary source flue gas, specifically involving a gas sensor and multi-component detection system suitable for complex environments (high temperature, high humidity, and multi-component cross-interference). Addressing the challenge of high-precision in-situ measurement of carbon pollution gases (such as CO2, CO, and volatile organic compounds) in stationary source flue gas, this invention develops an Fe-doped T-Nb2O5 and H-Nb2O5 dual-phase sensing unit. Utilizing the synergistic effect of Fe doping and phase modulation, it achieves differentiated responses to typical carbon pollution characteristic gases (such as acetone representing VOCs and hydrogen sulfide representing sulfur-containing pollutants). Furthermore, by combining a two-stage cascaded machine learning model (XGBoost qualitative + random forest quantitative), ppb-level detection (60 ppb for acetone and 25 ppb for hydrogen sulfide) is achieved in a simulated complex flue gas environment (humidity of 5-20 vol.%) with coexisting interfering gases. This invention overcomes the bottleneck of poor selectivity and low stability of single semiconductor sensors in complex flue gas environments, and provides a high-precision, low-cost in-situ sensing solution for the collaborative three-dimensional online monitoring of carbon pollution from stationary sources.

[0084] This invention first provides methods for preparing the first sensing unit and the second sensing unit: the gas-sensitive material of the first sensing unit is Fe-doped T-Nb2O5, and the gas-sensitive material of the second sensing unit is Fe-doped H-Nb2O5. A schematic diagram of the preparation process of the Fe-doped T-Nb2O5 and Fe-doped H-Nb2O5 is shown below. Figure 1 As shown, the gel was prepared using the Pechini gelation method: Step A: Dissolve ammonium niobate oxalate in water, add ferric chloride hexahydrate (FeCl3·6H2O), and stir for 0.5-2 h until completely dissolved, more preferably 1 h. This yields a precursor solution. The preferred mass-to-volume ratio of ammonium niobate oxalate to water is 0.3 g : (5-20) mL, more preferably 0.3 g : 10 mL; the amount of ferric chloride hexahydrate added corresponds to 3% of the molar amount of Nb2O5 doping, preferably 0.009 g.

[0085] Step B: Add citric acid (CA) to the precursor solution from Step A, and heat and stir at 50-75 °C, more preferably 60 °C, for 0.5-2 h, more preferably 1 h, to allow the citric acid to react with Nb. 5+ The cations are completely complexed to obtain a complexed solution. The molar ratio of citric acid to ammonium niobate oxalate is preferably (1-5):1, more preferably 3:1.

[0086] Step C: Add ethylene glycol (EG) to the complexing solution from Step B, heat to 100 °C and stir for 0.5-2 h, more preferably 1 h, until the polymerization reaction is complete, and obtain a gel. The preferred volume molar ratio of ethylene glycol to citric acid is (1-10) mL : 1 mol, more preferably 5 mL : 1 mol.

[0087] Step D: The gel obtained in step C is pretreated at 220 °C to remove residual water and organic matter, yielding the pretreated product. The pretreatment time is preferably 0.5-2 h, more preferably 1 h.

[0088] Step E: The product pretreated in Step D is calcined at 700 °C for 2 h to obtain Fe-doped T-Nb₂O₅; and the product pretreated in Step D is calcined at 1050 °C for 2 h to obtain Fe-doped H-Nb₂O₅. The heating rate of the calcination is preferably 1-10 °C / min, more preferably 10 °C / min; the calcination atmosphere is preferably air.

[0089] The first and second sensing units are independently disposed on two heated interdigital electrodes. The gas-sensitive material is preferably coated onto the surface of the interdigital electrodes by drop coating, dispersing the gas-sensitive material in a solvent to obtain a dispersion. The concentration of the gas-sensitive material in the dispersion is 10 mg / mL, the solvent is water and ethanol in a 1:1 volume ratio, and 5 wt% Nafion solution is added (20 μL per 1 mL of dispersion). The operating temperature of the gas sensor is preferably 325 °C.

[0090] The present invention further provides a gas sensing system for distinguishing and detecting acetone and hydrogen sulfide, the schematic diagram of which is shown below. Figure 2 As shown, the system includes a gas sensor array (containing a first sensing unit and a second sensing unit), a constant-temperature chamber, a signal acquisition unit, and a data processing unit. The constant-temperature chamber contains two independent thermocouples 1 and 2 to stabilize the ambient temperature at 300-400 °C, eliminating interference from ambient temperature fluctuations on the sensing signal. Both the first and second sensing units integrate heating resistors to heat the gas-sensitive material coated on the surface of the sensing unit to an operating temperature of 325 °C, ensuring stable gas-sensitive response activity at this target temperature. The signal acquisition unit and data processing unit electrically connect current signal acquisition channel 1 to the two electrodes of the first sensing unit to apply a constant bias voltage and acquire its response current in real time. Current signal acquisition channel 2 is connected to the second sensing unit in the same manner to acquire information. An electrochemical workstation is communicatively connected to the data processing unit to transmit the acquired two current signals to the data processing unit for feature extraction and pattern recognition. The process includes the following steps: The constant temperature chamber is used to stably control the operating temperature of the first and second sensing units at 300-400 ℃. The signal acquisition unit is used to acquire the response current signals of the first and second sensing units to the measured gas and generate the It response curve. The data processing unit is used to receive the It response curve and use a machine learning model to perform qualitative identification and quantitative detection of acetone and hydrogen sulfide. The machine learning model preferably includes: using an XGBoost model for qualitative identification and a random forest model for quantitative prediction.

[0091] The present invention also provides the application of the above-described gas sensing system in distinguishing and detecting acetone and hydrogen sulfide.

[0092] The embodiments of the present invention are described in detail below. These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments.

[0093] In the examples below, unless otherwise specified, the reagents used are commercially available products and the methods employed are those known in the art.

[0094] Example 1 This embodiment provides a gas sensing system (Fe-doped T-phase Nb2O5 and Fe-doped H-phase Nb2O5 dual gas sensing units, combined with a machine learning algorithm), including the following steps: (1) Preparation of precursor solution: Weigh 0.3 g of ammonium niobate oxalate ((NH4)2[Nb(C2O4)4]·2H2O) and dissolve it in 10 mL of deionized water. Stir until completely dissolved to obtain a niobium precursor solution. Weigh 0.009 g of ferric chloride hexahydrate (FeCl3·6H2O) and dissolve it in the above niobium precursor solution. Stir until completely dissolved to obtain an Fe-doped niobium precursor solution. Take another 0.3 g of ammonium niobate oxalate and dissolve it in 10 mL of deionized water without adding iron salt (ferric chloride hexahydrate) as an undoped control.

[0095] Citric acid (CA) was weighed as a complexing agent, with a molar ratio of citric acid to ammonium niobate oxalate of 3:1. Ethylene glycol (EG) was measured as a polymerization agent, with a volume molar ratio of ethylene glycol to citric acid of 5 mL:1 mol.

[0096] (2) Preparation of Fe-doped Nb2O5 (Fe-doped T-Nb2O5 and Fe-doped H-Nb2O5, respectively) by Pechini gel method: The above Fe-doped niobium precursor solution was placed in a 60 °C water bath, citric acid was added, and the mixture was heated and stirred at 60 °C for 1 h to allow the citric acid to react with Nb2O5. 5+The cations were completely complexed. Ethylene glycol was then added to the solution, and the temperature was raised to 100 °C and stirred for 1 h until the polymerization reaction was complete, yielding a gel. The gel was transferred to a 220 °C oven for pretreatment for 1 h to remove residual water and organic matter, yielding the pretreated product. The pretreated product was divided into two portions and calcined at 700 °C and 1050 °C for 2 h respectively (heating rate 10 °C / min, air atmosphere) to obtain Fe-doped T-Nb₂O₅ (700 °C) and Fe-doped H-Nb₂O₅ (1050 °C), respectively. The crystal phases of the products were characterized by X-ray diffraction (XRD) and Raman spectroscopy. Figure 3-4 The XRD patterns show that the diffraction peak positions of Fe-doped T-Nb₂O₅ and Fe-doped H-Nb₂O₅ are significantly different, confirming the successful control of the crystal phase. The morphology of the material was observed using scanning electron microscopy (SEM). Figure 5 It can be clearly observed that Fe-doped T-Nb2O5 exhibits nanoscale characteristics, while Fe-doped H-Nb2O5 exhibits larger crystal particles with increasing heating temperature. Using both Fe-doped T-Nb2O5 and Fe-doped H-Nb2O5 as the gas-sensitive materials in the first sensing unit, a gas sensor array based on Fe-doped bicrystalline Nb2O5 is formed.

[0097] (3) Sensing unit assembly: The Fe-doped T-phase Nb2O5 and Fe-doped H-phase Nb2O5 prepared above were dispersed in a mixed solvent of water and ethanol at a volume ratio of 1:1, with a concentration of 10 mg / mL. 20 μL of 5wt% Nafion solution was added to each 1 mL of dispersion, and the mixture was ultrasonically dispersed for 30 minutes to obtain a uniform dispersion. Two interdigital electrodes (3 mm × 3 mm) with heating resistors were taken, and 20 μL of the uniform dispersion was measured with a pipette and dropped onto the surface of the interdigital electrodes. After natural drying, the first sensing unit (Fe-doped T-phase Nb2O5) and the second sensing unit (Fe-doped H-phase Nb2O5) were obtained. The first and second sensing units were placed in the test gas chamber of the constant temperature chamber. The current signal acquisition channel 1 was electrically connected to the two electrodes of the first sensing unit to apply a constant bias voltage to the first sensing unit and acquire its response current in real time. The current signal acquisition channel 2 was connected to the second sensing unit in the same way to acquire information. The electrochemical workstation includes a current signal acquisition channel 1 and a current signal acquisition channel 2. The electrochemical workstation is communicatively connected to the data processing unit to transmit the acquired two current signals to the data processing unit for feature extraction and pattern recognition. (4) Gas sensing performance test: The operating temperature of the gas sensing system was stably controlled at 300-400 ℃ (325 ℃ in this embodiment) through a constant temperature chamber. A simulated complex flue gas atmosphere (simulating a complex flue gas atmosphere from a fixed source) was introduced into the test gas chamber: the moisture content was 15 vol.% (adjusted by bubbling method), and CO2, CO, and SO2 coexisting interference gases with a concentration of 1 ppm were added. On this basis, acetone or hydrogen sulfide target gases of different concentrations were introduced respectively, and the current-time (It) response curves of the first sensing unit and the second sensing unit were recorded. The target gases were provided by gas cylinders containing acetone and hydrogen sulfide standard gases. These were prepared using a gas mixing device at the following concentration gradients and introduced into the test chamber: 200 ppb, 300 ppb, 400 ppb, 500 ppb, 600 ppb, 700 ppb, 800 ppb, 900 ppb, and 1000 ppb. Each time, a single concentration of the target gas was introduced for 5 minutes, followed by 5 minutes of air introduction into the test chamber for recovery before proceeding to the next concentration test. The test results are as follows: Figure 6 , 7 As shown. The results indicate that under complex flue gas conditions, the response values ​​of the first sensing unit (Fe-doped T-phase Nb₂O₅) to 1 ppm acetone and 1 ppm hydrogen sulfide are 1.9 and 1.5, respectively; the response values ​​of the second sensing unit (Fe-doped H-phase Nb₂O₅) to 1 ppm acetone and 1 ppm hydrogen sulfide are 5.9 and 6.5, respectively. The two sensing units still exhibit significant differences in response characteristics to the two gases. Figure 9 As shown, comparing Fe-doped T-phase Nb₂O₅ and Fe-doped H-phase Nb₂O₅ with their respective undoped T-phase Nb₂O₅ and H-phase Nb₂O₅ samples, it can be seen that the improvement in gas sensing performance for acetone and hydrogen sulfide by Fe-doped T-phase Nb₂O₅ compared to single T-phase Nb₂O₅ is not significant. However, Fe-doped H-phase Nb₂O₅ shows a significant improvement compared to single H-phase Nb₂O₅, with a 4.9-fold increase in response value for acetone and a 5.1-fold increase in response value for hydrogen sulfide. This gas sensing system achieves a detection limit of 60 ppb for acetone and 25 ppb for hydrogen sulfide in complex flue gas environments.

[0098] (5) Machine learning-based detection: Steady-state response values ​​of the first and second sensing units at different concentrations of acetone and hydrogen sulfide were collected, and ambient humidity and temperature sensor data were collected simultaneously. The dataset contained test data at the following concentration gradients: acetone 200 ppb, 300 ppb, 400 ppb, 500 ppb, 600 ppb, 700 ppb, 800 ppb, 900 ppb, 1000 ppb; hydrogen sulfide 200 ppb, 300 ppb, 400 ppb, 500 ppb, 600 ppb, 700 ppb, 800 ppb, 900 ppb, 1000 ppb; and acetone and hydrogen sulfide mixtures at the above concentration gradients. During the steady-state response phase, sampling was performed every 10 seconds, for a total of 540 sets of data. The corrected dataset was divided into a training set and a test set in an 8:2 ratio. A two-stage cascaded machine learning model is employed: the first stage uses an XGBoost classifier for qualitative identification, with input parameters being the corrected steady-state response values ​​of the first and second sensing units, and output parameters being gas type labels. These gas type labels are uniquely encoded and include three categories: acetone, hydrogen sulfide, and mixed gases. In the second stage, based on the gas type labels output from the first stage, a corresponding random forest regressor is invoked for quantitative concentration prediction. The input parameters are the corrected steady-state response values ​​of the first and second sensing units, and the output parameter is the concentration value of the target gas. Figure 8 As shown, the XGBoost confusion matrix indicates a classification accuracy of 90% for acetone and hydrogen sulfide. These results demonstrate that the constant-temperature chamber, dual-sensor unit, and two-stage cascaded machine learning and humidity / temperature compensation methods constructed in this invention can effectively distinguish and quantitatively detect acetone and hydrogen sulfide in complex flue gas environments.

[0099] Comparative Example 1 This comparative example provides a gas sensing system (undoped H-phase Nb2O5), comprising the following steps: (1) Material preparation: Undoped H-phase Nb2O5 (undoped H-Nb2O5) was prepared using the same Pechini gelation method as in Example 1, except that iron salt (ferric chloride hexahydrate) was not added. 0.3 g of ammonium niobate oxalate was weighed and dissolved in 10 mL of deionized water and stirred for 1 h until completely dissolved. Citric acid (CA) was added, with a molar ratio of citric acid to ammonium niobate oxalate of 3:1, and the mixture was heated and stirred at 60 °C for 1 h. Then ethylene glycol (EG) was added, with a volume molar ratio of ethylene glycol to citric acid of 5 mL: 1 mol, and the mixture was heated to 100 °C and stirred for 1 h to obtain a gel. The gel was pretreated at 220 °C for 1 h and then calcined at 1050 °C for 2 h (heating rate 10 °C / min, air atmosphere) to obtain undoped H-phase Nb2O5.

[0100] (2) Assembly and performance testing of the sensing unit: The undoped H-phase Nb2O5 was dispersed in a solvent of water:ethanol = 1:1 (volume ratio) at a concentration of 10 mg / mL, and then 5wt% Nafion solution (20 μL per 1 mL dispersion) was added. It was drop-coated onto the heated interdigital electrode and allowed to dry naturally to obtain the sensing unit. The prepared sensing unit was placed in the test gas chamber of the constant temperature chamber; the working electrode and the counter electrode were electrically connected to the signal acquisition unit; the signal acquisition unit was used to apply a bias voltage to each sensing unit and acquire the response current signal; the signal acquisition unit was communicatively connected to the data processing unit to transmit the acquired current signal to the data processing unit for analysis and processing. The heating temperature and gas introduction method were the same as in Example 1. (3) Test results: Figure 9 The undoped H-phase Nb₂O₅ sensor showed a response value of 1.2 for 1 ppm acetone and 1.19 for 1 ppm hydrogen sulfide. Compared to the Fe-doped H-phase Nb₂O₅ in Example 1, the response values ​​decreased by 4.9 times and 5.1 times, respectively. The response curves for the two gases showed no significant difference and could not be effectively distinguished.

[0101] Comparative Example 2 This comparative example provides a gas sensing system (undoped T-phase Nb2O5), comprising the following steps: (1) Material preparation: Same as Comparative Example 1, but the calcination temperature is 700 °C to obtain undoped T-phase Nb2O5 (undoped T-Nb2O5).

[0102] (2) Sensor unit assembly and performance testing: Same as comparative example 1.

[0103] (3) Test results: such as Figure 9 The undoped T-phase Nb2O5 sensor had a response value of 1.06 to 1 ppm acetone and a response value of 1.04 to 1 ppm hydrogen sulfide, which were also much lower than those of the Fe-doped T-phase Nb2O5 in Example 1.

[0104] Comparative Example 3 This comparative example provides a gas sensing system (providing only dual gas sensing units, without incorporating machine learning), including the following steps: (1) Sensing unit preparation: Only the Fe-doped T-phase Nb2O5 and Fe-doped H-phase Nb2O5 prepared in Example 1 were used to assemble single sensing units in the same way, without using any machine learning algorithm.

[0105] (2) Gas sensing test: At a working temperature of 325 °C, acetone and hydrogen sulfide of different concentrations were introduced and the response signals were recorded.

[0106] (3) Test results: such as Figure 10 (a) and (b) show the responses of Fe-doped T-phase Nb₂O₅ and Fe-doped H-phase Nb₂O₅ to different concentrations of acetone and hydrogen sulfide, respectively. It can be seen that both Fe-doped T-phase Nb₂O₅ and Fe-doped H-phase Nb₂O₅ produce significant responses to acetone and hydrogen sulfide, but the response curves for the two gases highly overlap, making it impossible to accurately identify the gas species and thus impossible to quantify them. For example, taking Fe-doped H-phase Nb₂O₅ as an example, it can be seen that the response value to 1 ppm acetone is 5.9, and the response value to 1 ppm hydrogen sulfide is 6.1. The two values ​​are similar, and their trends with concentration are almost identical. Figure 10 It can be seen that although the slopes of the standard curves for the two gases are different, since there is only one response signal, it is impossible to distinguish whether the gas to be tested is acetone or hydrogen sulfide without the addition of a machine learning algorithm. That is, there is cross-interference between the two. This result shows that relying solely on dual sensing units, no matter how high the sensitivity, cannot solve the problem of gas type identification.

[0107] Comparative Example 4 This comparative example provides a gas sensing system (a dual gas sensing unit consisting of undoped T-phase Nb2O5 and undoped H-phase Nb2O5, combined with a machine learning algorithm), comprising the following steps: (1) Sensing unit preparation: Undoped T-phase Nb2O5 and undoped H-phase Nb2O5 were assembled into single sensing units using the same method as in Example 1, and the same machine learning algorithm was used.

[0108] (2) Gas sensing test: At a working temperature of 325 °C, acetone and hydrogen sulfide of different concentrations were introduced and the response signals were recorded.

[0109] (3) Test results: such as Figure 11 As shown, the undoped Nb₂O₅ sensor array, using the same machine learning method, identified hydrogen sulfide, acetone, and their mixtures. Its classification accuracy, corresponding to the confusion matrix, was only about 60%, far lower than the 90% achieved by the doped sensor array. This may be because the undoped sensor's gas-sensitive response signal is weaker, its noise level is higher, and the boundaries between the response feature values ​​of different gases are blurred and fluctuate significantly, making it difficult for the machine learning model to effectively extract discriminative features. Especially in the identification of mixed gases, a large number of samples were misclassified as single-component gases, resulting in poor generalization performance.

[0110] The above examples and comparative examples show that Fe doping significantly improves the response values ​​of Nb₂O₅ to acetone and hydrogen sulfide. The response values ​​of Fe-doped H-Nb₂O₅ are 4.9 times (acetone) and 5.1 times (hydrogen sulfide) higher than those of the undoped sample. Simultaneously, the response values ​​of Fe-doped T-phase Nb₂O₅ are also significantly better than those of the undoped T-phase. This indicates that Fe doping introduces active sites and oxygen vacancies, enhancing gas adsorption and surface reactions.

[0111] This also proves the necessity of dual-sensor units + machine learning. While the single-sensor unit in Comparative Example 3 had high sensitivity (thanks to Fe doping), it could not distinguish between acetone and hydrogen sulfide. In contrast, Example 1, by constructing Fe-doped T-phase and H-phase dual-sensor units, obtained two independent differential response signals. Combined with the XGBoost qualitative model and the random forest quantitative model, it successfully achieved high-accuracy differentiation between the two gases (classification accuracy as high as 90%). Therefore, the core of this invention in solving the differentiation detection problem lies in "dual-phase Fe-doped Nb2O5 dual-sensor unit + machine learning," rather than simply improving sensitivity.

[0112] Example 1 was significantly superior to all comparative examples in terms of response value, detection limit (acetone 60 ppb, hydrogen sulfide 25 ppb) and selectivity.

[0113] In summary, the gas sensing system provided by this invention has high sensitivity, low detection limit, fast response, and distinguishable detection capabilities, and has important application value in fields such as environmental monitoring and industrial safety.

[0114] The above description of the embodiments is provided to enable those skilled in the art to understand and use the invention. It will be apparent to those skilled in the art that various modifications can be made to these embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the invention should be within the protection scope of the present invention.

Claims

1. A gas sensor array based on Fe-doped bicrystalline Nb₂O₅, characterized in that, The gas sensor array includes: The first sensing unit includes a first gas-sensitive material, which is Fe-doped T-Nb2O5; The second sensing unit includes a second gas-sensitive material, which is Fe-doped H-Nb2O5; The operating temperature of the first sensing unit and the second sensing unit is 300-400 ℃.

2. The gas sensor array based on Fe-doped bicrystalline Nb₂O₅ according to claim 1, characterized in that, The preparation method of the Fe-doped T-Nb2O5 includes the following steps: Step A: Dissolve ammonium niobate oxalate in water, add ferric chloride hexahydrate, and stir until completely dissolved to obtain a precursor solution; Step B: Add citric acid to the precursor solution from Step A, heat at 50-75 °C and stir to obtain a complex solution; Step C: Add ethylene glycol to the complexing solution from step B, heat to 90-110 °C and keep stirring to obtain a gel; Step D: The gel obtained in step C is pretreated by heating at 200-250 °C to obtain the pretreated product; Step E: Calcine the pretreated product obtained in step D at 600-800 °C for 1-4 h to obtain Fe-doped T-Nb2O5.

3. The gas sensor array based on Fe-doped bicrystalline Nb₂O₅ according to claim 2, characterized in that, In the Fe-doped T-Nb2O5, the Fe doping molar ratio is 1-10%; In step A, the ratio of ammonium niobate oxalate to water is 0.1-0.5 g : 5-20 mL; In step A, the mass ratio of ferric chloride hexahydrate to ammonium niobate oxalate is 0.001-0.01 : 0.1-0.5; In step E, the calcination heating rate is 1-10 °C / min, and the calcination is carried out in an air atmosphere.

4. The gas sensor array based on Fe-doped bicrystalline Nb₂O₅ according to claim 1, characterized in that, The preparation method of the Fe-doped H-Nb2O5 includes the following steps: Step A: Dissolve ammonium niobate oxalate in water, add ferric chloride hexahydrate, and stir until completely dissolved to obtain a precursor solution; Step B: Add citric acid to the precursor solution from Step A, heat and stir at 50-75 °C to obtain a complex solution; Step C: Add ethylene glycol to the complexing solution from step B, heat to 90-110 °C and keep stirring to obtain a gel; Step D: The gel obtained in step C is pretreated by heating at 200-250 °C to obtain the pretreated product; Step E': Calcine the pretreated product obtained in step D at 900-1100 °C for 1-4 h to obtain Fe-doped H-Nb2O5.

5. The gas sensor array based on Fe-doped bicrystalline Nb₂O₅ according to claim 4, characterized in that, In the Fe-doped H-Nb2O5, the Fe doping molar ratio is 1-10%; In step A, the ratio of ammonium niobate oxalate to water is 0.1-0.5 g : 5-20 mL; In step A, the mass ratio of ferric chloride hexahydrate to ammonium niobate oxalate is 0.001-0.01 : 0.1-0.5; In step E', the calcination heating rate is 1-10 °C / min, and the calcination is carried out in an air atmosphere.

6. A gas sensing system, said gas sensing system comprising a gas sensor array based on Fe-doped bicrystalline Nb₂O₅ as described in any one of claims 1-5, characterized in that, The gas sensing system also includes: The signal acquisition unit is used to acquire the response current signals of the first sensing unit and the second sensing unit and generate the It response curve. The data processing unit employs a two-level cascaded machine learning model; In the two-level cascaded machine learning model, the first level uses a first machine learning algorithm to qualitatively identify gas types, and the second level calls the corresponding second machine learning algorithm to quantitatively predict concentration based on the qualitative identification results of gas types, correcting cross-interference.

7. The gas sensing system according to claim 6, characterized in that, The signal acquisition unit is used to acquire the response current signals of the first sensing unit and the second sensing unit under a complex flue gas atmosphere from a fixed source, and generate the It response curve. The moisture content of the complex flue gas atmosphere from the stationary source is 5-20 vol.%; The stationary source complex flue gas atmosphere includes acetone, hydrogen sulfide, and interfering gases; The interfering gases in the complex flue gas atmosphere of the stationary source include CO2, CO, and SO2; The first machine learning algorithm is the XGBoost model; The second type of machine learning algorithm is the random forest model.

8. The gas sensing system according to claim 6, characterized in that, The gas sensing system also includes: The constant temperature chamber is used to stably control the operating temperature of the first and second sensing units at 300-400 ℃.

9. The gas sensing system according to claim 6 or 7, characterized in that, The data processing unit is further configured to: extract the peak response ratio and recovery time ratio based on the It response curves of the first sensing unit and the second sensing unit; when the peak response ratio and / or recovery time ratio deviate from the preset proportion range corresponding to the known gas in the fixed source complex flue gas atmosphere, it is determined that there is an unknown interference component, and the peak response ratio, recovery time ratio and the time series features of the It response curve are jointly input into the first-level machine learning model to identify the type of unknown interference component.

10. An application of the gas sensing system as described in any one of claims 6-9, characterized in that, The gas sensing system is used for the synergistic detection of carbon pollution in complex flue gas environments with fixed sources. The atmosphere of the stationary source complex flue gas environment is a stationary source complex flue gas atmosphere.

Images