TMDCs parameter optimization method for detecting dissolved gas in oil

Abstract

The invention provides a TMDCs parameter optimization method for detecting dissolved gas in oil, which belongs to the technical field of transformer oil detection and comprises the following steps: firstly, constructing a multi-dimensional parameter matrix containing factors such as doping concentration, energy band structure, carrier concentration, surface active site density and working temperature; a charge transfer matrix of a TMDCs material and gas molecules is obtained by adopting a gas molecule adsorption experiment, and main influence factors are extracted by utilizing a singular value decomposition method. Dimensionality reduction optimization is carried out on the parameter matrix based on the influence factors, and a response function matrix is established; key structure sensitive factors are extracted through principal component analysis, an optimized subspace is established, and an optimal solution is searched by adopting a gradient descent method. Finally, an optimization result is fed back to an original parameter space, an optimal TMDCs material parameter combination including doping concentration, an energy band structure, carrier concentration, surface active site density and working temperature is obtained, and the result is verified through a gas chromatography experiment.

Description

Technical Field

[0001] The present invention belongs to the technical field of transformer oil detection, and in particular, relates to a TMDCs parameter optimization method for detecting dissolved gas in oil. Background Art

[0002] The detection of dissolved gas in transformer oil is an important means to evaluate the operating status of the transformer and predict potential faults. Traditional gas detection methods mainly include gas chromatography and infrared spectroscopy. Although these methods have high detection accuracy, the equipment is expensive, the operation is complicated, and it is difficult to achieve online real-time monitoring. In recent years, gas sensors based on two-dimensional transition metal chalcogenides have attracted widespread attention due to their unique electronic structure and excellent gas-sensitive properties. TMDCs materials have a layered structure, the layers are bonded by van der Waals forces, and the thickness of the single layer is only nanometers. They have a large specific surface area and abundant surface active sites, which are conducive to the adsorption and charge transfer of gas molecules. At present, TMDCs-based gas sensors have shown broad application prospects in environmental monitoring, industrial safety, and medical diagnosis. However, the gas-sensing properties of TMDCs materials are heavily dependent on their structural parameters, including doping concentration, band structure, carrier concentration, surface active site density, and operating temperature. Existing optimization methods mainly rely on empirical trial and error, lack of systematic theoretical guidance, resulting in a time-consuming and inefficient optimization process, and it is difficult to obtain the optimal parameter combination. Summary of the invention

[0003] In view of this, the present invention provides a TMDCs parameter optimization method for detecting dissolved gas in oil, which can solve the technical problems that the prior art mainly relies on empirical trial and error and lacks systematic theoretical guidance, resulting in a time-consuming and inefficient optimization process and difficulty in obtaining the optimal parameter combination.

[0004] The present invention is achieved in that:

[0005] The present invention provides a TMDCs parameter optimization method for detecting dissolved gas in oil, comprising the following steps:

[0006] S10, constructing multiple sets of TMDCs material characteristic parameter matrices, including n-dimensional matrices consisting of doping concentration, band structure, carrier concentration, surface active site density, and operating temperature;

[0007] S20, using a gas molecule adsorption device to collect and calculate the charge transfer matrix of the interaction between the gas molecules and the TMDCs material;

[0008] S30, using a singular value decomposition method to perform dimensionality reduction processing on the charge transfer matrix to obtain a weight matrix of main influencing factors;

[0009] S40, performing dimensionality reduction transformation on the characteristic parameter matrix of the TMDCs material according to the weight matrix to obtain an optimization target matrix;

[0010] S50, establishing a mapping relationship between the optimization target matrix and the gas-sensitive response value, and constructing a response function matrix;

[0011] S60, processing the response function matrix using a principal component analysis method to extract TMDCs material structure sensitivity factors;

[0012] S70, establishing a TMDCs material parameter optimization subspace based on the structure sensitive factor, and using a gradient descent method to search for an optimal solution;

[0013] S80, mapping the optimal solution back to the original parameter space through inverse matrix transformation to obtain an optimized parameter combination of TMDCs materials;

[0014] S90. Use a gas chromatograph to verify the gas-sensitive response value under the optimized parameter combination, and feed back the verification result to the weight matrix, and iteratively execute steps S40-S90 for optimization. After the iteration is completed, the final optimal parameter combination of TMDCs materials is obtained, including doping concentration, band structure, carrier concentration, surface active site density, and operating temperature.

[0015] Furthermore, in step S10, multiple sets of TMDCs material characteristic parameter matrices are constructed. The doping concentration and band structure are known parameters, and the doping concentration is fixed at 5×10 18 cm ；3 The band structure is a direct bandgap semiconductor with a bandgap width of 1.5 eV. The carrier concentration is determined by the Hall effect method with a measurement range of 10 12 -10 18 cm ；3 The surface active site density was determined by programmed temperature desorption method, with a temperature range of 25-800°C, a heating rate of 10°C / min, and an operating temperature range of 25-200°C. After standardizing all parameters, the characteristic parameter matrix X was constructed:

[0016]

[0017] Among them, x i1 represents the doping concentration of the i-th group of samples, x i2 represents the band structure parameters, x i3 represents the carrier concentration, x i4 represents the surface active site density, x i5 represents the working temperature, and m is the number of sample groups. The standardization process uses the minimum and maximum normalization method:

[0018]

[0019] where x j,max and x j,min are the maximum and minimum values ​​of the jth feature parameter respectively. The standardized feature parameter matrix is ​​recorded as X ′ The purpose of this step is to unify the parameters of different dimensions to the same scale to facilitate subsequent analysis and processing.

[0020] In step S20, a gas molecule adsorption device is used to collect and calculate the charge transfer matrix of the interaction between gas molecules and TMDCs materials. The surface potential change of the sample is measured by a circular electrostatic probe array to construct the charge transfer matrix T:

[0021]

[0022] where t ij It represents the charge transfer between the i-th sample and the j-th gas molecule, and the calculation formula is:

[0023] t ij =C i ΔV ij ;

[0024] In the formula, C i is the equivalent capacitance of the i-th sample, ΔV ij is the potential change of the i-th sample exposed to the j-th gas. The potential change is obtained through time series measurement:

[0025] ΔV ij (t) = V ij (t)-V ij (0);

[0026] Where t∈[0,300s] and the sampling interval is 0.01s.

[0027] In step S30, the charge transfer matrix is ​​subjected to dimensionality reduction processing using a singular value decomposition method. The singular value decomposition of the charge transfer matrix T is expressed as:

[0028] T=U∑V T ;

[0029] in:

[0030]

[0031] Select the first k singular values ​​whose cumulative contribution rate exceeds 95%:

[0032]

[0033] Construct the truncated singular value matrix ∑ k And the weight matrix W = U k ∑k .

[0034] In step S40, the TMDCs material characteristic parameter matrix is ​​transformed into a dimension reduction matrix according to the weight matrix. The calculation formula of the optimized target matrix Y is:

[0035] Y=X ′ W;

[0036] It is expressed as:

[0037]

[0038] In step S50, a mapping relationship between the optimization target matrix and the gas sensitive response value is established. The gas sensitive response value is defined as:

[0039]

[0040] Construct the response function matrix:

[0041]

[0042] Build a polynomial regression model:

[0043]

[0044] where a ijl and b ij is the regression coefficient.

[0045] In step S60, the response function matrix is ​​processed using the principal component analysis method. The covariance matrix is ​​calculated:

[0046]

[0047] Solve the characteristic equation:

[0048] (C-λI)v=0;

[0049] Get the eigenvalue λ i and the corresponding eigenvector v i , construct the structure sensitive factor matrix:

[0050] S=[v1,v2,…,v r ];

[0051] Where r satisfies:

[0052]

[0053] In step S70, the objective function is calculated in the optimization subspace:

[0054]

[0055] The objective function is defined as:

[0056]

[0057] where w j is the weight coefficient, is the expected response value. The gradient descent method is used to solve:

[0058]

[0059] Where η is the learning rate, which ranges from 0.01 to 0.1.

[0060] In step S80, the parameters are restored by inverse matrix transformation:

[0061] Y * =Z * S T ;

[0062] X * =Y * (W T W) ；1 W T ;

[0063] Parameter restoration formula:

[0064]

[0065] In step S90, the optimization result is verified. The relative error is calculated:

[0066]

[0067] When the maximum relative error max(e ij )>10%, the training set is updated and the optimization process is repeated. The optimal parameter combination of TMDCs materials is finally obtained: the carrier concentration is 10 16 -10 17 cm ；3 In the range of 10 13 -10 14 cm ；2 The operating temperature is within the range of 100-150°C.

[0068] Furthermore, the stopping condition of the iteration is that the cosine similarity of the vector formed by the material optimization parameter combination obtained after 20 consecutive iterations or two adjacent iterations is greater than 0.97.

[0069] Compared with the prior art, the beneficial effects of the TMDCs parameter optimization method for detecting dissolved gas in oil provided by the present invention are as follows: first, the material characteristic parameters and gas-sensitive response are described in matrix form, and a mathematical model of the structure-performance relationship is established, so that the optimization process has theoretical guiding significance. Secondly, the charge transfer matrix is ​​reduced in dimension by singular value decomposition, the main influencing factors of the material's gas-sensitive properties are extracted, the dimension of the optimization space is reduced, and the calculation efficiency is improved. Third, the principal component analysis method is used to identify the structural sensitive factors, the coupling effect between the parameters is considered, and the local optimal solution is avoided. Fourth, the gradient descent method is used to search for the optimal solution in the optimization subspace, and the actual parameters are restored by inverse matrix transformation, which ensures the feasibility of the optimization result. Fifth, an iterative verification mechanism is introduced, the optimization results are fed back through gas chromatography testing, the training data is continuously updated, and the optimization accuracy is improved. It solves the technical problems that the prior art mainly relies on empirical trial and error, lacks systematic theoretical guidance, and leads to a long optimization process, low efficiency, and difficulty in obtaining the optimal parameter combination. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A flow chart of the method provided by the present invention;

[0071] Figure 2 It is a schematic diagram of the structure of the gas adsorption experimental device;

[0072] Figure 3 It is a schematic diagram of the first and second flange connection structure of the gas adsorption experimental device;

[0073] Figure 4 is a graph showing the relationship between the charge transfer amount, temperature and doping concentration in Example 2;

[0074] Figure 5 The principal component analysis results in Example 2;

[0075] Figure 6 This is a diagram showing the change in the optimization iterative process in Example 2;

[0076] Figure 7 is a graph showing the relationship between the sensitivity and selectivity of different gases in Example 2;

[0077] Figure 8It is a comparison chart of the response characteristics of the optimized MoS2 sensor to different gases in Example 2; in the accompanying drawings, the reference numerals represent: 10, vacuum chamber; 11, exhaust port; 12, vacuum system; 13, observation window; 101, first flange; 102, second flange; 103, third flange; 21, sample fixing table; 211, lifting mechanism; 212, rotating mechanism; 213, groove; 31, gas storage bottle; 32, flow control valve; 33, pressure regulating valve; 34, injection head; 41, heating wire; 42, radiator; 43, temperature sensor. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0079] like Figure 1 FIG. 1 is a flow chart of a TMDCs parameter optimization method for detecting dissolved gas in oil provided by the present invention. The method comprises the following steps:

[0080] S10, constructing multiple sets of TMDCs material characteristic parameter matrices, including n-dimensional matrices consisting of doping concentration, band structure, carrier concentration, surface active site density, and operating temperature;

[0081] S20, using a gas molecule adsorption device to collect and calculate the charge transfer matrix of the interaction between the gas molecules and the TMDCs material;

[0082] S30, using a singular value decomposition method to perform dimensionality reduction processing on the charge transfer matrix to obtain a weight matrix of main influencing factors;

[0083] S40, performing dimension reduction transformation on the characteristic parameter matrix of the TMDCs material according to the weight matrix to obtain an optimization target matrix;

[0084] S50, establishing a mapping relationship between the optimization target matrix and the gas-sensitive response value, and constructing a response function matrix;

[0085] S60, using principal component analysis method to process the response function matrix and extract the structural sensitivity factors of TMDCs materials;

[0086] S70. Establish the TMDCs material parameter optimization subspace based on the structure sensitive factor and use the gradient descent method to search for the optimal solution;

[0087] S80, mapping the optimal solution back to the original parameter space through inverse matrix transformation to obtain the optimized parameter combination of TMDCs materials;

[0088] S90. Use a gas chromatograph to verify the gas-sensitive response value under the optimized parameter combination, and feed the verification result back to the weight matrix, and iteratively execute steps S40-S90 for optimization. After the iteration is completed, the final optimal parameter combination of TMDCs materials is obtained, including doping concentration, band structure, carrier concentration, surface active site density, and operating temperature.

[0089] The specific implementation methods of the above steps are described in detail below:

[0090] The specific implementation of step S10 is to construct multiple sets of TMDCs material characteristic parameter matrices. First, the TMDCs material is prepared by chemical vapor deposition. During the preparation process, the doping concentration of the material is regulated by controlling parameters such as precursor concentration, growth temperature and growth time. The doping concentration range is set to 1×10 17 -1×10 19 cm ；3 The doping concentration was determined by secondary ion mass spectrometry, with oxygen ions as the incident ion, energy of 5 keV, ion beam intensity of 50 nA, and a scanning range of 100 × 100 μm. 2 , with a step size of 0.1 μm. The band structure was characterized by angle-resolved photoelectron spectroscopy. The excitation light source was HeⅠα rays, the photon energy was 21.2 eV, the energy resolution was better than 10 meV, the angle resolution was 0.1°, the measurement temperature was 300 K, and the vacuum degree was better than 5×10 ；10 The Hall effect method is used to determine the carrier concentration. The specific steps include: preparing the sample into a van der Pauw Hall bar structure with a size of 10 mm × 2 mm; measuring the resistivity of the sample using a four-probe method; applying a uniform magnetic field perpendicular to the sample surface with a magnetic field strength of 0.5 T; measuring the Hall voltage, and using the formula Calculate the carrier concentration, where I is the current passing through the sample, B is the magnetic field strength, q is the electron charge, and V H is the Hall voltage, and d is the sample thickness. The surface active site density was determined by a programmed temperature desorption experiment. The specific steps are as follows: the sample was loaded into a quartz reaction tube; the vacuum was evacuated to 1×10 ； 6Pa; introduce nitrogen carrier gas with a flow rate of 30mL / min; heat from room temperature to 800℃ at a rate of 10℃ / min; use a mass spectrometer to monitor the type and concentration of desorbed gas in real time; calculate the surface active site density based on the desorption peak area. The operating temperature range is determined by thermogravimetric analysis, the heating rate is 5℃ / min, and the test atmosphere is nitrogen. All parameters are standardized using the minimum and maximum normalization method to ensure the comparability of parameters of different dimensions. The constructed characteristic parameter matrix contains five dimensions: doping concentration, band structure, carrier concentration, surface active site density, and operating temperature. The matrix elements are all standardized values. The purpose of this step is to establish a mathematical description of the characteristic parameters of the material and provide basic data for subsequent optimization.

[0091] The specific implementation of step S20 is to use a gas molecule adsorption experiment to measure and calculate the interaction between the gas molecules and the TMDCs material to obtain a charge transfer matrix. In the prior art, the following four methods are commonly used:

[0092] 1. In-situ X-ray Photoelectron Spectroscopy (XPS) Method The charge transfer behavior between gas molecules and TMDCs materials can be studied by using an in-situ XPS device equipped with a differential pumping system. The core components of this method include: a high-pressure gas supply chamber (pressure can reach 10 atmospheres), a differential gas pressure control system, a photoelectron spectroscopy analysis chamber, and a data acquisition system. During the experiment, the TMDCs sample is placed on a special sample stage in the analysis chamber, and the introduction of gas molecules is controlled by the differential system, while maintaining an ultra-high vacuum environment in the analysis chamber (better than 10 ；9 By continuously monitoring the core energy level peak shifts of transition metal elements and sulfur elements in TMDCs materials, as well as the fine structural changes in the valence band region, the amount of charge transfer before and after gas molecules are adsorbed can be quantitatively obtained. The advantage of this method is that it can simultaneously obtain chemical state information and electronic structure changes, especially through angle-resolved XPS, it can also study the difference in charge transfer between the surface and near-surface regions.

[0093] 2. Kelvin probe force microscopy (KPFM) technology uses an ultra-high vacuum KPFM system to study the changes in contact potential difference caused by gas molecule adsorption at the atomic scale. The system is mainly composed of: an ultra-high vacuum SPM host, a gas molecule pulse jet, a temperature control system, and an in-situ annealing device. During the experiment, the morphology and contact potential difference of the surface of the TMDCs material are first obtained in an ultra-high vacuum, and then a certain amount of gas molecules are introduced through a pulse valve to monitor the evolution of the surface potential in real time. By establishing a quantitative relationship between the contact potential difference and the charge transfer amount, a charge transfer matrix with high spatial resolution can be constructed. This method is characterized by nanoscale spatial resolution and the ability to distinguish adsorption behaviors on different crystal planes and defect locations.

[0094] 3. Synchrotron radiation soft X-ray absorption spectroscopy (XAS) technology uses the in-situ XAS experimental station of the synchrotron radiation device to study the electronic structure evolution of TMDCs materials during the adsorption of gas molecules. This technology mainly relies on: high-throughput synchrotron radiation light source, gas mixing control system, differential vacuum system and fluorescence yield detector and other components. By adjusting the X-ray energy to the L edge of the transition metal and the K edge of sulfur in the TMDCs material and measuring the near-edge X-ray absorption fine structure (NEXAFS) spectrum, element-specific electronic state density information can be obtained. Combined with density functional theory simulation, the charge redistribution process caused by gas molecule adsorption can be quantitatively described. The advantage of this method is that it has element selectivity and can study the electronic structure changes of transition metals and sulfur atoms during the adsorption process separately.

[0095] 4. In-situ Raman-electrochemical coupling technology The in-situ Raman-electrochemical coupling system can simultaneously obtain molecular vibration information and charge transfer dynamics. The system mainly includes: grating spectrometer, electrochemical workstation, gas flow control device and constant temperature control system. By designing a special spectral electrolytic cell, gas molecules are introduced while controlling the electrode potential to record the displacement and intensity changes of the characteristic Raman peaks of TMDCs materials. Combined with the charge amount measured by cyclic voltammetry, a quantitative relationship between Raman spectral parameters and charge transfer amount can be established. The characteristic of this method is that it can study the gas adsorption behavior under electric field regulation and obtain electrochemically driven charge transfer information.

[0096] In addition, the following method can also be used to achieve the first assembly of a gas molecule adsorption experimental device, including: a vacuum chamber 10, a sample table assembly, a gas introduction assembly, a temperature adjustment assembly, and a pressure monitoring assembly; the vacuum chamber is a cylindrical stainless steel structure, the inner surface is polished, and the surface roughness Ra ≤ 0.4μm; the top of the vacuum chamber is provided with a sealable and openable observation window 13, and the observation window is made of pressure-resistant glass; the bottom of the vacuum chamber is provided with an exhaust port 11 connected to an external vacuum system 12; three flange interfaces are evenly arranged on the side wall of the vacuum chamber for connecting various functional components; the sample table assembly is installed at the center of the vacuum chamber, including a sample fixing table 21, a lifting mechanism 211 and a rotating mechanism 212; the sample fixing table is made of high thermal conductivity copper material, the surface is chrome-plated, and a groove for fixing the TMDCs material is provided in the central position, and the groove depth is 2mm; the lifting mechanism is composed of a precision screw pair, which can realize the precise vertical displacement of the sample table; the rotating mechanism adopts a worm gear transmission structure, which can make the sample table rotate 360 ​​degrees in the horizontal plane; the gas introduction assembly The components are installed at the first flange 101 interface of the vacuum chamber, including a gas storage bottle 31, a flow control valve 32, a pressure regulating valve 33 and an injection head 34; the gas storage bottle is provided with an explosion-proof protective cover on the outside and a buffer chamber on the inside; the flow control valve adopts a needle valve structure, which can achieve precise adjustment of 0-100mL / min; the pressure regulating valve adopts a diaphragm structure to ensure the stability of the gas injection pressure; the injection head has a conical structure, the nozzle diameter is 0.5mm, and the distance between the nozzle and the sample stage is adjustable; the temperature adjustment component is installed on the second flange 102 The interface includes a heating wire 41, a temperature sensor 43 and a radiator 42; the heating wire is made of nickel-chromium alloy material and is spirally wound around the bottom of the sample stage; the temperature sensor is a K-type thermocouple, and the probe directly contacts the sample stage; the radiator adopts a fin-type structure and fits tightly with the sample stage; the pressure monitoring component is installed at the interface of the third flange 103, including a vacuum gauge, a pressure transmitter and a sealing joint; the vacuum gauge adopts an ionization gauge structure with a measurement range of 10-7 to 10-2Pa; the pressure transmitter adopts a capacitive structure and has a temperature compensation function. A circular electrostatic probe array is installed at the bottom of the sample stage, consisting of 12 equally spaced micro-electrostatic probes; each probe is made of a platinum wire with a diameter of 0.1mm, and the probe tip is electrolytically polished; the distance between the probe and the sample surface can be accurately adjusted within the range of 0.1-1.0mm; a Faraday shield is installed outside the vacuum chamber to reduce environmental electromagnetic interference. It also includes a low-energy electron diffraction (LEED) window set on the side wall of the vacuum chamber. The sample stage can be moved to the position directly below the LEED window. The electron beam energy can be adjusted in the range of 20-200eV. The surface structure changes are analyzed through the diffraction pattern, and the observation window is used for real-time Raman spectroscopy measurement. The laser wavelength is 532nm and the power density does not exceed 1mW / μm 2 , the spot size is about 1μm.

[0097] Afterwards, the gas molecule adsorption experiment was carried out using a gas molecule adsorption experimental device: first, the TMDCs material was fixed on the sample stage, and the sample position was adjusted by the lifting mechanism to keep a distance of 0.5 mm from the annular electrostatic probe array. ；7 Pa, the gas to be tested is introduced, and the gas flow rate is controlled to be 50mL / min. Start the electrostatic probe array, the sampling frequency is 100Hz, and the measurement range is 300s. The calculation of the charge transfer amount is based on the principle of capacitor charging and discharging. The equivalent capacitance of the sample is measured by an impedance analyzer. The measurement frequency range is 20Hz-1MHz, and the excitation signal amplitude is 10mV. The potential change is obtained by time series measurement, and a 12-bit analog-to-digital converter is used for data acquisition with a sampling rate of 100kS / s. After the raw data is median filtered to remove noise, the charge transfer matrix is ​​calculated. The purpose of this step is to obtain a quantitative description of the interaction between the material and the gas molecules.

[0098] The specific implementation method of step S30 is to use the singular value decomposition method to reduce the dimension of the charge transfer matrix. First, the charge transfer matrix is ​​decomposed into the product of the left singular matrix, the singular value matrix and the right singular matrix, wherein the left singular matrix and the right singular matrix are both orthogonal matrices, reflecting the correlation between the sample and the gas; the singular value matrix is ​​a diagonal matrix, and the diagonal elements are arranged in descending order, indicating the importance of each principal component. By calculating the cumulative contribution rate of the singular values, the singular values ​​with a cumulative contribution rate exceeding 95% are selected to construct a truncated singular value matrix. The left singular matrix is ​​multiplied by the truncated singular value matrix to obtain a weight matrix, which reflects the main characteristics of the response of different samples to the gas. The purpose of this step is to reduce the data dimension and extract key feature information.

[0099] The specific implementation method of step S40 is to perform a dimensionality reduction transformation on the characteristic parameter matrix of the TMDCs material according to the weight matrix. First, the standardized characteristic parameter matrix is ​​multiplied by the weight matrix to obtain the optimization target matrix. Each row of the optimization target matrix corresponds to a sample, each column corresponds to a principal component, and the matrix elements represent the projection coordinates of the sample in the principal component space. The dimensionality reduction transformation maintains the relative relationship between samples while reducing redundant information. The purpose of this step is to map the high-dimensional feature space to the low-dimensional optimization space.

[0100] The specific implementation method of step S50 is to establish a mapping relationship between the optimization target matrix and the gas-sensitive response value. The gas-sensitive response value is defined as the relative rate of change of the resistance value of the sample in the target gas relative to the resistance value in the air. The resistance value is measured using a four-probe method, with a probe spacing of 1 mm and a measuring current of 1 mA. In the constructed response function matrix, each element represents the response value of a specific sample to a specific gas. A cubic polynomial regression is used to establish a mapping relationship between the principal component and the response value, and the regression coefficient is solved by the least squares method. The purpose of this step is to establish a quantitative relationship between material structure and performance.

[0101] The specific implementation method of step S60 is to process the response function matrix using the principal component analysis method. First, the covariance matrix of the response function matrix is ​​calculated. The covariance matrix reflects the correlation between different gas responses. Solve the characteristic equation of the covariance matrix to obtain the eigenvalue and the corresponding eigenvector. The eigenvalue represents the variance contribution of the principal component, and the eigenvector represents the direction of the principal component. Select eigenvectors with a cumulative variance contribution rate of more than 90% to construct a structural sensitivity factor matrix. The purpose of this step is to identify the key structural factors that affect the gas-sensitive response.

[0102] The specific implementation of step S70 is to search for the optimal solution in the optimization subspace. First, the optimization target matrix is ​​projected into the subspace determined by the structural sensitive factor matrix to obtain the representation of the sample in the sensitive factor space. The objective function is defined as the weighted square error sum of the actual response value and the expected response value, and the weight coefficient is determined according to the importance of different gases. The gradient descent method is used for optimization and solution, and the learning rate is dynamically adjusted according to the change of the objective function value, and the initial value is set to 0.05. When the objective function value changes less than 1×10 ；6 The purpose of this step is to find the optimal parameter combination in the dimensionality reduction space.

[0103] The specific implementation of step S80 is to map the optimization results back to the original parameter space through matrix inverse transformation. First, the optimal solution in the sensitive factor space is transformed to the principal component space through the transposition of the structural sensitive factor matrix, and then the result is transformed to the original feature space using the generalized inverse matrix of the weight matrix. The standardized parameters obtained by the transformation are denormalized to obtain the actual material parameter values. The purpose of this step is to convert the mathematical optimization results into achievable material parameters.

[0104] The specific implementation of step S90 is to verify the validity of the optimized parameters. Prepare TMDCs material samples according to the optimized parameters, and measure their gas-sensitive response characteristics using a gas chromatograph. The working conditions of the gas chromatograph are: the carrier gas is high-purity nitrogen with a flow rate of 30mL / min; the injection volume is 1mL; the column temperature is 40°C; and the detector temperature is 200°C. Calculate the relative error between the measured response value and the predicted response value. When the maximum relative error exceeds 10%, add the group of data to the training set and repeat the optimization process. After multiple iterative optimizations, the optimal parameter combination of the TMDCs material is finally determined.

[0105] Specifically, the principle of the present invention is: based on the electronic structure characteristics of TMDCs materials and the gas molecule adsorption mechanism. TMDCs materials are sandwich structures formed by covalent bonds between transition metal atoms and chalcogen atoms. The atomic bonding within the layer is strong and the interlayer van der Waals force is weak, which makes the material have a unique electronic band structure. When gas molecules are adsorbed on the surface of TMDCs, charge transfer occurs, resulting in changes in the conductivity of the material. This change is closely related to the intrinsic properties of the material. The doping concentration affects the position of the Fermi level and directly determines the direction and size of the charge transfer. The band structure, especially the band gap width, affects the thermal excitation process of the carrier and has an important influence on the operating temperature range of the material. The carrier concentration determines the intrinsic conductivity of the material and affects the baseline and noise level of the signal. The surface active site density determines the adsorption capacity and adsorption strength of the gas molecules, which directly affects the sensitivity of the sensor. There are complex interactions between these parameters, which jointly determine the gas-sensitive properties of the material. The present invention uses matrix algebra methods to describe this multi-parameter coupling relationship. By constructing a characteristic parameter matrix, discrete experimental data is converted into a continuous mathematical model. The physical meaning of singular value decomposition is to decompose the complex gas adsorption process into several independent basic modes, and each singular value represents the importance of the corresponding mode. Principal component analysis further reveals the intrinsic relationship between material structure and performance, and identifies the key factors that play a decisive role in gas-sensitive response. In the optimization process, the gradient descent method is used to search for the optimal solution in the reduced dimensional space. Its physical meaning is to adjust the parameters in the direction of the fastest performance improvement. This solution, which combines mathematical processing methods with material science principles, not only improves the optimization efficiency, but also ensures the physical rationality of the optimization results. The optimization results are verified by gas chromatograph and iteratively updated to further ensure the reliability and practicality of the method.

[0106] A specific embodiment 1 of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows: The specific implementation of step S10 is to construct multiple sets of TMDCs material characteristic parameter matrices. This step establishes a standardized characteristic parameter matrix by systematically measuring and collecting the key parameters of the TMDCs material to provide basic data support for subsequent optimization. First, the doping concentration of the TMDCs material is determined. In this embodiment, the thermal diffusion method is used for doping. The doping source is molybdenum sulfide powder. The doping temperature is controlled in the range of 800 to 1000°C, the doping time is 2 to 4 hours, and the doping concentration is determined by secondary ion mass spectrometry (SIMS). The measurement depth distribution range is 0 to 500nm, the resolution is better than 10nm, and the doping concentration is controlled at 5×10 18 cm ；3 The band structure was characterized by angle-resolved photoelectron spectroscopy (ARPES). The excitation light source was a helium discharge lamp with a photon energy of 21.2 eV, an energy resolution better than 10 meV, and an angle resolution better than 0.2 degrees. The sample was measured to be a direct bandgap semiconductor with a bandgap width of 1.5 eV. The carrier concentration was determined by the Hall effect method. The measurement system adopted the Van der Pauw configuration. The magnetic field strength was 0.5 T, the measurement current range was 1 μA to 1 mA, the temperature range was 25 to 300 K, and the measurable carrier concentration range was 10 12 Up to 10 18 cm ；3 The surface active site density was determined by programmed temperature desorption. The experiment was carried out in an ultra-high vacuum chamber with a background vacuum better than 10 ；8 Pa, desorption gas analysis was performed using a mass spectrometer with a mass range of 1 to 200 amu and a minimum detectable partial pressure of 10 ；11 Pa, the temperature range is 25 to 800 °C, and the heating rate is 10 °C / min. The operating temperature range is achieved by a program temperature control system, the temperature range is 25 to 200 °C, and the temperature control accuracy is better than ±0.1 °C. All measured data are standardized to construct the characteristic parameter matrix X, and the matrix elements x ij They represent the jth characteristic parameter of the i-th group of samples, where i ranges from 1 to m, indicating the number of sample groups, and j ranges from 1 to 5, corresponding to doping concentration, band structure, carrier concentration, surface active site density, and operating temperature, respectively. The minimum and maximum normalization method is used for standardization, and the standardization formula is: where x j,max and x j,min are the maximum and minimum values ​​of the jth feature parameter respectively. The standardized feature parameter matrix is ​​recorded as X ′ .

[0107] The specific implementation of step S20 is to use a gas molecule adsorption experimental device to collect and calculate the charge transfer matrix of the interaction between gas molecules and TMDCs materials. This step obtains the charge transfer characteristics between gas molecules and materials by accurately measuring the dynamic changes of the material surface potential during the gas molecule adsorption process. First, fix the TMDCs material sample on the sample table and use double-sided conductive tape to achieve reliable electrical contact. Vacuum to 10 ；6 Below Pa, start the temperature control system to adjust the sample temperature to the preset working temperature. The annular electrostatic probe array consists of 12 equally spaced micro electrostatic probes. The probes use platinum wire with a diameter of 0.1 mm. The probe tips are electrolytically polished and have a curvature radius of less than 1 μm. The distance between the probe and the sample surface is adjusted by a precision translation stage, which can be continuously adjusted in the range of 0.1 to 1.0 mm, and the positioning accuracy is better than 1 μm. The introduction of gas molecules adopts a precision flow control system with a flow range of 0 to 100 mL / min and a control accuracy better than 0.1 mL / min. The change of sample surface potential over time is measured by the annular electrostatic probe array. The sampling time is 300 s and the sampling interval is 0.01 s. The change in potential is calculated by ΔV ij (t) = V ij (t)-V ij (0) obtain, where V ij (t) represents the surface potential of the i-th sample when it is exposed to the j-th gas at time t, V ij (0) is the initial potential value. The charge transfer amount is calculated using the equivalent capacitance model, t ij =C i ΔV ij , where C i is the equivalent capacitance of the i-th sample, measured by the AC impedance method. Finally, the charge transfer matrix T is constructed, and the matrix element t ij Represents the amount of charge transfer between the i-th sample and the j-th gas molecule.

[0108] The specific implementation of step S30 is to use the singular value decomposition method to reduce the dimension of the charge transfer matrix. This step aims to extract the main influencing factors in the charge transfer process and reduce data redundancy. Singular value decomposition is an important matrix decomposition method that can decompose any matrix into the product of three matrices: T = U∑V T . Where U is an m×m dimensional orthogonal matrix, representing the left singular vector; ∑ is an m×p dimensional diagonal matrix, the diagonal elements are singular values, and are arranged in descending order; V is a p×p dimensional orthogonal matrix, representing the right singular vector. The size of the singular value reflects the importance of the corresponding eigenvector in the original data. By selecting the first k singular values ​​with a cumulative contribution rate exceeding 95%, data dimensionality reduction can be achieved while retaining the main information. The calculation formula for the cumulative contribution rate is where σi is the i-th singular value. Construct the truncated singular value matrix ∑ according to the selected k singular values k , and calculate the weight matrix W = U k ∑ k , where U k are the first k columns of U. According to experience, the default k=12.

[0109] The specific implementation of step S40 is to perform dimensionality reduction conversion on the TMDCs material characteristic parameter matrix according to the weight matrix. This step maps the high-dimensional feature space to the low-dimensional optimization space, simplifying the subsequent optimization process. The calculation of the optimization target matrix Y adopts matrix multiplication: Y = X ′ W, where X ′ is the normalized feature parameter matrix, and W is the weight matrix obtained in step S30. The dimension of the optimization target matrix Y is m×k, and each row represents the coordinate of a sample in the space after dimensionality reduction.

[0110] The specific implementation of step S50 is to establish a mapping relationship between the optimization target matrix and the gas-sensitive response value. This step realizes the quantitative association between material parameters and performance indicators by constructing a response function model. The definition of the gas-sensitive response value adopts the resistance change rate: Where R g,ij is the resistance value of the i-th sample in the j-th gas, R a,i is the resistance value in air. The element f in the response function matrix F ij Represents the response function of the i-th sample to the j-th gas. The mapping relationship is established using a cubic polynomial regression model: where a ijl and b ij are regression coefficients, determined by the least squares method.

[0111] The specific implementation of step S60 is to process the response function matrix using the principal component analysis method. This step extracts the material structure sensitivity factor by analyzing the correlation between the response functions. First, the covariance matrix of the response function matrix is ​​calculated: Solve the characteristic equation (C-λI)v=0 to obtain the eigenvalue λ i and the corresponding eigenvector v i The eigenvalue reflects the variance of the data in the direction of the corresponding eigenvector. The eigenvectors with a cumulative contribution rate exceeding 90% are selected to construct the structural sensitivity factor matrix S = [v1, v2, …, v r ], where r satisfies

[0112] The specific implementation of step S70 is to establish a TMDCs material parameter optimization subspace based on the structural sensitivity factor. This step searches for the optimal solution in the optimization space after dimensionality reduction to improve the optimization efficiency. First, calculate the coordinate matrix in the optimization subspace: Z = YS. Define the optimization objective function: where w j is the weight coefficient, which indicates the importance of different gas responses. is the expected response value. The gradient descent method is used for optimization and solution, and the iterative formula is: Where η is the learning rate, which ranges from 0.01 to 0.1. The objective function value is calculated in real time during the iteration process. When the change in the objective function value between two consecutive iterations is less than a preset threshold (such as 10 ；6 ) stops iteration.

[0113] The specific implementation of step S80 is to map the optimal solution back to the original parameter space through matrix inverse transformation. This step converts the optimization result into practical and operable material parameters. First, the optimal solution Z in the optimization subspace is mapped back to the original parameter space through matrix inverse transformation. * By transposing the transformation, we can get Y * =Z * S T , and then use the generalized inverse matrix to calculate X * =Y * (W T W) ；1 W T . Finally, the actual parameter value is restored through denormalization:

[0114]

[0115] The specific implementation method of step S90 is to use a gas chromatograph to verify the gas-sensitive response value under the optimized parameter combination, and obtain the final material parameter combination through multiple rounds of iterative optimization. This step first uses a gas chromatograph to systematically test the gas-sensitive properties of the material. The chromatographic column of the gas chromatograph uses an HP-5 capillary column with a column length of 30m, an inner diameter of 0.32mm, and a film thickness of 0.25μm. The injector uses a split / non-split programmable injection port with a split ratio of 50:1 and an injection volume of 1μL. The carrier gas uses high-purity nitrogen with a purity of not less than 99.999%, and the carrier gas flow rate is controlled at 30mL / min in a constant flow mode. The program heating conditions of the chromatographic column are: the initial temperature is 80°C for 2 minutes, then the temperature is increased to 120°C at a heating rate of 5°C / min and maintained for 5 minutes, and finally the temperature is increased to 150°C at a heating rate of 10°C / min and maintained for 3 minutes. The detector uses a hydrogen flame ionization detector (FID), the detector temperature is set to 150°C, the hydrogen flow rate is 30mL / min, the air flow rate is 300mL / min, and the tail gas (nitrogen) flow rate is 25mL / min. The sample gas is prepared using a dynamic gas diluter, which can be continuously adjusted in the range of 0.1 to 1000ppm, and the gas distribution accuracy is better than ±2%. During the test, the ambient temperature is controlled at 25±1°C and the relative humidity is controlled at 45±5%. Each gas concentration point is tested three times, and the average value is taken as the final result. The gas-sensitive response value is calculated using the formula Where R g,ij Use a digital multimeter to measure online, with a sampling frequency of 10Hz. At the same time, record the response time t 90 (defined as the time required for the gas-sensitive response to reach 90% of the steady-state value) and the recovery time t 10 (defined as the time required for the response value to decrease to 10% of the maximum value). To evaluate the repeatability and stability of the material, 10 cycles of gas adsorption and desorption cycle tests were performed on each sample, and the standard deviation of the response value was calculated. The relative error between the experimental value and the theoretical prediction value was calculated by the formula Calculate, where f ij is the experimentally measured response function value, is the model prediction value. When the maximum relative error max(e ij ) exceeds 10%, the iterative optimization process is started. The iterative optimization process includes the following steps: first, the new experimental data is added to the training set to expand the charge transfer matrix T; then the weight matrix W is recalculated and the update formula is W new =(1-α)W old +αW exp , where α is the learning factor, ranging from 0.1 to 0.3, W expis the weight matrix calculated based on the new experimental data; then the optimization target matrix Y is reconstructed based on the updated weight matrix; the improved gradient descent method is used to search for a new optimal solution in the optimization subspace. The improvements include: dynamically adjusting the learning rate η=η0 / (1+βt), where η0 is the initial learning rate, which is 0.1, β is the attenuation coefficient, which is 0.05, and t is the number of iterations; introducing the momentum term to accelerate convergence, the update formula is Where γ is the momentum coefficient, which is 0.9. The early stopping strategy is adopted in the optimization process. When the error on the validation set does not decrease for 5 consecutive iterations, the optimization is stopped. Finally, the new optimal solution is mapped back to the original parameter space through matrix inverse transformation. The iterative optimization process continues until any of the following termination conditions is met: the maximum relative error is less than 10%; the parameter change in 3 consecutive iterations is less than 1%; the preset maximum number of iterations (such as 20) is reached. The optimal parameter combination of TMDCs materials is finally determined as follows: the carrier concentration is between 10 16 Up to 10 17 cm ；3 In the range of 10 13 Up to 10 14 cm ；2 In the range of 100 to 150 ° C, the working temperature is 100 to 150 ° C. Under this parameter combination, the material exhibits excellent gas-sensing performance: the response sensitivity to the target gas is increased by more than 50%, reaching αR / R0>80% (100ppm test gas); the response time t 90 Less than 30s, 40% shorter than before optimization; recovery time t 10 Less than 60s; repeatability error (relative standard deviation) less than 5%; long-term stability test (30 days) shows that the response value drift is less than 8%. In addition, comparative tests show that the material under this parameter combination also has good selectivity, and the cross-response coefficient to interfering gases is less than 0.1.

[0116] In order to further better understand and implement the present invention, the following provides Example 2 of a specific application scenario of the present invention: This Example 2 is aimed at the detection of dissolved gas in 220kV transformer oil, MoS2 is selected as the basic material, and a high-sensitivity gas sensor is designed by the parameter optimization method of the present invention. The specific implementation process is as follows.

[0117] First, step S10 is performed to construct the characteristic parameter matrix of MoS2 material. Platinum (Pt) is selected as the doping element, and the doping concentration range is 0.1% to 5%. It is discretized at intervals of 0.1%, and a total of 50 concentration points are obtained. In terms of energy band structure, the value range of the bottom energy level Ec of the conduction band is -4.2eV to -3.5eV, and the value range of the top energy level Ev of the valence band is -6.2eV to -5.5eV. Both are discretized at intervals of 0.1eV, and 8 and 8 energy level points are obtained respectively. The carrier concentration range is set to 1×1015 to 1×1018cm-3, and 10 points are taken by logarithmic uniform sampling. The surface active site density range is 1×1012 to 1×1015cm-2, and 8 points are also taken by logarithmic uniform sampling. The operating temperature range is 25℃ to 200℃, and it is discretized at intervals of 5℃, with a total of 36 temperature points. The above parameters are combined to construct the characteristic parameter matrix X, with a matrix dimension of 115200 × 7. After minimum and maximum normalization, the condition number of the matrix is ​​521, which meets the requirements of numerical calculation stability. Some parameter data after processing are shown in Table 1.

[0118] Table 1 Normalized MoS2 material characteristic parameter matrix (partial data)

[0119] Sample No. Doping concentration Ec Ev Carrier concentration Active site density Operating temperature 1 0.0204 0.375 0.250 0.167 0.143 0.286 2 0.0408 0.500 0.375 0.333 0.286 0.429 3 0.0612 0.625 0.500 0.500 0.429 0.571 4 0.0816 0.750 0.625 0.667 0.571 0.714 5 0.1020 0.875 0.750 0.833 0.714 0.857

[0120] Then, step S20 is performed to perform first-principles calculations using the VASP software package. A 4×4 MoS2 supercell model is established, and the vacuum layer thickness is set to The optimized lattice constant is H2, CH4, C2H2, C2H4 and C2H6 were selected as target gas molecules, and three adsorption positions, top position, bridge position and vacancy position, were considered. The calculated charge transfer matrix Q dimension is 115200×15, and some data are shown in Table 2.

[0121] Table 2 Charge transfer amount between gas molecules and MoS2 materials (partial data, unit: e)

[0122] Configuration number H2-Top position H2-Bridge Position CH4-top position CH4-Bridge C2H2-top position 1 0.052 0.038 0.086 0.072 0.154 2 0.067 0.045 0.095 0.081 0.168 3 0.078 0.056 0.112 0.094 0.185 4 0.089 0.064 0.128 0.108 0.197 5 0.096 0.075 0.142 0.125 0.212

[0123] Figure 4 This is a three-dimensional surface diagram showing the relationship between charge transfer and temperature and doping concentration. The color gradient allows you to clearly see the optimal working area.

[0124] In step S30, singular value decomposition analysis is performed on the charge transfer matrix. The first 10 singular values ​​and their cumulative contribution rates calculated are shown in Table 3.

[0125] Table 3 Singular value analysis results of charge transfer matrix

[0126] Serial number Singular Values Contribution rate (%) Cumulative contribution rate (%) 1 156.24 45.32 45.32 2 98.76 28.65 73.97 3 45.32 13.15 87.12 4 18.65 5.41 92.53 5 12.43 3.61 96.14

[0127] According to the requirement of 95% cumulative contribution rate, the first four singular values ​​are selected to construct the weight matrix W. After normalization, the stability index of the weight matrix s = 7.8 × 10-4 meets the orthogonality requirement.

[0128] In step S40, the weight matrix is ​​used to reduce the dimension of the feature parameter matrix to obtain the optimization target matrix Y. The dimension of the reduced matrix is ​​115200×4, and the matrix condition number is 235, which is significantly lower than the original feature parameter matrix. The maximum correlation coefficient between adjacent dimensions is 0.26, which proves that the dimensionality reduction result has good independence.

[0129] In step S50, 100 groups of different parameter combinations are selected for gas-sensitive response testing to obtain an experimental data set. A radial basis function neural network is used to construct a mapping model, and the number of hidden layer nodes is set to 8. After 3000 rounds of training, the mean square error on the validation set is reduced to 0.015, and a stable response function matrix F is obtained.

[0130] In step S60, principal component analysis is performed on the response function matrix, and the obtained eigenvalues ​​and their contribution rates are shown in Table 4.

[0131] Table 4. Results of principal component analysis of the response function matrix

[0132] principal component Eigenvalue Contribution rate (%) Cumulative contribution rate (%) PC1 3.856 48.20 48.20 PC2 2.143 26.79 74.99 PC3 1.256 15.70 90.69 PC4 0.452 5.65 96.34

[0133] The first three principal components are selected as structural sensitive factors, corresponding to the coupling effect of doping concentration and band structure, carrier transport characteristics, and surface active site distribution. Figure 5 As shown, the results of principal component analysis are shown, including the contribution rate of each principal component (bar chart) and the cumulative contribution rate (line chart).

[0134] In step S70, a gradient descent search is performed in the optimization subspace. The initial point is randomly selected, and the local optimal solution is obtained after 2500 iterations. In order to avoid falling into the local optimum, multiple random initialization strategies are adopted, and a total of 10 rounds of optimization are performed to finally obtain the global optimal solution, such as Figure 6 shown.

[0135] In step S80, the optimal solution of the optimized subspace is mapped back to the original parameter space. After Tikhonov regularization, the optimal parameter combination is: platinum doping concentration 1.2%, conduction band bottom energy level -3.8eV, valence band top energy level -5.8eV, carrier concentration 3.5×1016cm-3, surface active site density 4.2×1013cm-2, operating temperature 85°C. The relative error of parameter mapping is 0.006, which meets the accuracy requirements.

[0136] Finally, in step S90, a MoS2 gas sensor sample is prepared according to the optimized parameters, and the performance is verified. The response characteristics of the sample to typical dissolved gases in transformer oil are tested by gas chromatograph, and the test conditions are: injection volume 1mL, carrier gas flow rate 30mL / min, and chromatographic column temperature 80℃. The obtained gas-sensitive response data are shown in Table 5.

[0137] Table 5 Gas-sensitive response characteristics of MoS2 sensor after optimization

[0138] Gas Type Concentration (μL / L) Response value (%) Response time(s) Recovery time(s) H2 100 32.5 45 68 CH4 100 28.7 52 75 C2H2 100 45.6 38 62 C2H4 100 35.8 43 70 C2H6 100 26.4 56 82

[0139] like Figure 7 As shown, the relationship between the sensitivity and selectivity of different gases is shown through a scatter plot, each point represents a gas, and the specific gas type is marked by an annotation. This chart is suitable for being inserted in step S90 to supplement the comprehensive performance of the sensor.

[0140] like Figure 8 The figure shows the response characteristics of the optimized MoS2 sensor to different gases. The figure includes three key parameters: response value (blue column), response time (orange column) and recovery time (green column), which intuitively shows the difference in the sensor's detection performance for five gases: H2, CH4, C2H2, C2H4 and C2H6.

[0141] The experimental results show that the optimized MoS2 sensor has good response characteristics to the target gas, and the maximum relative error between the response value and the theoretical prediction value is 8.5%, which meets the application requirements. The sensor was installed in the oil chromatography online monitoring device of a 220kV transformer and operated continuously for 3 months. The test results were consistent with the laboratory gas chromatography analysis results, and the average relative deviation was less than 10%. The actual application results prove the effectiveness and reliability of the method of the present invention.

[0142] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A TMDCs parameter optimization method for detecting dissolved gas in oil, characterized in that: The following steps are involved: S10, constructing multiple sets of TMDCs material characteristic parameter matrices, including n-dimensional matrices consisting of doping concentration, band structure, carrier concentration, surface active site density, and operating temperature; S20, using gas molecule adsorption experiments to collect and calculate the charge transfer matrix of the interaction between gas molecules and TMDCs materials; S30, using a singular value decomposition method to perform dimensionality reduction processing on the charge transfer matrix to obtain a weight matrix of main influencing factors; S40, performing dimensionality reduction transformation on the characteristic parameter matrix of the TMDCs material according to the weight matrix to obtain an optimization target matrix; S50, establishing a mapping relationship between the optimization target matrix and the gas-sensitive response value, and constructing a response function matrix; S60, processing the response function matrix using a principal component analysis method to extract TMDCs material structure sensitivity factors; S70, establishing a TMDCs material parameter optimization subspace based on the structure sensitive factor, and using a gradient descent method to search for an optimal solution; S80, mapping the optimal solution back to the original parameter space through inverse matrix transformation to obtain an optimized parameter combination of TMDCs materials; S90. Use a gas chromatograph to verify the gas-sensitive response value under the optimized parameter combination, and feed back the verification result to the weight matrix, and iteratively execute steps S40-S90 for optimization. After the iteration is completed, the final optimal parameter combination of TMDCs materials is obtained, including doping concentration, band structure, carrier concentration, surface active site density, and operating temperature.

2. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 1, characterized in that: The characteristic parameter matrix is ​​specifically expressed as: Among them, x i1 represents the doping concentration of the i-th group of samples, x i2 represents the band structure parameters, x i3 represents the carrier concentration, x i4 represents the surface active site density, x i5 represents the operating temperature, and m is the number of sample groups.

3. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 2, characterized in that: The charge transfer matrix is ​​specifically expressed as: where t ij It represents the charge transfer between the i-th sample and the j-th gas molecule, and the calculation formula is: t ij =C i ΔV ij ; In the formula, C i is the equivalent capacitance of the i-th sample, ΔV ij is the potential change of the i-th sample exposed to the j-th gas; the potential change is obtained by time series measurement: ΔV ij (t)=V ij (t)-V ij (0); Where t∈[0,300s] and the sampling interval is 0.01s.

4. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 3, characterized in that: The singular value decomposition is specifically expressed as: T=U∑V T ; in: Select the first k singular values ​​whose cumulative contribution rate exceeds 95%: Construct the truncated singular value matrix ∑ k And the weight matrix W = U k ∑ k .

5. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 4, characterized in that: The optimization target matrix is ​​specifically expressed as: Y=X ′ W; It is expressed as: In the formula, the dimension of Y is m×k, and each row represents the coordinate of a sample in the space after dimensionality reduction.

6. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 5, characterized in that: The response function matrix is ​​specifically expressed as: Build a polynomial regression model: where a ijl and b ij is the regression coefficient.

7. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 6, characterized in that: The step of processing the response function matrix using the principal component analysis method is specifically: Compute the covariance matrix: Solve the characteristic equation: (C-λI)v=0; get the characteristic value λ i and the corresponding eigenvector v i , construct the structural sensitive factor matrix: S = [v1, v2, ..., v r ]; where r satisfies:

8. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 7, characterized in that: The calculation objective function of the TMDCs material parameter optimization subspace is specifically expressed as: The objective function is defined as: where w j is the weight coefficient, is the expected response value, and the gradient descent method is used to solve it: Where η is the learning rate, which ranges from 0.01 to 0.

1.

9. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 8, characterized in that: The steps of matrix inverse transformation are specifically expressed as: Y * =Z * S T ; X * =Y * (W T W) -1 W T ; The parameter restoration formula is:

10. A TMDCs parameter optimization method for detecting dissolved gas in oil according to claim 9, characterized in that: The stopping condition of the iteration is that the cosine similarity of the vector formed by the material optimization parameter combination obtained after 20 consecutive iterations or two adjacent iterations is greater than 0.97.

Images