Optimal selection method of trace characteristic gas detection sensing material in transformer oil

Abstract

The invention provides a method for optimizing a detection sensing material for trace characteristic gas in transformer oil, and belongs to the technical field of detection of gas in transformer oil, and the method comprises the following steps: firstly, obtaining basic data of candidate materials, including material composition, atomic structure, electron energy band structure and theoretical gas sensing performance parameters; and establishing a sensing performance prediction model considering the material composition and structure, and training a material-performance correlation model by using a machine learning algorithm. A large number of theoretical material data sets are generated through computer traversal, and multi-dimensional clustering analysis is performed on candidate materials to obtain a clustering center. And carrying out laboratory performance testing by taking a clustering center corresponding material as a to-be-tested set, and collecting actual performance data. And carrying out multi-generation iterative optimization on the material with the optimal performance by adopting a genetic algorithm to finally obtain the transformer oil trace characteristic gas detection sensing material. The problem that an existing gas sensing material is difficult to meet harsh requirements of an electric power system on online detection of a transformer in the aspects of sensitivity, selectivity and stability is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of gas detection in transformer oil, and in particular relates to a method for optimizing sensing materials for detecting trace characteristic gases in transformer oil. Background Art

[0002] As a key component of power systems, transformers play an important role in power transmission and energy conversion. To ensure stable transformer operation, online monitoring of the transformer's internal operating status has long been a research focus in the field of power engineering. Monitoring trace characteristic gases in transformer oil is one of the most commonly used and effective diagnostic methods. Transformer oil produces several characteristic gases, such as hydrogen (H2), carbon monoxide (CO), carbon dioxide (CO2), methane (CH4), ethane (C2H6), ethylene (C2H4), and acetylene (C2H2). Changes in the concentration of these gases reflect abnormal operating conditions within the transformer's insulation, such as partial discharge, overheating, and arcing. Therefore, online monitoring and analysis of these characteristic gases can promptly identify potential transformer faults and provide an important basis for maintaining reliable transformer operation.

[0003] Existing methods for detecting trace characteristic gases in transformer oil primarily rely on gas chromatography-mass spectrometry. This method offers advantages such as high sensitivity and good analytical accuracy. However, its stringent sample pretreatment requirements, large instrument size, long detection times, and high operating costs limit its application in the field. Therefore, there is an urgent need to develop a simpler, faster, and more cost-effective method for detecting characteristic gases in transformer oil. Gas sensing materials, as an emerging detection technology, have garnered widespread attention in recent years. This method exploits the adsorption and electron transfer mechanisms of specific gases by materials, detecting gases by measuring changes in material parameters such as resistance and capacitance. It offers advantages such as rapid detection, low cost, and compact size. However, existing gas sensing materials still require further improvement in key performance indicators such as sensitivity, selectivity, and stability, and are unable to meet the stringent requirements of power systems for online transformer monitoring. Summary of the Invention

[0004] In view of this, the present invention provides a method for optimizing sensor materials for detecting trace characteristic gases in transformer oil, which can solve the problem that existing gas sensor materials are still difficult to meet the stringent requirements of power systems for online transformer detection in terms of sensitivity, selectivity and stability.

[0005] The present invention is achieved in that:

[0006] The present invention provides a method for optimizing a sensing material for detecting trace characteristic gases in transformer oil, comprising the following steps:

[0007] S10, obtaining basic data of multiple candidate materials for detecting trace characteristic gas in transformer oil, wherein the basic data includes: material composition, atomic structure, electronic band structure, theoretical gas sensing performance parameters, material preparation process parameters, and theoretical response data of the material to different gas molecules;

[0008] S20. Establishing a sensing performance prediction model that takes into account material composition, atomic structure, and electronic band structure, wherein the sensing performance prediction model includes: a material sensitivity equation, a material selectivity equation, a material response time equation, a material stability equation, a material detection limit equation, and an equation for interaction between the material and gas molecules;

[0009] S30, using a machine learning algorithm to train the sensing performance prediction model using the basic data of the multiple candidate materials obtained to obtain a preliminary material-performance correlation model;

[0010] S40, generating a large number of theoretical material data sets using a computer traversal method according to a preset material composition range and adjustment step size, and substituting the large number of theoretical material data sets into a preliminary material-property correlation model for calculation to obtain a large number of candidate sensing materials and their theoretical performance data;

[0011] S50, performing multi-dimensional cluster analysis on a large number of candidate sensing materials based on the theoretical performance data to obtain a number of cluster centers;

[0012] S60, using the sensing material corresponding to each cluster center as a set of test materials, conducting a laboratory trace gas detection performance test and collecting actual performance data;

[0013] S70, selecting the material with the best performance from the set of materials to be tested, forming a plurality of materials as an initial population, normalizing the performance index data as a fitness function, setting the crossover rate and mutation rate, and using a genetic algorithm to perform multi-generation iterative optimization to obtain a final sensing material for detecting trace characteristic gases in transformer oil;

[0014] S80. Perform system performance verification on the final obtained sensing material, including: gas detection sensitivity test, selectivity evaluation, response time determination, stability evaluation and detection limit confirmation.

[0015] On the basis of the above technical solution, the method for optimizing the sensing material for detecting trace characteristic gases in transformer oil of the present invention can also be improved as follows:

[0016] Among them, the various candidate materials include metal oxide materials, transition metal sulfides, doped metal composite oxides, carbon-based composite materials, semiconductor heterojunction materials, nanostructured materials, rare earth metal doped materials and precious metal modified materials.

[0017] Furthermore, the characteristic gases include: H2, CO, CO2, CH4, C2H6, C2H4, and C2H2.

[0018] Furthermore, the material sensitivity equation is used to predict the sensitivity response of the material to the target gas. The input includes the material's band structure parameters, surface active site density, structural characteristics and concentration of gas molecules, and the output is the material's sensitivity coefficient to the specific gas.

[0019] Furthermore, the material selectivity equation is used to evaluate the material's ability to distinguish between different gases. The input includes the material's surface chemical properties, the adsorption energy of gas molecules, the material's surface charge density distribution, and the gas molecule polarizability. The output is the material's selectivity coefficient for the target gas relative to the interfering gas.

[0020] Furthermore, the material response time equation is used to calculate the response speed of the material to gas. The input includes the material's specific surface area, porosity, surface defect density, gas molecular diffusion coefficient, and operating temperature. The output is the time required for the material to reach 90% of its maximum response value.

[0021] Furthermore, the material stability equation is used to predict the long-term performance of the material. The input includes the material's lattice energy, structural stability index, surface bond energy, operating temperature, and ambient humidity. The output is the material's stability index, which includes a comprehensive score of temperature stability, humidity stability, and time stability.

[0022] Furthermore, the material detection limit equation is used to estimate the minimum gas detection concentration of the material. The input includes the material's intrinsic noise level, signal amplification factor, baseline drift rate, temperature coefficient, and response curve slope. The output is the theoretical detection lower limit of the material for a specific gas.

[0023] Furthermore, the gas molecule interaction equation is used to simulate the interaction mechanism between gas molecules and material surfaces. The input includes the atomic arrangement on the material surface, the gas molecular orbital energy level, the electron transfer barrier, the adsorption site distribution, and the surface charge density. The output is the interaction intensity between the gas molecules and the material surface, including the adsorption energy, the electron transfer amount, and the chemical bond strength.

[0024] Wherein, the multi-dimensional cluster analysis adopts weighted Euclidean distance clustering.

[0025] Furthermore, the calculation steps involved in the present invention are described in detail as follows:

[0026] 1. The material sensitivity equation is specifically expressed as follows:

[0027] ;

[0028] Where, is the material sensitivity coefficient (range 0~100); is the material structure coefficient (range 0.1~10); is the material band gap (eV); is the Boltzmann constant; is the operating temperature (K); is the surface active site density (number / nm²); is the concentration of gas molecules (ppm); is the activation energy of gas molecule adsorption (eV); is the charge transfer coefficient (range 0~1); is the electron diffusion coefficient (cm² / s); is the amount of transferred charge (e); is the sensitivity error term; is the number of gas components;

[0029] 2. The material selectivity equation is specifically expressed as follows:

[0030] ;

[0031] Where, is the selectivity coefficient; is the weight coefficient (range 0~1); are the adsorption energies of the target gas and interfering gas (eV), respectively; are the charge density of the target gas and interfering gas on the material surface (C / m²); are the polarizabilities of the target gas and interfering gas (cm³), respectively; is the gas molecule diffusion barrier (eV); is the selective error term; are the component numbers of target gas and interfering gas respectively;

[0032] 3. The material response time equation is specifically expressed as follows:

[0033] ;

[0034] Where, 90% response time (s); is the material characteristic size (nm); is the effective diffusion coefficient (cm² / s); is the specific surface area (m² / g); is the theoretical maximum specific surface area (m² / g); is the diffusion activation energy (eV); is the porosity influence coefficient (range 0~1); is the defect density (pieces / nm²); is the response time error term;

[0035] 4. The material stability equation is specifically expressed as follows:

[0036] ;

[0037] Where, is the stability index (range 0~1); is the weight coefficient and satisfies ; is the lattice energy (kJ / mol); is the surface bond energy (kJ / mol); is the gas constant; is the relative humidity (%); is the critical relative humidity (%); is the humidity impact index; is the stability error term;

[0038] 5. The material detection limit equation is specifically expressed as follows:

[0039] ;

[0040] Where, is the detection limit (ppm); is the noise standard deviation; is the temperature coefficient (K⁻¹); is the signal amplification factor; is the initial sensitivity; is the baseline drift rate (s⁻¹); is the test time (s); is the baseline change; is the initial baseline value; is the detection limit error term;

[0041] 6. The gas molecule interaction equation is specifically expressed as follows:

[0042] ;

[0043] Where, is the interaction energy (eV); are the local charges of the material and gas molecules, respectively; is the interatomic distance (Å); is the potential energy parameter; is the charge transfer energy; is the interaction error term; are the atomic numbers of the material and gas molecules respectively.

[0044] The parameters of the above equation are obtained as follows:

[0045] 1) Band structure parameters ( ): obtained by density functional theory (DFT) calculation;

[0046] 2) Surface active site density ( ): Determined by temperature-course desorption experiment;

[0047] 3) Adsorption energy ( ): obtained by first-principles calculations combined with molecular dynamics simulations;

[0048] 4) Charge density distribution ( ): measured using electrostatic force microscopy (EFM);

[0049] 5) Specific surface area ( ): Determined by nitrogen adsorption-desorption experiments;

[0050] 6) Lattice energy ( ): obtained through quantum chemical calculations;

[0051] 7) Noise level ( ): Obtained through long-term baseline testing statistics;

[0052] 8) Charge transfer amount ( ): determined by in situ X-ray photoelectron spectroscopy (XPS).

[0053] The core principles and significance of these equations are as follows:

[0054] 1. The sensitivity equation takes into account four main influencing factors: band structure, surface activity, gas adsorption, and charge transfer. It uses an exponential relationship to reflect the effect of temperature on electron transport, and a linear relationship to represent the concentration response, reflecting the semiconductor gas sensing mechanism.

[0055] 2. The selectivity equation is based on the ratio of the interaction strength between gas molecules and the material surface, taking into account the synergistic effects of adsorption energy, charge density, and polarizability, and reflects the material's ability to recognize different gases;

[0056] 3. The response time equation is based on Fick's diffusion law, taking into account the influence of the material's microstructural characteristics and using an exponential relationship to describe the effect of temperature on the diffusion process.

[0057] 4. The stability equation comprehensively considers the influence of thermodynamic stability, chemical bond stability and environmental factors, and uses a polynomial weighted approach to evaluate the comprehensive stability performance of the material;

[0058] 5. The detection limit equation is based on the signal-to-noise ratio principle and takes into account the influence of practical factors such as system noise, temperature fluctuation, signal amplification and baseline drift;

[0059] 6. The interaction equation is based on the principle of superposition of intermolecular interaction potential energy, includes the contributions of Coulomb force, van der Waals force and charge transfer, and can accurately describe the interaction mechanism between gas molecules and material surfaces.

[0060] Regarding the calculation process and algorithm in steps S40-S70, the relevant equations are expanded in detail below:

[0061] 1. The component ratio equation generated by theoretical material data in S40:

[0062] ;

[0063] ;

[0064] ;

[0065] Where, For the The component in The content in each ratio; For the Minimum content of each component; For the The adjustment step of the components; is an integer coefficient; is the mole fraction; is the maximum allowable mole fraction; is the total number of components.

[0066] 2. Distance calculation equation for multidimensional cluster analysis in S50:

[0067] ;

[0068] ;

[0069] Where, For samples and The weighted Euclidean distance between them; For the The weight coefficient of the dimension feature; For samples In the The eigenvalue of the dimension; For the The standard deviation of the dimension feature; is the feature dimension.

[0070] 3. Cluster center update equation:

[0071] ;

[0072] ;

[0073] Where, For the Class The cluster centers of the iteration; For the Class sample The eigenvector of For the The number of samples contained in the class; is the clustering objective function; is the number of clusters; For samples To The distance to the cluster center.

[0074] 4. Standardized processing equation for performance test data in S60:

[0075] ;

[0076] ;

[0077] Where, For the Sample No. Normalized value of each indicator; is the original test data; is the vector normalized value; is the number of samples.

[0078] 5. Fitness function of genetic algorithm in S70:

[0079] ;

[0080] ;

[0081] ;

[0082] Where, is the comprehensive fitness value; For the The weight coefficient of each performance indicator; is the normalized performance index value; is the performance fluctuation penalty coefficient; is the standard deviation of the performance index; is the number of performance indicators.

[0083] 6. The probability calculation equation of the crossover operation:

[0084] ;

[0085] ;

[0086] Where, is the crossover probability; are the upper and lower limits of the crossover probability respectively; is the larger fitness value; is the average fitness value of the population; is the maximum fitness value of the population.

[0087] 7. Probability calculation equation for mutation operation:

[0088] ;

[0089] ;

[0090] Where, is the mutation probability; are the upper and lower limits of the mutation probability respectively; is the individual fitness value.

[0091] 8. Selection operator for population update:

[0092] ;

[0093] Where, For individuals Probability of being selected; For individuals The fitness value of is the population size; is the evolutionary pressure coefficient; For the current algebra; is the maximum algebra.

[0094] Parameter acquisition method:

[0095] 1) The original data of performance indicators are obtained through experimental testing, including:

[0096] Gas sensitivity: Standard gas is prepared in proportion through the gas distribution device to measure the resistance change of the material under different gas concentrations;

[0097] Selectivity coefficient: the ratio of the response signal of the measuring material to the target gas and the interfering gas;

[0098] Response recovery time: the time required for the recording material to reach 90% of the maximum response value and recover to 10% of the response value;

[0099] Stability index: Test continuously for 30 days, 3 times a day, and calculate the signal drift rate.

[0100] 2) The weight coefficient is determined by the analytic hierarchy process (AHP):

[0101] Constructing a judgment matrix ;

[0102] Compute eigenvalues ​​and eigenvectors;

[0103] Perform consistency check.

[0104] The core principles and significance of these equations:

[0105] 1. The material ratio equation adopts the equal-step traversal method to ensure full exploration of the component space;

[0106] 2. The cluster analysis equation introduces weighted Euclidean distance to highlight the influence of key features;

[0107] 3. Standardized processing equations eliminate dimensionality effects and achieve unified evaluation of multiple indicators;

[0108] 4. The fitness function comprehensively considers performance level and stability and introduces a penalty mechanism;

[0109] 5. Adaptive genetic operators dynamically adjust operation probabilities based on population evolutionary status, improving algorithm efficiency;

[0110] 6. The selection operator takes into account the influence of evolutionary generations and increases the selection pressure in the later stages.

[0111] Innovations include: introducing performance fluctuation penalty items to improve material stability; using adaptive genetic operators to avoid premature convergence; considering the impact of evolutionary stages to optimize selection strategies; and multi-dimensional performance evaluation to ensure the comprehensive performance of materials.

[0112] Compared with the prior art, the method for selecting the optimal sensor material for detecting trace characteristic gases in transformer oil provided by the present invention has the following beneficial effects:

[0113] 1. Significantly improved sensitivity. By establishing a material sensitivity equation that comprehensively considers factors such as the material's energy band structure, surface active site density, and gas concentration, it is possible to accurately predict the material's response sensitivity to specific gases.

[0114] 2. Significantly improved selectivity. This invention proposes a material selectivity equation that comprehensively evaluates factors such as the material's surface chemical properties, gas molecule adsorption energy, and the material's surface charge density distribution, effectively distinguishing target gases from interfering gases.

[0115] 3. Significantly improved response speed. The material response time equation established by this invention takes into account factors such as the material's specific surface area, porosity, surface defect density, and gas molecular diffusion coefficient, and can accurately predict the material's response speed to gas.

[0116] 4. Long-term stability is greatly improved. The material stability equation established by this invention comprehensively considers factors such as the material's lattice energy, structural stability index, and surface bond energy, and can predict the long-term stability of the material under conditions such as working temperature and ambient humidity.

[0117] In summary, the method for optimizing sensor materials for detecting trace characteristic gases in transformer oil proposed in the present invention fully leverages the advantages of multidisciplinary technologies such as materials science, machine learning, and genetic algorithms, and effectively solves the problem that existing gas sensing materials are still unable to meet the stringent requirements of the power system for online transformer detection in terms of sensitivity, selectivity, and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0118] Figure 1 A flow chart of the method provided by the present invention;

[0119] Figure 2 Schematic diagram of the gas sensing test platform. DETAILED DESCRIPTION

[0120] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0121] like Figure 1 FIG. 1 is a flow chart of a method for optimizing a sensing material for detecting trace characteristic gases in transformer oil provided by the present invention. The method comprises the following steps:

[0122] S10. Obtain basic data of multiple candidate materials for detecting trace characteristic gases in transformer oil, including material composition, atomic structure, electronic band structure, theoretical gas sensing performance parameters, material preparation process parameters, and theoretical response data of the materials to different gas molecules;

[0123] S20. Establish a sensing performance prediction model that takes into account material composition, atomic structure, and electronic band structure. The sensing performance prediction model includes: a material sensitivity equation, a material selectivity equation, a material response time equation, a material stability equation, a material detection limit equation, and an equation for interaction between the material and gas molecules;

[0124] S30, using a machine learning algorithm to train a sensing performance prediction model using the basic data of multiple candidate materials to obtain a preliminary material-performance correlation model;

[0125] S40, generating a large number of theoretical material data sets using a computer traversal method according to a preset material composition range and adjustment step size, and substituting the large number of theoretical material data sets into a preliminary material-performance correlation model for calculation to obtain a large number of candidate sensing materials and their theoretical performance data;

[0126] S50, performing multi-dimensional cluster analysis on a large number of candidate sensing materials based on theoretical performance data to obtain several cluster centers;

[0127] S60, using the sensing material corresponding to each cluster center as a set of test materials, conducting a laboratory trace gas detection performance test and collecting actual performance data;

[0128] S70, using multiple materials with the best performance among the materials to be tested as the initial population, using the performance index data for normalization as the fitness function, setting the crossover rate and mutation rate, and using a genetic algorithm to obtain the final sensing material for detecting trace characteristic gases in transformer oil through multi-generation iterative optimization;

[0129] S80. Perform system performance verification on the final obtained sensing material, including: gas detection sensitivity test, selectivity evaluation, response time determination, stability evaluation and detection limit confirmation.

[0130] The specific implementation of the above steps is described in detail below:

[0131] The specific implementation method of step S10 is to obtain the basic data of candidate materials through professional databases and material characterization methods. First, the material genome project database is used to collect the crystal structure parameters of the candidate materials, including lattice constants, space groups, atomic position coordinates and other information. Secondly, the density functional theory calculation software is used to obtain the electronic structure characteristics of the material, including band structure, state density, Fermi level position, etc. Then, the first principle calculation method is used to simulate the adsorption configuration and adsorption energy of different gas molecules on the surface of the material. Then, through literature research and laboratory preliminary research data, the preparation process parameters of the material are collected, such as synthesis temperature in the range of 400 to 1200 degrees Celsius, pressure in the range of 0.1 to 10 MPa, and reaction time in the range of 0.5 to 24 hours. Finally, a material database is established to systematically organize and classify all basic data. These data provide the basis for the subsequent establishment of a sensing performance prediction model.

[0132] The specific implementation of step S20 is to construct a multi-dimensional sensing performance prediction model.

[0133] The material sensitivity equation is used to predict the sensitivity response of the material to the target gas. The input includes the material's band structure parameters, surface active site density, and the structural characteristics and concentration of the gas molecules. The output is the material's sensitivity coefficient to the specific gas.

[0134] The material selectivity equation is used to evaluate the material's ability to distinguish between different gases. The input includes the material's surface chemical properties, the adsorption energy of gas molecules, the material's surface charge density distribution, and the gas molecule polarizability. The output is the material's selectivity coefficient for the target gas relative to the interfering gas.

[0135] The material response time equation is used to calculate the material's response speed to gas. The inputs include the material's specific surface area, porosity, surface defect density, gas molecular diffusion coefficient, and operating temperature. The output is the time required for the material to reach 90% of its maximum response value.

[0136] The material stability equation is used to predict the long-term performance of a material. The input includes the material's lattice energy, structural stability index, surface bond energy, operating temperature, and ambient humidity. The output is the material's stability index, which includes a comprehensive score of temperature stability, humidity stability, and time stability.

[0137] The material detection limit equation is used to estimate the minimum gas detection concentration of a material. The input includes the material's intrinsic noise level, signal amplification factor, baseline drift rate, temperature coefficient, and response curve slope. The output is the theoretical lower detection limit of the material for a specific gas.

[0138] The gas molecule interaction equation is used to simulate the interaction mechanism between gas molecules and material surfaces. The input includes the atomic arrangement on the material surface, the gas molecular orbital energy level, the electron transfer barrier, the adsorption site distribution, and the surface charge density. The output is the interaction intensity between the gas molecules and the material surface, including the adsorption energy, the electron transfer amount, and the chemical bond strength.

[0139] The specific implementation method of step S30 is to use a machine learning algorithm to train a sensing performance prediction model. First, data preprocessing technology is used to standardize and normalize the basic material data to eliminate the dimensional effect. Secondly, the principal component analysis method is used to reduce the data dimension and extract the main characteristic parameters. Then, a deep neural network algorithm is used to establish a multi-layer perceptron model. The number of input layer nodes is the same as the number of material characteristic parameters. The hidden layer adopts a 3-layer structure, and the number of nodes in each layer is 128, 64, and 32 respectively. The number of output layer nodes is the same as the number of performance indicators. Then, the stochastic gradient descent optimization algorithm is used, and the learning rate is set to 0.001, the batch size is 32, and the number of training rounds is 1000 to minimize the mean square error loss function. After that, the cross-validation method is used to divide the data set into a training set and a validation set at a ratio of 8 to 2 to evaluate the generalization ability of the model. Finally, the network weights are continuously adjusted by the back propagation algorithm until the model prediction error is less than 5%.

[0140] The specific implementation method of step S40 is to generate and screen theoretical material data. First, set the material composition range. For doped materials, the doping concentration range is 0.1% to 10%, and the adjustment step size is 0.1%. Secondly, use the Monte Carlo algorithm to randomly generate 1 million sets of theoretical material component data within the set range. Then use a high-throughput calculation method to input the generated material data into the trained performance prediction model. Then use parallel computing technology to process multiple sets of data at the same time to improve computing efficiency. The calculation results are preliminarily screened again to eliminate data that does not conform to the basic physical and chemical laws. Finally, retain materials whose theoretical performance indicators meet the following conditions: sensitivity greater than 50 microvolts per microliter per liter, selectivity coefficient greater than 10, response time less than 60 seconds, stability loss less than 10% per year, and detection limit less than 5 microliters per liter.

[0141] The specific implementation method of step S50 is to perform multidimensional cluster analysis. First, the theoretical performance data of the material is standardized to eliminate the dimensional differences between different indicators. Secondly, the principal component analysis method is used to extract the main components that can explain 90% of the variance. Then, the improved K-means clustering algorithm is used to determine the optimal number of clusters through the silhouette coefficient, which is initially set to 5 to 10 clusters. Then, the density peak clustering method is used to identify the cluster centers of high-density areas. The hierarchical clustering algorithm is used again to construct a cluster dendrogram to verify the rationality of the clustering results. Finally, the three materials with the best theoretical performance in the neighborhood of each cluster center are selected as representatives to form the set of materials to be tested.

[0142] The specific implementation method of step S60 is to conduct laboratory performance testing. First, the material to be tested is prepared according to the optimized process parameters, using methods such as hydrothermal method, sol-gel method, and chemical vapor deposition method. Secondly, the morphology and structure of the material are characterized by means of scanning electron microscopy, transmission electron microscopy, X-ray diffraction, etc. Then a gas sensing test platform is built, and then a standard gas detection experiment is carried out. The gas concentration range is 1 to 1000 microliters per liter, the temperature range is 25 to 200 degrees Celsius, and the humidity range is 30% to 70%. The response characteristics of the material to the mixed gas are tested again to evaluate the anti-interference ability. Finally, the experimental data are recorded and organized, including parameters such as response curve, sensitivity, selectivity, and response recovery time.

[0143] The specific implementation of step S60 is further described in detail as follows.

[0144] First, the materials to be tested are prepared using a variety of advanced material preparation methods. For metal oxide materials, a sol-gel method is used. The specific steps include: preparing a metal salt precursor solution with a concentration controlled within the range of 0.1 to 0.5 mol / L; adding a chelating agent such as ethanolamine, controlling the molar ratio of the chelating agent to the metal ion to be 1.5 to 2; stirring at 60 to 80 degrees Celsius for 4 to 6 hours to form a sol; aging the sol at 80 to 100 degrees Celsius for 12 to 24 hours to form a gel; drying the gel at 120 degrees Celsius for 8 to 12 hours; and finally calcining in a muffle furnace at a heating rate of 2 degrees Celsius per minute to a temperature between 500 and 800 degrees Celsius for 2 to 4 hours. Transition metal sulfide materials are prepared using a hydrothermal method, which includes dissolving a metal salt and a sulfur source in deionized water in a stoichiometric ratio; adjusting the solution pH to 6 to 8; transferring the mixture to a polytetrafluoroethylene-lined stainless steel reactor; reacting at 160 to 220 degrees Celsius for 12 to 36 hours; and centrifuging, washing, and drying the product. Carbon-based composite materials are grown using chemical vapor deposition in a tube furnace using methane or acetylene as the carbon source at 800 to 1000 degrees Celsius, with a controlled gas flow rate of 50 to 100 ml / min and a growth time of 30 to 60 minutes.

[0145] Next, the materials were characterized and analyzed. The crystal structure of the materials was analyzed using an X-ray diffractometer with a scanning angle range of 10 to 80 degrees, a scanning rate of 5 degrees per minute, and a step size of 0.02 degrees. The surface morphology of the materials was observed using a field-emission scanning electron microscope with an accelerating voltage of 5 to 15 kilovolts and a working distance of 8 to 12 mm. The microstructure of the materials was analyzed using a transmission electron microscope with an accelerating voltage of 200 kilovolts and observations in at least five different fields of view. The surface elemental composition and chemical state of the materials were analyzed using an X-ray photoelectron spectrometer, using a monochromated aluminum target X-ray excitation source, a vacuum better than 10-8 Pa, and an energy resolution better than 0.5 electron volts. The specific surface area and pore structure of the materials were determined using a surface area analyzer. The samples were first degassed at 150 degrees Celsius for 12 hours, followed by nitrogen adsorption and desorption tests at liquid nitrogen temperature. The molecular structure and defect characteristics of the materials were analyzed using a Raman spectrometer with an excitation wavelength of 532 nanometers, a power density controlled below 2 milliwatts per square micron, and an acquisition time of 60 seconds, with three cumulative acquisitions.

[0146] Then build a gas sensing test platform, such as Figure 2As shown. The test system includes a gas supply unit, a mixing reaction unit, a detection unit, and a data acquisition unit. The gas supply unit is equipped with seven standard gas cylinders with a purity of not less than 99.999%, and is equipped with a high-precision mass flow controller with a control accuracy of better than 1%. Zero-grade air is used as the carrier gas and dilution gas. The mixing reaction unit is made of quartz glass, with an inner diameter of 50 mm and a length of 200 mm, and has good airtightness and corrosion resistance. The cavity temperature is controlled by a precision thermostat with a temperature range of room temperature to 300 degrees Celsius and a temperature control accuracy of ±0.1 degrees Celsius. The relative humidity is controlled by a saturated salt solution method with an adjustable range of 20% to 90%. The detection unit uses a four-probe method to measure the resistance change of the material, using a high-precision digital source meter with a measurement accuracy of better than 0.1%. The data acquisition unit includes an analog-to-digital converter and a computer, with a sampling frequency of up to 1000 Hz, and has real-time display and automatic storage functions.

[0147] Next, a standard gas detection experiment was conducted. First, the test system was aged and calibrated, and zero-grade air was introduced at operating temperature for at least 24 hours. Before testing, the sensing material was pretreated at 200 degrees Celsius for 2 hours to remove surface adsorbents. Standard gases of varying concentrations were prepared using a dynamic gas distribution method, with concentration gradients set at 1, 2, 5, 10, 20, 50, 100, 200, 500, and 1000 microliters per liter. For each concentration point, the gas was introduced for 10 minutes and recovered for 20 minutes, with the test repeated at least three times. The test temperatures were set at 25, 50, 100, 150, and 200 degrees Celsius, respectively, and the relative humidity was set at 30%, 50%, and 70%. The resistance curves of the material under different conditions were recorded over time, and parameters such as the response value, response time, and recovery time were calculated. The response value is defined as the ratio of the resistance change in the presence of gas to the initial resistance, the response time is defined as the time required for the signal to reach 90% of its maximum value, and the recovery time is defined as the time required for the signal to return to 10% of its initial value.

[0148] Conduct the mixed gas detection experiment again. Prepare mixed gases with different component ratios, including binary mixtures and multi-component mixtures. The concentration ratio of the target gas to the interfering gas in the binary mixture is set to 1:1, 1:5, 1:10, 5:1, and 10:1. Multi-component mixtures contain more than three gas components, and the total concentration is controlled within 1000 microliters per liter. Investigate the selectivity and anti-interference ability of the material for the mixed gas and calculate the selectivity coefficient. The selectivity coefficient is defined as the ratio of the material's response value to the target gas to its response value to the interfering gas. Evaluate the cross-interference effect between different gas components and analyze the practical application potential of the material.

[0149] Finally, the experimental data is systematically organized and analyzed. A database is established to record all test results, including basic material information, test conditions, response curves, performance parameters, etc. Statistical analysis is performed on the data from repeated tests, and the average and standard deviation are calculated. A curve is plotted showing the relationship between the material response value and gas concentration, and the linear range and saturation characteristics are analyzed. A model is established for the relationship between response time and temperature to optimize the operating temperature. The influence of humidity on material performance is analyzed to determine the optimal operating humidity range. The reproducibility of the material is evaluated, requiring the relative standard deviation of the test under the same conditions to be less than 5%. Based on the experimental results, the comprehensive performance of the material is scored and ranked to provide a basis for subsequent optimization design.

[0150] Evaluate the practical application feasibility of the material. Conduct a continuous working test, running continuously for 72 hours under optimal working conditions, and record the baseline value and response value every 4 hours. Conduct environmental adaptability tests, including temperature cycle tests, humidity cycle tests, vibration tests, etc. The temperature cycle test range is -20 degrees Celsius to 200 degrees Celsius, and the number of cycles is not less than 50 times. The humidity cycle test range is 20% to 90% relative humidity, and the number of cycles is not less than 100 times. The vibration test frequency is 10 to 55 Hz, with an amplitude of 0.5 mm and a duration of 2 hours. Evaluate the material's anti-poisoning ability and test the effects of common interferences such as hydrogen sulfide, ammonia, and methanol on material performance. Analyze the service life of the material and predict the replacement cycle based on the performance attenuation trend. Comprehensively consider factors such as the cost of the material, the complexity of the preparation process, and environmental friendliness to determine the candidate material with the most practical value.

[0151] The specific implementation method of step S70 is to use a genetic algorithm to optimize the material composition. First, the 10 materials with the best experimental performance are used as the initial population and encoded as a binary string. Secondly, the fitness function is designed, comprehensively considering indicators such as sensitivity, selectivity, response time, stability, and detection limit, and a weighted summation method is used for normalization. Then the genetic algorithm parameters are set, including a population size of 100, a crossover rate of 0.8, a mutation rate of 0.1, and a maximum number of iterations of 200 generations. Then, the tournament selection strategy is adopted to select dominant individuals for crossover recombination. The adaptive mutation mechanism is introduced again to dynamically adjust the mutation probability according to the population diversity. Finally, when the optimal fitness value changes by less than 0.1% for 20 consecutive generations, the iteration is terminated and the material composition corresponding to the optimal solution is output.

[0152] The specific implementation method of step S80 is to verify the performance of the system. First, prepare sensor samples according to the optimized material composition, and the preparation quantity is not less than 10. Secondly, under standard laboratory conditions, use the constructed gas sensing test platform to test the detection sensitivity of the samples to 7 characteristic gases, and require the detection limit to reach 1 microliter per liter. Then evaluate the selectivity of the material, calculate the ratio of the response signals of different gases, and require the selectivity coefficient of the target gas to the interfering gas to be greater than 20. Then measure the response time and recovery time, and require the response time to be less than 30 seconds and the recovery time to be less than 60 seconds. Perform stability tests again, including thermal cycle tests, humidity cycle tests, long-term drift tests, etc., and require that the performance degradation within 6 months is less than 5%. Finally, confirm the detection limit, use the 3 times signal-to-noise ratio method, and require it to reach the sub-microliter per liter level.

[0153] In the specific experimental process, a sensitivity test is first carried out. No less than 10 prepared sensor samples are placed in the test chamber and pretreated with zero gas at 200°C for 2 hours to remove surface adsorbents. The temperature is then adjusted to the target test temperature (ranging from 25-200°C). A mass flow controller is used to accurately prepare standard gas with concentrations ranging from 1 to 1000μL / L. The test concentration points are set to ten gradients of 1, 2, 5, 10, 20, 50, 100, 200, 500 and 1000μL / L. The test is repeated three times at each concentration point to ensure the reliability of the data. By recording the resistance change of each sample at each concentration point, the response value (defined as the percentage of the resistance change to the initial resistance) is calculated. Finally, a curve of the relationship between the response value and the concentration is plotted to determine the sensitivity characteristics of the material.

[0154] In the selectivity evaluation experiment, the response characteristics of seven target gases are tested separately. The response values ​​of each gas at the same concentration are recorded. The selectivity coefficient is calculated by calculating the ratio of the target gas response value to the interfering gas response value. At the same time, gas mixtures of different proportions are prepared to evaluate the cross-interference effect to ensure the material's accurate detection capability in complex atmospheres. For the response characteristic test, a dynamic response curve is obtained by rapidly switching between the target gas and the zero gas. Key parameters such as response time (time required for the signal to reach 90% of the maximum value), recovery time (time required for the signal to return to 10% of the initial value), and baseline drift are extracted from the curve. Digital filtering technology is used to eliminate test noise, and the average response time and standard deviation of the material are finally obtained.

[0155] In the stability evaluation experiment, a temperature cycling test is first conducted, with the temperature range set between -20°C and 200°C, and a cumulative number of cycles of no less than 50, with detailed records of changes in material performance with temperature. Secondly, a humidity cycling test is conducted, with the relative humidity cycling between 20% and 90%, with a cumulative number of cycles of no less than 100, and real-time monitoring of changes in the material's response value. Finally, a long-term stability test is conducted, with continuous operation for 72 hours under optimal working conditions, and performance parameters recorded every 4 hours. Data analysis is used to evaluate the material's performance degradation within 6 months. In the detection limit confirmation experiment, low-concentration gas (concentration range between 0.1-1μL / L) is prepared for precise measurement, recording the response signal and system noise, and using the triple signal-to-noise ratio method to determine the theoretical detection limit of the material, while also verifying the repeatability of the test data.

[0156] Ultimately, all test data must meet the following performance indicators: detection limit no higher than 1 μL / L, selectivity coefficient no lower than 20, response time no longer than 30 seconds, recovery time no longer than 60 seconds, performance degradation no more than 5% within 6 months, and relative standard deviation of test data within 5%. This complete experimental setup and rigorous testing process allows for comprehensive evaluation of the various performance indicators of sensing materials, providing reliable experimental basis and technical support for their practical applications.

[0157] Specifically, the present invention establishes a comprehensive sensor performance prediction model and intelligently optimizes material composition through machine learning and genetic algorithms. This comprehensive simulation optimization technology effectively overcomes the limitations of existing empirical testing methods, significantly improving the targeted and efficient development of sensor materials.

[0158] Specifically, the present invention first establishes six key sensing performance prediction models: material sensitivity equation, material selectivity equation, material response time equation, material stability equation, material detection limit equation, and gas molecule-material interaction equation. These models comprehensively consider numerous factors, including the material's energy band structure, surface chemical properties, pore characteristics, and gas molecular properties, enabling accurate prediction of the material's detection performance under various operating conditions.

[0159] Secondly, the present invention uses a machine learning algorithm to train and optimize the aforementioned sensor performance prediction model. First, using techniques such as data preprocessing and principal component analysis, the basic data of multiple candidate materials is normalized to extract the most critical characteristic parameters. Next, a deep neural network model is constructed, with the material characteristic parameters as the input layer, a multi-layered hidden layer structure, and the various performance indicators as the output layer. Using a backpropagation algorithm to continuously adjust the network weights, the model's prediction error is ultimately reduced to less than 5%, achieving accurate prediction of sensor performance.

[0160] Third, the present invention uses a computer-generated traversal approach to generate a large amount of theoretical material data within a range of material compositions. This data is then calculated using the previously trained prediction model to obtain a large number of candidate sensing materials and their theoretical performance data. To further select high-quality materials, the present invention employs cluster analysis techniques to classify candidate materials across multiple dimensions based on performance indicators, identifying the core points of each material category. Laboratory trace gas detection performance testing is then performed on the materials corresponding to each core point to obtain actual performance data.

[0161] Finally, the present invention uses a genetic algorithm to optimize experimental test data. First, multiple materials with the best performance are selected as an initial population and encoded into a binary string. A fitness function is designed that comprehensively considers indicators such as sensitivity, selectivity, response speed, stability, and detection limit, and normalization is performed using a weighted summation method. Through genetic algorithm operations such as population selection, crossover recombination, and adaptive mutation, a stepwise iterative optimization is performed, ultimately obtaining the optimal sensing material composition that meets all performance indicators.

[0162] The following provides an example of a specific application scenario of the present invention: This example describes a process for applying the present method to develop a high-performance tin dioxide-based sensing material for acetylene gas detection in transformer oil. Acetylene was chosen as the target gas because it is a characteristic gas of transformer arc discharge faults, and its detection is crucial for safe transformer operation.

[0163] First, step S10 is performed to collect basic data for candidate materials. This example selected the following materials as initial research targets: pure tin dioxide, tungsten-doped tin dioxide, platinum-modified tin dioxide, palladium-modified tin dioxide, and nickel-doped tin dioxide. The crystal structure parameters of these materials were obtained from the Materials Genome Project database. Pure tin dioxide belongs to the tetragonal system with a space group of P42 / mnm, lattice constants a=b=4.738 Å, and c=3.187 Å. Density functional theory calculations were performed using the VASP software package, employing the PBE functional and PAW pseudopotentials, a plane wave cutoff energy set to 500 eV, and a 7×7×7 Brillouin zone k-point grid. The calculated band gap of pure tin dioxide is 3.6 eV, the conduction band bottom energy is -4.5 eV, and the valence band top energy is -8.1 eV. For the doped material, a 2×2×2 supercell model was used, with a doping concentration of 3.125%. Calculation results show that tungsten doping reduces the band gap to 3.2 electron volts, and platinum modification forms a deep energy level on the surface with an energy located 2.1 electron volts above the top of the valence band.

[0164] Next, a material preparation process database was established. Pure tin dioxide was prepared by the sol-gel method. The specific process parameters were: stannous chloride as a precursor at a concentration of 0.2 mol / L, ethanolamine as a chelating agent at a molar ratio of 1.8 to 1, stirring at 70 degrees Celsius for 5 hours, aging for 18 hours, drying at 120 degrees Celsius for 10 hours, and calcining at 600 degrees Celsius for 3 hours. Tungsten-doped tin dioxide was prepared by adding ammonium metatungstate to the above process, with a doping concentration of 3 weight percent. Precious metal modification was performed by impregnation, with chloroplatinic acid and palladium chloride solution as precursors, a loading of 1 weight percent, and reduction at 350 degrees Celsius for 2 hours.

[0165] In step S20, a sensing performance prediction model is constructed. Taking acetylene detection as an example, the material structure coefficient alpha in the sensitivity equation is set to 2.5, the surface active site density is determined to be 4.2 per square nanometer by temperature-dependent desorption, and the adsorption activation energy of acetylene molecules is calculated to be 0.82 electron volts by first-principles calculations. The weight coefficient gamma in the selectivity equation is set to 0.6, the charge density is measured to be 3.8 microcoulombs per square meter by electrostatic force microscopy, and the polarizability of acetylene molecules is 3.33 cubic angstroms. The material characteristic dimension L in the response time equation is 50 nanometers, and the effective diffusion coefficient is 2.1×10-4 square centimeters per second. The lattice energy in the stability equation is calculated to be 725 kilojoules per mole by quantum chemical calculations.

[0166] In step S30, a deep learning algorithm was used to train the model. A three-layer neural network was constructed, with an input layer containing 12 nodes, corresponding to the basic characteristic parameters of the material. The first hidden layer had 64 nodes, the second hidden layer had 32 nodes, and the output layer had 5 nodes, corresponding to sensitivity, selectivity, response time, stability, and detection limit, respectively. The Adam optimizer was used with a learning rate of 0.001 and a batch size of 32. Using mean squared error as the loss function, the prediction error on the validation set dropped to 4.2% after 1000 training rounds.

[0167] In step S40, the material composition optimization range is set. For tungsten-doped tin dioxide, the doping concentration range is 0.5% to 5%, with a step size of 0.1%. For noble metal modification, the loading range is 0.2% to 2%, with a step size of 0.05%. The Monte Carlo method is used to generate 1 million sets of theoretical material compositions, which are substituted into the trained model for calculation. After preliminary screening, 20,000 sets of material composition data that meet the basic requirements are retained.

[0168] In step S50, cluster analysis was performed on 20,000 groups of materials. Principal component analysis was used to extract features, and the cumulative variance contribution of the first three principal components reached 92.3%. Using the K-means clustering algorithm, silhouette coefficient analysis determined that the optimal number of clusters was 6. The resulting cluster centers are shown in the table below.

[0169] Table 1 Performance indicators of material cluster centers.

[0170]

[0171] In step S60, laboratory performance tests are performed. First, samples of the six materials in Table 1 are prepared, and the morphology is observed by scanning electron microscopy. The particle size distribution is in the range of 30 to 80 nanometers. X-ray diffraction analysis shows that all samples are tetragonal tin dioxide with no obvious impurity phase. The specific surface area test results are between 45 and 60 square meters per gram. Subsequently, a gas sensing performance test is performed, and the test conditions are: temperature 150 degrees Celsius, relative humidity 50%, and acetylene concentration range of 1 to 1000 microliters per liter. The test results show that the Pt-W co-doped SnO2 material has the best performance, with a response value of 8.2% to 1 microliter per liter of acetylene, a response time of 18.5 seconds, and a recovery time of 42.3 seconds.

[0172] Table 2 Experimentally measured gas sensing properties of the materials.

[0173]

[0174] In step S70, the three best performing materials (Pt-W co-doped SnO2, Pd-W co-doped SnO2, and tungsten-doped SnO2) are used as the initial population for genetic algorithm optimization. The fitness function is designed as follows: , where R is the normalized response value, S is the selectivity coefficient, T is the response time score, and P is the stability score. The population size was set to 100, the crossover rate to 0.85, and the mutation rate to 0.08. After 150 generations of iterative optimization, the optimal material composition was obtained: 0.92 weight percent platinum doping and 2.35 weight percent tungsten doping.

[0175] In step S80, the optimized material was systematically verified. Ten sensor samples were prepared under identical conditions and their performance was tested at various temperatures (100°C to 200°C) and humidity levels (30% to 70%). The results showed optimal performance at 150°C and 50% relative humidity. The selectivity test results for seven characteristic gases are shown in the table below.

[0176] Table 3 Selectivity test results of optimized materials for different gases (gas concentration is 100 μL / L).

[0177]

[0178] Long-term stability testing showed that after 72 hours of continuous operation, baseline drift was less than 3% and response attenuation was less than 4%. Performance retention reached 95.8% after temperature cycling testing (-20°C to 200°C, 50 cycles). Performance retention reached 94.2% after humidity cycling testing (20% to 90% relative humidity, 100 cycles). Detection limit testing, using a 3x signal-to-noise ratio method, showed a theoretical detection limit of 0.8 μL / L for acetylene and an actual detection limit of 1 μL / L.

[0179] Anti-interference performance tests show that the detection error for 10 microliters per liter of acetylene is less than 8% in the presence of 100 microliters per liter of hydrogen sulfide or ammonia. Testing using transformer oil samples from different batches revealed a detection error within ±5%, demonstrating the material's excellent repeatability and reliability. Cost analysis shows that the material's preparation cost is approximately 200 yuan per gram, which is very cost-effective considering that a single sensor only requires a few milligrams of material.

[0180] The Pt-W co-doped SnO2 sensing material obtained through system optimization in this example demonstrated excellent comprehensive performance: the detection limit for acetylene reached 1 microliter per liter, the response time was less than 20 seconds, the selectivity coefficient was greater than 20, and the stability retention rate exceeded 94%.

[0181] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A method for optimizing sensor materials for detecting trace characteristic gases in transformer oil, characterized in that: The following steps are involved: S10, obtaining basic data of multiple candidate materials for detecting trace characteristic gas in transformer oil, wherein the basic data at least includes: material composition, atomic structure, electronic energy band structure, theoretical gas sensing performance parameters, material preparation process parameters, and theoretical response data of the material to different gas molecules; S20. Establishing a sensing performance prediction model that takes into account material composition, atomic structure, and electronic band structure, wherein the sensing performance prediction model includes: a material sensitivity equation, a material selectivity equation, a material response time equation, a material stability equation, a material detection limit equation, and an equation for interaction between the material and gas molecules; S30, using a machine learning algorithm to train the sensing performance prediction model using the basic data of the multiple candidate materials obtained to obtain a preliminary material-performance correlation model; S40, generating a large number of theoretical material data sets using a computer traversal method according to a preset material composition range and adjustment step size, and substituting the large number of theoretical material data sets into a preliminary material-property correlation model for calculation to obtain a large number of candidate sensing materials and their theoretical performance data; S50, performing multi-dimensional cluster analysis on a large number of candidate sensing materials based on the theoretical performance data to obtain a number of cluster centers; S60, using the sensing material corresponding to each cluster center as a set of test materials, conducting a laboratory trace gas detection performance test and collecting actual performance data; S70, selecting the material with the best performance from the set of materials to be tested, forming a plurality of materials as an initial population, normalizing the performance index data as a fitness function, setting the crossover rate and mutation rate, and using a genetic algorithm to perform multi-generation iterative optimization to obtain a final sensing material for detecting trace characteristic gases in transformer oil; S80. Perform system performance verification on the final obtained sensing material, including: gas detection sensitivity test, selectivity evaluation, response time determination, stability evaluation and detection limit confirmation.

2. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 1, wherein: The multiple candidate materials include metal oxide materials, transition metal sulfides, doped metal composite oxides, carbon-based composite materials, semiconductor heterojunction materials, nanostructured materials, rare earth metal doped materials and precious metal modified materials.

3. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 2, characterized in that: The characteristic gases include: H2, CO, CO2, CH4, C2H6, C2H4, and C2H2.

4. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 3, wherein: The material sensitivity equation is used to predict the sensitivity response of the material to the target gas. The input includes the material's band structure parameters, surface active site density, structural characteristics and concentration of gas molecules, and the output is the material's sensitivity coefficient to the specific gas.

5. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 4, characterized in that: The material selectivity equation is used to evaluate the material's ability to distinguish between different gases. The input includes the material's surface chemical properties, the adsorption energy of gas molecules, the material's surface charge density distribution, and the gas molecule polarizability. The output is the material's selectivity coefficient for the target gas relative to the interfering gas.

6. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 5, characterized in that: The material response time equation is used to calculate the response speed of a material to gas. The input includes the material's specific surface area, porosity, surface defect density, gas molecular diffusion coefficient, and operating temperature. The output is the time required for the material to reach 90% of its maximum response value.

7. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 6, characterized in that: The material stability equation is used to predict the long-term performance of a material. The input includes the material's lattice energy, structural stability index, surface bond energy, operating temperature, and ambient humidity. The output is the material's stability index, which includes a comprehensive score of temperature stability, humidity stability, and time stability.

8. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 7, characterized in that: The material detection limit equation is used to estimate the minimum gas detection concentration of a material. The input includes the material's intrinsic noise level, signal amplification factor, baseline drift rate, temperature coefficient, and response curve slope. The output is the material's theoretical lower detection limit for a specific gas.

9. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 8, characterized in that: The gas molecule interaction equation is used to simulate the interaction mechanism between gas molecules and material surfaces. The input includes the atomic arrangement on the material surface, the gas molecular orbital energy level, the electron transfer barrier, the adsorption site distribution, and the surface charge density. The output is the interaction intensity between the gas molecules and the material surface, including the adsorption energy, the electron transfer amount, and the chemical bond strength.

10. The method for selecting a sensor material for detecting trace characteristic gases in transformer oil according to claim 1, characterized in that: The multidimensional cluster analysis adopts weighted Euclidean distance clustering.

Images