System and method for characterizing water content of fuel cell membrane

Abstract

The invention discloses a fuel cell membrane water content characterization system and method, and belongs to the technical field of fuel cells. According to the method, an impedance test platform composed of a gas supply device, a gas pre-cooling device, a constant-temperature water area, a controller, a high and low temperature environment box and an electrochemical workstation is built, a cold start variable working condition impedance test and a water content calibration test are carried out, and a working condition-impedance data set and a working condition-water content data set are built. Fitting and predicting the data set by means of a Gaussian process regression model, and obtaining a high-precision variable-working-condition impedance-water content data set through experimental verification and correction. The problem that an existing membrane water content characterization method cannot cope with variable working conditions is effectively solved, accurate mapping of the membrane water content and the variable working conditions is achieved, and powerful technical support is provided for fuel cell internal water content state characterization research under the variable working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of fuel cells, and in particular to a fuel cell membrane water content characterization system and method. Background Art

[0002] Accurate characterization of the membrane's hydration state is a prerequisite for conducting accelerated stress cycling tests to investigate cold-start durability damage. Currently, there are several methods for accurately characterizing the internal hydration state of fuel cells, but they still face the following challenges when responding to variable operating conditions:

[0003] Optical imaging methods are complex and rely on steady-state conditions. Methods capable of accurately characterizing the water content within fuel cells primarily include neutron imaging, X-ray imaging, nuclear magnetic resonance (NMR), and electron microscopy. These methods generally suffer from the following limitations: They rely on complex optical imaging systems, resulting in cumbersome procedures and long testing cycles; they only capture localized static water distribution information under steady-state conditions; and they are highly sensitive to test conditions, limited by the equipment's resolution and penetration depth, making them incapable of detecting water content under variable operating conditions. These optical detection methods are subject to long testing cycles and limited test conditions.

[0004] The impedance test method is susceptible to multivariate nonlinear interference, and the accuracy of water content characterization under variable operating conditions is insufficient: EIS can indirectly reflect the coupling mechanism of water transport and electrochemical reaction in the fuel cell membrane electrode through the frequency domain response characteristics. Existing studies have verified the quantitative relationship between water content and membrane conductivity and diffusion layer mass transfer impedance under controllable steady-state conditions in the laboratory through EIS. However, the nonlinear interaction between actual operating variables (such as the synergistic effect of temperature and humidity on membrane hydration) will cause the characteristic frequency band of the impedance spectrum to shift significantly. The single steady-state laboratory test data relied on by traditional methods is difficult to cover the variable combination space of actual complex operating scenarios, and the equivalent circuit model cannot decouple the coupling effect of multiple variables, resulting in an increase in the error of water content inversion based on empirical formulas.

[0005] In summary, membrane moisture content needs to be characterized through indirect measurement methods. Current characterization methods are ineffective in exploring the low-damage boundary under variable operating conditions. Therefore, an effective membrane moisture content characterization scheme is proposed to accurately map membrane moisture content to variable cold start operating conditions, providing a key technical guarantee for accurate membrane moisture content characterization under variable cold start operating conditions. Summary of the Invention

[0006] In order to make up for the deficiencies of the prior art, the embodiments of the present application propose a fuel cell membrane water content characterization system and method to solve the problems existing in the prior art.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] A fuel cell membrane water content characterization system, comprising:

[0009] A hydrogen supply device is used to provide hydrogen with the required purity to the fuel cell module to meet the hydrogen demand of the fuel cell electrochemical reaction;

[0010] An air supply device, used to provide the fuel cell module with the required air, which contains the oxygen required for the reaction;

[0011] The hydrogen circuit solenoid valve is used to control the on / off of the hydrogen circuit and realize the on / off control of the hydrogen supply;

[0012] Air circuit solenoid valve, used to control the on / off of the air circuit, to achieve on / off control of the air supply;

[0013] Hydrogen gas line humidifier, used to humidify hydrogen;

[0014] Air path humidifier, used to humidify the air;

[0015] The hydrogen gas mass flow controller is used to precisely control the flow of hydrogen to ensure that hydrogen is supplied to the fuel cell module at a set flow rate;

[0016] Air path mass flow controller, used to accurately control the air flow rate to ensure that the air is supplied to the fuel cell module at a set flow rate;

[0017] The gas temperature range control device is used to adjust the gas temperature by pre-cooling or preheating the gas, thereby accurately controlling the gas dew point temperature to change the relative humidity of the gas to meet the gas humidity requirements under different working conditions;

[0018] An impedance testing device, used to perform impedance testing on a fuel cell module and obtain impedance data;

[0019] Constant temperature and humidity chamber, used to provide a stable temperature and humidity environment for the fuel cell module, ensuring that the test is carried out under the set operating conditions;

[0020] Hydrogen line back pressure control valve, used to adjust the hydrogen line pressure;

[0021] Air line back pressure control valve, used to accurately control the air line back pressure;

[0022] The hydrogen supply device is connected to the hydrogen path humidifier through a hydrogen path solenoid valve, the hydrogen path humidifier is connected to the gas temperature range control device through a hydrogen path mass flow controller, the air supply device is connected to the air path humidifier through an air path solenoid valve, the air path humidifier is connected to the gas temperature range control device through an air path mass flow controller, the gas temperature range control device is also connected to a fuel cell, and the fuel cell is also respectively connected to an impedance testing device, a hydrogen path back pressure control valve and an air path back pressure control valve.

[0023] As a further technical solution of the present invention: the gas temperature range control device controls the gas dew point temperature by adjusting the gas temperature, thereby achieving precise adjustment of the relative humidity of the gas. The temperature adjustment range of the gas temperature range control device covers the entire operating temperature range of cold start.

[0024] As a further technical solution of the present invention: the impedance testing device can perform real-time impedance testing on the fuel cell module under variable operating conditions to obtain impedance data under different operating conditions. The test frequency range of the impedance testing device can cover the characteristic frequency band of the fuel cell membrane electrode process.

[0025] A method for characterizing the water content of a fuel cell membrane comprises the following steps:

[0026] S1. Conduct impedance test. Use the test platform to conduct cold start impedance test under different working conditions. During the impedance data collection process under different working conditions, the impedance data collection is completed by judging the relaxation change stage and adjusting the sampling frequency.

[0027] S2. Conduct a water content calibration test, using a balanced purge test and impedance upper and lower limit tests to establish a variable operating condition impedance-water content data set;

[0028] S3. Fitting and prediction of variable operating condition impedance-water content data based on Gaussian process regression. The established operating condition-impedance data set and operating condition-water content data set are fitted and predicted using the Gaussian process regression model. The model is verified and modified through experiments to establish a variable operating condition high reliability data set.

[0029] S4. Research on impedance normalization method based on SVM-RFE algorithm, explore the correlation between various factors and impedance and water content in the impedance-water content data set, establish a strong correlation data set, and complete the establishment of impedance normalization method.

[0030] As a further technical solution of the present invention: Step S1 specifically includes:

[0031] S11. Use single-factor impedance testing to explore the influence of different factors on impedance and divide the sensitive range;

[0032] S12. Explore the influence of each factor on impedance through orthogonal experiments with multi-factor coupling;

[0033] S13. Based on the above two tests, representative impedance test operating points are selected, and the operating condition-impedance data set is established through impedance testing.

[0034] As a further technical solution of the present invention: Step S2 specifically includes:

[0035] S21. Select multiple sets of data within the cold start variable operating condition range and perform impedance upper and lower boundary tests on the platform. First, perform dry gas purge tests under different operating conditions to obtain the upper impedance boundary under different operating conditions. Then, perform load tests under different operating conditions to obtain the lower impedance boundary under different operating conditions.

[0036] S22. Conduct balanced purge tests under different operating conditions on the platform, maintaining constant parameters such as flow rate, pressure, and temperature. Conduct balanced purge tests at different humidity levels. Combine the empirical formula for relative humidity and moisture content of the purge gas to calibrate the moisture content under each operating condition, and use this as the baseline value for moisture content under each operating condition.

[0037] S23. Based on the above test results, a working condition-water content data set is established.

[0038] As a further technical solution of the present invention: Step S3 specifically includes:

[0039] S31. Determine the value boundaries and precision of the Gaussian process regression independent variables based on the test analysis results and the cold start operating range. Establish a Gaussian process regression impedance training data set through impedance testing. Combined with the water content calibration test results, establish a variable operating condition impedance-water content training set for the Gaussian process regression model.

[0040] S32. Fit the training data using a Gaussian process regression model, and repeatedly correct the model's fit using methods such as kernel function optimization, hyperparameter optimization, and ensemble learning;

[0041] S33. Using a high-fitting model, a prediction is performed on a variable operating range and a high-precision data set to obtain an impedance-water content mapping relationship within the variable operating range;

[0042] S34. In addition, multiple sets of operating condition data are selected for impedance testing and water content calibration tests, and the test and predicted impedance errors are compared and analyzed. Through repeated verification and control of error accuracy, the verification and correction of the variable operating condition data set are completed, and finally a high-precision variable operating condition impedance-water content data set is obtained.

[0043] As a further technical solution of the present invention: Step S4 specifically includes:

[0044] S41. Use the SVM algorithm to evaluate the contribution of each feature in the data set to determine the weight of each feature;

[0045] S42. Use the RFE algorithm to perform feature screening on the "impedance-water content" data set and eliminate features with less influence. The judgment condition for smaller features is: minimum feature weight / previous level feature weight < 0.01. This is done until the features with smaller influence are completely eliminated. A strong correlation and high-precision variable working condition impedance-water content data set is established, and finally an impedance normalization method is established.

[0046] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0047] The present invention builds an impedance testing platform, conducts comprehensive impedance testing experiments and water content calibration experiments, and combines the Gaussian process regression model and SVM-RFE algorithm to establish a variable operating condition "impedance-water content" mapping relationship and an impedance normalization method. It can effectively deal with the characterization problem of membrane water content under variable operating conditions and realize the accurate mapping of membrane water content and variable operating conditions.

[0048] The method of the present invention takes into account the coupling influence of multiple factors, improves the accuracy and reliability of membrane water content characterization, and provides strong technical support for the study of fuel cell cold start durability damage.

[0049] The system of the present invention has a reasonable structure and is easy to operate, can realize variable operating condition characterization of the water content of the fuel cell membrane, and has high practical value and promotion and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a schematic diagram of the system of the present invention;

[0051] Figure 2 Flowchart of the method for characterizing the water content of fuel cell membranes under variable operating conditions. DETAILED DESCRIPTION

[0052] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] like Figure 1As shown, a fuel cell membrane moisture content characterization system and method includes a hydrogen supply device 1, an air supply device 2, a hydrogen circuit solenoid valve 3, an air circuit solenoid valve 4, a hydrogen circuit humidifier 5, an air circuit humidifier 6, a hydrogen circuit mass flow controller 7, an air circuit mass flow controller 8, a gas temperature range control device 9, an impedance testing device 10, a constant temperature and humidity chamber 11, a fuel cell module 12, a hydrogen circuit backpressure control device 13, and an air circuit backpressure control device 14. The hydrogen supply device 1 can provide the fuel cell module 12 with hydrogen of the required purity, meeting the hydrogen demand of the fuel cell electrochemical reaction. The air supply device 2 can provide the required air for the fuel cell module 12, which contains the oxygen required for the reaction. The hydrogen path solenoid valve 3 can control the on-off of the hydrogen path to achieve on-off control of the hydrogen supply. The air path solenoid valve 4 can control the on-off of the air path to achieve on-off control of the air supply. The hydrogen path humidifier 5 can humidify the hydrogen. The air path humidifier 6 can humidify the air. The hydrogen path mass flow controller 7 can accurately control the flow of hydrogen to ensure that hydrogen is supplied to the fuel cell module 12 at a set flow rate. The air path mass flow controller 8 can accurately control the flow of air to ensure that the air It is supplied to the fuel cell module 12 at a set flow rate. The gas temperature range control device 9 can adjust the temperature of the gas by pre-cooling or preheating the gas, and then accurately control the dew point temperature of the gas to change the relative humidity of the gas to meet the requirements of gas humidity under different working conditions. The impedance testing device 10 can perform impedance testing on the fuel cell module 12 and obtain impedance data. The constant temperature and humidity chamber 11 can provide a stable temperature and humidity environment for the fuel cell module 12 to ensure that the test is carried out under the set working conditions. The hydrogen path back pressure control device 13 is used to adjust the hydrogen path pressure, and the air path back pressure control device 14 is used to accurately control the air path back pressure.

[0054] like Figure 2 As shown, the present invention is specifically used for characterizing the water content of fuel cell membranes under cold start variable operating conditions. Taking a certain type of fuel cell as an example, the characterization of membrane water content under variable operating conditions is performed:

[0055] S1: Construction of multi-dimensional impedance characteristic test system;

[0056] Single-factor, full-operation impedance test matrix: Based on an equivalent single-cell model, impedance characterization testing was conducted using the fuel cell membrane moisture characterization system of claim 1. The test conditions encompassed temperature gradients, relative humidity ranges, pressure ranges, gas flow rates, and current load ranges, encompassing the entire cold start range. During the test, a constant temperature and humidity chamber (11) was used to control ambient temperature, a gas temperature range control device (9) was used to adjust purge gas humidity, and hydrogen and air mass flow controllers (8 and 9) were used to precisely control gas flow.

[0057] Quantitative analysis of sensitive ranges: By using the impedance test device 10 to test impedance under various operating conditions, response curves between each factor and impedance are plotted to determine the sensitive range of each factor. Taking the temperature factor as an example, within the -20°C to 80°C range, the impedance decays exponentially with increasing temperature. In the low-temperature range of -10°C to 20°C, the impedance decreases by 15%-20% for every 5°C increase in temperature, indicating a high sensitivity range. In the high-temperature range above 60°C, the impedance change rate drops below 5%, indicating a low sensitivity range.

[0058] Construction of the operating condition-impedance dataset: Based on the single-factor sensitivity intervals and orthogonal test results, we used the Latin hypercube sampling method to select n groups of representative operating conditions. These included cold start at low temperatures, normal temperature operation, and high temperature conditions with varying humidity, intake pressure, intake flow, and current. Each operating condition was tested three times, and the average values ​​were taken to construct a two-dimensional dataset containing the operating condition parameters (temperature, humidity, pressure, flow, and current) and impedance values.

[0059] S2: High-precision water content calibration and mapping relationship establishment;

[0060] Impedance boundary test: The upper boundary test uses high-purity dry hydrogen provided by hydrogen supply device 1 and dry air provided by air supply device 2 for purging. The purge flow and pressure are maintained for 30 minutes at each operating point. After the impedance value stabilizes, the upper boundary value is recorded. The lower boundary test uses a step-by-step current loading method, starting at 10% of the rated current and increasing by 10% each time. When the impedance change rate is less than 0.5% after two consecutive current loadings, the lower boundary value of the impedance is recorded. For example, under operating conditions of -20°C and 50% humidity, the upper boundary of the impedance is 125mΩ after dry gas purging. When the load is increased to 400A, the impedance stabilizes at 82mΩ, thus determining the boundary range.

[0061] Balanced purge test: The gas dew point temperature is adjusted using a gas temperature range control device to achieve humidity conditions corresponding to each temperature. For example, using an 80°C battery temperature as an example, the dew point temperature is controlled at 31.7°C (corresponding to 10% humidity) and 44.7°C (corresponding to 20% humidity). Purge is continued under each humidity condition until the impedance value stabilizes. High-frequency impedance is simultaneously collected using an impedance test device, establishing a multi-dimensional calibration system.

[0062] Data integration: Integrate the balanced purge test data under various operating conditions, smooth the data, and construct a four-dimensional data set containing operating parameters (temperature, humidity, pressure, flow) and water content.

[0063] S3: Gaussian process regression modeling and dynamic prediction;

[0064] Constructing a multidimensional independent variable space: Based on the sensitive intervals determined by S1 and S2, the independent variables of Gaussian process regression (GPR) were set to temperature, humidity, pressure, flow rate, and current (with a precision of 0.1%), forming a five-dimensional input space. A KD tree algorithm was used to partition the input space, ensuring that each leaf node contained at least 20 sets of training data to avoid the curse of dimensionality.

[0065] GPR model optimization framework: A linear combination of the squared exponential kernel (SE) and the Matern kernel (ν=2.5) is used, with weight coefficients determined by maximizing the marginal likelihood function. This combination improves the fit of the temperature-humidity coupling effect by 12%. The expected improvement (EI) acquisition function is used to iteratively optimize the hyperparameter space (length scale, signal variance, noise variance) to reduce the root mean square error (RMSE).

[0066] Prediction verification and iterative correction: Use the fuel cell membrane water content characterization system of claim 1 to verify the prediction results, calculate the error between the test and the prediction results, and when the error of a certain set of data is greater than 5%, automatically add the data to the training set, re-optimize the kernel function and hyperparameters until the accuracy requirements of the engineering application are met, and finally establish a high-precision variable operating condition "impedance-water content" prediction model and a cold start full operating condition "impedance-water content" data set.

[0067] S4: Impedance normalization method driven by SVM-RFE;

[0068] Quantitative evaluation of feature weights: The SVM-C-SVC model is used to perform weight analysis on the five-dimensional features (temperature, humidity, pressure, flow, and current). The Leave-One-Out (LOO) cross-validation score of each feature is calculated to obtain a standardized weight vector.

[0069] Implementation of recursive feature elimination algorithm: The RFE algorithm is used to screen the features of the full-condition impedance dataset. According to the criterion of "minimum feature weight / previous level feature weight < 0.01", the features with little impact on the impedance are recursively and iteratively eliminated until all features with strong correlation and high accuracy to the impedance are completely retained, thus constructing a high-quality full-condition impedance dataset.

[0070] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations that come within the meaning and range of equivalents of the claims be embraced therein.

[0071] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment have also been appropriately combined to form other implementation methods that are easy for those skilled in the art to understand.

Claims

1. A fuel cell membrane water content characterization system, characterized in that: include: A hydrogen supply device is used to provide hydrogen with the required purity to the fuel cell module to meet the hydrogen demand of the fuel cell electrochemical reaction; An air supply device, used to provide the fuel cell module with the required air, which contains the oxygen required for the reaction; The hydrogen circuit solenoid valve is used to control the on / off of the hydrogen circuit and realize the on / off control of the hydrogen supply; Air circuit solenoid valve, used to control the on / off of the air circuit, to achieve on / off control of the air supply; Hydrogen gas line humidifier, used to humidify hydrogen; Air path humidifier, used to humidify the air; The hydrogen gas mass flow controller is used to precisely control the flow of hydrogen to ensure that hydrogen is supplied to the fuel cell module at a set flow rate; Air path mass flow controller, used to accurately control the air flow rate to ensure that the air is supplied to the fuel cell module at a set flow rate; The gas temperature range control device is used to adjust the gas temperature by pre-cooling or preheating the gas, thereby accurately controlling the gas dew point temperature to change the relative humidity of the gas to meet the gas humidity requirements under different working conditions; An impedance testing device, used to perform impedance testing on a fuel cell module and obtain impedance data; Constant temperature and humidity chamber, used to provide a stable temperature and humidity environment for the fuel cell module, ensuring that the test is carried out under the set operating conditions; Hydrogen line back pressure control valve, used to adjust the hydrogen line pressure; Air line back pressure control valve, used to accurately control the air line back pressure; The hydrogen supply device is connected to the hydrogen path humidifier through a hydrogen path solenoid valve, the hydrogen path humidifier is connected to the gas temperature range control device through a hydrogen path mass flow controller, the air supply device is connected to the air path humidifier through an air path solenoid valve, the air path humidifier is connected to the gas temperature range control device through an air path mass flow controller, the gas temperature range control device is also connected to a fuel cell, and the fuel cell is also respectively connected to an impedance testing device, a hydrogen path back pressure control valve and an air path back pressure control valve.

2. A fuel cell membrane water content characterization system according to claim 1, characterized in that: The gas temperature range control device controls the gas dew point temperature by adjusting the gas temperature, thereby achieving precise adjustment of the relative humidity of the gas. The temperature adjustment range of the gas temperature range control device covers the entire operating temperature range of cold start.

3. A fuel cell membrane water content characterization system according to claim 1, characterized in that: The impedance testing device can perform real-time impedance testing on the fuel cell module under variable operating conditions to obtain impedance data under different operating conditions. The test frequency range of the impedance testing device can cover the characteristic frequency band of the fuel cell membrane electrode process.

4. A method for characterizing the water content of a fuel cell membrane based on the system according to any one of claims 1 to 3, characterized in that: The following steps are involved: S1. Conduct impedance test. Use the test platform to conduct cold start impedance test under different working conditions. During the impedance data collection process under different working conditions, the impedance data collection is completed by judging the relaxation change stage and adjusting the sampling frequency. S2. Conduct a water content calibration test, using a balanced purge test and impedance upper and lower limit tests to establish a variable operating condition impedance-water content data set; S3. Fitting and prediction of variable operating condition impedance-water content data based on Gaussian process regression. The established operating condition-impedance data set and operating condition-water content data set are fitted and predicted using the Gaussian process regression model. The model is verified and modified through experiments to establish a variable operating condition high reliability data set. S4. Research on impedance normalization method based on SVM-RFE algorithm, explore the correlation between various factors and impedance and water content in the impedance-water content data set, establish a strong correlation data set, and complete the establishment of impedance normalization method.

5. The method for characterizing the water content of a fuel cell membrane according to claim 4, wherein: Step S1 specifically includes: S11. Use single-factor impedance testing to explore the influence of different factors on impedance and divide the sensitive range; S12. Use orthogonal experiments with multiple factors to explore the influence of each factor on impedance; S13. Based on the above two tests, representative impedance test operating points are selected, and the operating condition-impedance data set is established through impedance testing.

6. The method for characterizing the water content of a fuel cell membrane according to claim 4, wherein: Step S2 specifically includes: S21. Select multiple sets of data within the cold start variable operating condition range and perform impedance upper and lower boundary tests on the platform. First, perform dry gas purge tests under different operating conditions to obtain the upper impedance boundary under different operating conditions. Then, perform load tests under different operating conditions to obtain the lower impedance boundary under different operating conditions. S22. Conduct balanced purge tests under different operating conditions on the platform, maintaining constant parameters such as flow rate, pressure, and temperature. Conduct balanced purge tests at different humidity levels. Combine the empirical formula for relative humidity and moisture content of the purge gas to calibrate the moisture content under each operating condition, and use this as the baseline value for moisture content under each operating condition. S23. Based on the above test results, a working condition-water content data set is established.

7. The method for characterizing the water content of a fuel cell membrane according to claim 4, wherein: Step S3 specifically includes: S31. Determine the value boundaries and precision of the Gaussian process regression independent variables based on the test analysis results and the cold start operating range. Establish a Gaussian process regression impedance training data set through impedance testing. Combined with the water content calibration test results, establish a variable operating condition impedance-water content training set for the Gaussian process regression model. S32. Fit the training data using a Gaussian process regression model, and repeatedly correct the model's fit using methods such as kernel function optimization, hyperparameter optimization, and ensemble learning; S33. Using a high-fitting model, a prediction is performed on a variable operating range and a high-precision data set to obtain an impedance-water content mapping relationship within the variable operating range; S34. In addition, multiple sets of operating condition data are selected for impedance testing and water content calibration tests, and the test and predicted impedance errors are compared and analyzed. Through repeated verification and control of error accuracy, the verification and correction of the variable operating condition data set are completed, and finally a high-precision variable operating condition impedance-water content data set is obtained.

8. The method for characterizing the water content of a fuel cell membrane according to claim 4, wherein: Step S4 specifically includes: S41. Use the SVM algorithm to evaluate the contribution of each feature in the data set to determine the weight of each feature; S42. Use the RFE algorithm to perform feature screening on the "impedance-water content" dataset and eliminate features with less influence. The judgment condition for smaller features is: minimum feature weight / previous level feature weight < 0.

01. This process is repeated until all features with smaller influence are eliminated. A strong correlation and high-precision variable working condition impedance-water content dataset is established, and finally an impedance normalization method is established.

Images