Accelerated life test method for fuel cell

Abstract

The invention discloses a fuel cell accelerated life test method. Comprising the following steps: firstly, under a preset service life test working condition, collecting performance characterization data of different preset service life stages in a fuel cell service life test process; then, according to the performance characterization data of different preset life stages, calculating to obtain difference curves corresponding to different performance characterization data, so as to obtain all difference curves of the current fuel cell; then, extracting preset target life stage features corresponding to each difference curve, and establishing a feature set; and finally, inputting the feature set into a life prediction model, and outputting a prediction result of a corresponding life test by the model. The time cost and the economic cost consumed by the fuel cell life test can be remarkably reduced, and the research and development efficiency of the fuel cell is improved.

Description

Technical Field

[0001] The invention relates to a fuel cell life test method in the field of batteries, and in particular to a fuel cell accelerated life test method and system. Background Art

[0002] Life testing (also known as durability testing) plays an important role in fuel cell R&D and is commonly used in the R&D of new materials, new structures, new control strategies, etc. Life testing is the most expensive of all fuel cell R&D test items, including economic and time costs. Accelerating the speed of life testing can significantly reduce the R&D cost of fuel cells and improve R&D efficiency.

[0003] At present, fuel cells are mainly accelerated by designing equivalent accelerated test conditions to test their lifespan. According to the equivalent accelerated life test results of fuel cells released by Toyota, the life attenuation of a fuel cell running for 75 hours under laboratory accelerated test conditions is equivalent to running for 6000 hours on an actual vehicle, and the equivalent acceleration ratio can reach 80 times. In addition to using equivalent accelerated test conditions, life test time can also be shortened through life prediction technology. In the field of lithium batteries, some researchers have used life prediction technology to reduce the life test that originally took more than 500 days to complete to 16 days, greatly reducing the R&D and testing costs of lithium battery fast charging strategies and improving R&D efficiency. However, an accelerated life test method based on life prediction has not yet been proposed in the fuel cell field. Summary of the invention

[0004] In order to solve the problem of high cost and low efficiency of life test in the process of fuel cell research and development, the present invention proposes a fuel cell accelerated life test method and system. The present invention reduces the economic cost and time cost of fuel cell life test and improves the research and development efficiency of fuel cells.

[0005] The technical solution adopted by the present invention is:

[0006] 1) Under the preset life test conditions, collect performance characterization data of different preset life stages during the fuel cell life test;

[0007] 2) According to the performance characterization data of different preset life stages, the difference curves corresponding to different performance characterization data are calculated, thereby obtaining all the difference curves of the current fuel cell;

[0008] 3) Extracting the preset target life stage characteristics corresponding to each difference curve and establishing a feature set;

[0009] 4) The feature set is input into the life prediction model, and the model outputs the corresponding prediction results of the life test.

[0010] The preset life stage includes a preset test time value and a preset life index value. The life index of the fuel cell includes but is not limited to the limiting current (I lim ), total cathode oxygen transport resistance (R o2，total ), electrochemical active area (ECSA), hydrogen permeation current (I cross ) and the output voltage under a specific current load (V out )wait.

[0011] The performance characterization data of the fuel cell include, but are not limited to, electrochemical impedance spectroscopy (EIS), polarization (I / V) curves, cyclic voltammetry (CV) curves, and linear sweep voltammetry (LSV) curve data.

[0012] The different preset life stages are generated through expert experience, filtering method, packaging method or embedding method.

[0013] The 2) is specifically:

[0014] After standardizing all performance characterization data of different preset life stages respectively, the standardized performance characterization data are obtained; for each type of standardized performance characterization data, the performance characterization data of the later preset life stage is subtracted from the performance characterization data of the earlier preset life stage to obtain the corresponding difference curve; after traversing and processing all performance characterization data of all different preset life stages, all difference curves of the current fuel cell are obtained, such as the difference curves corresponding to the EIS, I / V curve, CV curve and LSV curve are the EIS difference curve, I / V difference curve, CV difference curve and LSV difference curve respectively.

[0015] In the above 3), the preset target life stage characteristics corresponding to each difference curve include one or more of instantaneous characteristics, statistical characteristics and model parameter characteristics, that is, it can be a single feature or a combination of characteristics. Therefore, the feature set composed of the preset target life stage characteristics corresponding to different difference curves can include a combination of the same type of features and a combination of different types of features.

[0016] In the above 3), when the preset target life stage is a preset test time value, the preset target life stage characteristics are characteristics related to the life index value corresponding to the preset test time value; when the preset target life stage is a preset life index value, the preset target life stage characteristics are characteristics related to the test time value corresponding to the preset life index value.

[0017] In the above 4), when the preset target life stage is the preset test time value, the life prediction model used is the life index prediction model, and the life index prediction model outputs the life index value when the fuel cell reaches the preset test time value, and obtains the prediction result of the life test; when the preset target life stage is the preset life index value, the life prediction model used is the test time prediction model, and the test time prediction model outputs the test time value required for the fuel cell to reach the preset life index value, and obtains the prediction result of the life test.

[0018] The life prediction model is an integration of one or more machine learning models, empirical models or semi-empirical models.

[0019] 2. A fuel cell accelerated life test system

[0020] A performance characterization data acquisition unit, used to acquire performance characterization data at different preset life stages during a fuel cell life test;

[0021] A difference curve generating unit, used for generating difference curve data of corresponding performance characterization data according to different performance characterization data obtained at different preset life stages;

[0022] A feature extraction unit, used to extract preset target life stage features from the calculated difference curve data and establish a feature set;

[0023] The test result generating unit is used to input the feature set into the life prediction model, and the model outputs the predicted value of the corresponding life test result.

[0024] 3. A computer device

[0025] The device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the steps of the fuel cell accelerated life test method are implemented.

[0026] 4. A computer-readable storage medium

[0027] The medium stores a computer program, and when the computer program is executed by a processor, the steps of the fuel cell accelerated life test method are implemented.

[0028] The beneficial effects of the present invention are:

[0029] The method proposed in the present invention does not require a complete fuel cell life test, but only requires a small amount of preliminary testing to accurately predict the life index value of the fuel cell when it reaches the preset test time or the test time required to reach the preset life index value. Compared with the full life test, it can greatly save time and economic costs and significantly improve the research and development efficiency of fuel cells. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is the overall flow chart of the present invention.

[0031] Figure 2 Schematic diagram of the spline function fitting and linear interpolation results of I / V curve data in an embodiment of the present invention.

[0032] Figure 3 It is a schematic diagram of the spline function fitting and linear interpolation results of CV curve data in an embodiment of the present invention.

[0033] Figure 4 Schematic diagram of a method for calculating an EIS imaginary impedance difference curve in an embodiment of the present invention.

[0034] Figure 5 Schematic diagram of the calculation results of the EIS imaginary impedance difference curve in an embodiment of the present invention.

[0035] Figure 6 Schematic diagram of a method for calculating an I / V difference curve in an embodiment of the present invention.

[0036] Figure 7 Schematic diagram of calculation results of the I / V difference curve in an embodiment of the present invention.

[0037] Figure 8 Schematic diagram of the calculation method of the CV difference curve in the embodiment of the present invention.

[0038] Fig. 9 Schematic diagram of the calculation results of the CV difference curve in the embodiment of the present invention.

[0039] Fig.10 It is the 30000th life test condition in the embodiment of the present invention. lim The prediction result diagram.

[0040] Fig.11 is the 30000th life test condition in the embodiment of the present invention. o2，total The prediction result diagram.

[0041] Fig.12 It is a graph showing the predicted results of ECSA at the 30,000th life test condition in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] Example 1

[0044] The preset life stage of this embodiment is a preset test time value, which is generated through expert experience, including the test time of the 0th life test condition (0s) and the test time of the 1000th life test condition (6500s). Figure 1 As shown, the specific process of the accelerated life test method is as follows:

[0045] 1) Under the preset life test conditions, collect performance characterization data at different preset test times during the fuel cell life test. The preset life test conditions have the same test cycle (6.5s) and different upper potential limits (UPL) and inlet gas humidity. The residence time at high potential (0.95V or 1.15V) and low potential (0.6V) is 3s, and the switching time between high potential and low potential is 0.25s. The life indicator uses the limiting current (I lim ), total cathode oxygen transport resistance (R o2,total ) and electrochemical active area (ECSA). The performance of fuel cells is characterized by polarization (I / V) curves, electrochemical impedance spectroscopy (EIS), and cyclic voltammetry (CV) curves.

[0046] 2) According to the performance characterization data collected at different preset test times, the difference curves of different performance characterization data are calculated. The performance characterization data collected at different preset test times are uniformly standardized, and the difference curve is obtained by subtracting the performance characterization data collected at the previous preset test time from the performance characterization data collected at the later preset test time. The difference curves of different performance characterization data are also different. The difference curves corresponding to the I / V curve, EIS, and CV curve are the I / V difference curve, the EIS imaginary impedance difference curve, and the CV difference curve, respectively. After traversing and processing all the performance characterization data collected at all different preset test times, all the difference curves of the current fuel cell are obtained.

[0047] In this embodiment, the performance characterization data is fitted using a spline function and linear interpolation is performed. For the I / V curve data, the voltage is fitted as a function of the current density using a spline function, and the current density is linearly interpolated. The current density value range is [0mA, 3.1mA], and the linear interpolation interval of the current density is 0.16mA. Figure 2 As shown. For the CV curve data, the current density is fitted as a function of voltage using a spline function, and the voltage is linearly interpolated. The voltage value range is [0.051V, 0.4V], and the linear interpolation interval of the voltage is 0.003V, as shown in Figure 3As shown. For EIS data, since the frequency points of all EIS data are set to the same during acquisition, there is no need for unified standardization. The EIS imaginary impedance difference curve is calculated by subtracting the EIS imaginary impedance data at the 1000th life test condition (6500s) from the EIS imaginary impedance data at the 0th life test condition (0s), as shown in Figure 4 The calculation results of the EIS imaginary impedance difference curve are shown in Figure 5 The I / V difference curve is calculated by subtracting the I / V curve data at the 0th life test condition (0s) from the I / V curve data at the 1000th life test condition (6500s), as shown in Figure 6 The calculation results of the I / V difference curve are shown in Figure 7 The CV difference curve is calculated by subtracting the CV curve data at the 1000th life test condition (6500s) from the CV curve data at the 0th life test condition (0s), as shown in Figure 8 The calculation results of the CV difference curve are shown in Fig. 9 shown.

[0048] 3) Extract the preset target life stage features corresponding to each difference curve and establish a feature set; specifically, the features extracted from the I / V difference curve, the EIS imaginary impedance difference curve and the CV difference curve are all single statistical features, and are the life index I corresponding to the 30000th life test condition (125h) at the preset target test time. lim , R o2，total Features that are respectively related to the values ​​of ECSA.

[0049] The method for extracting a single statistical feature on the I / V difference curve is:

[0050] (1) Subtract the voltage values ​​corresponding to any two current density values ​​on the I / V difference curve and take the absolute value as the candidate two-point feature; (2) Traverse all combinations of two current density values ​​on the I / V difference curve, and a total of (n 2 -n) / 2 candidate two-point features, where n is the number of current density values ​​on the difference curve and n=20; (3) Calculate the life index I corresponding to each candidate two-point feature and the fuel cell at the 30000th life test condition (125h) lim The Pearson correlation coefficient between the values ​​is used to select the candidate two-point features with the largest absolute value of the correlation coefficient as the best two-point features of the I / V difference curve.

[0051] The method for extracting a single statistical feature on the EIS imaginary impedance difference curve is:

[0052] (1) Subtract the imaginary impedance values ​​corresponding to any two frequency values ​​on the EIS imaginary impedance difference curve, and take the absolute value as the candidate two-point feature; (2) Traverse all combinations of two frequency values ​​on the EIS imaginary impedance difference curve, and a total of (n 2 -n) / 2 candidate two-point features, where n is the number of frequency values ​​on the difference curve and n=41; (3) Calculate the life index R corresponding to each candidate two-point feature and the fuel cell life test condition (125h) of the 30000th life test condition o2，total The Pearson correlation coefficient between the values ​​is calculated, and the candidate two-point features with the largest absolute value of the correlation coefficient are selected as the best two-point features of the EIS imaginary impedance difference curve.

[0053] The method for extracting a single statistical feature on the CV difference curve is:

[0054] (1) Subtract the current density values ​​corresponding to any two voltage values ​​on the CV difference curve and take the absolute value as the candidate two-point feature; (2) Traverse all combinations of two voltage values ​​on the CV difference curve, and a total of (n 2 -n) / 2 candidate two-point features, where n is the number of voltage values ​​on the difference curve and n=100; (3) calculating the Pearson correlation coefficient between each candidate two-point feature and the corresponding life indicator ECSA value at the 30000th life test condition (125h) of the fuel cell, and selecting the candidate two-point feature with the largest absolute value of the correlation coefficient as the best two-point feature of the CV difference curve.

[0055] 4) Input the feature set into the life prediction model, and the model outputs the corresponding prediction result of the life test. The specific life prediction model used is the life index prediction model, which includes I lim Prediction Model, R o2，total The prediction model and ECSA prediction model are both random forest regression models. lim The input of the prediction model is the best two-point feature of the I / V difference curve, and the output is the I lim The value of Fig.10 As shown. o2，total The input of the prediction model is the best two-point feature of the EIS imaginary impedance difference curve, and the output is the R o2，total The value of Fig.11 The input of the ECSA prediction model is the best two-point feature of the CV difference curve, and the output is the ECSA value when the 30,000th test cycle (125h) is reached, as shown in Fig.12 shown.

[0056] Example 2

[0057] The preset life stage of this embodiment is a preset life index value, which is generated through expert experience and includes I lim The value when attenuation is 0% and I lim The value when the attenuation is 5%, R o2，total The value when the growth is 0% and the value when the growth is 5%, the value when the ECSA decays by 0% and the value when the ECSA decays by 5%. Figure 1 As shown, the specific process of the accelerated life test method is as follows:

[0058] 1) Under the preset life test conditions, collect performance characterization data when the fuel cell reaches different preset life index values ​​during the life test. The life index is I lim , R o2，total The performance characterization data of fuel cells are I / V curve, EIS, and CV curve.

[0059] 2) According to the performance characterization data collected when different preset life index values ​​are reached, the difference curves of different performance characterization data are calculated. The performance characterization data collected when different preset life index values ​​are reached are uniformly standardized, and the difference curve is obtained by subtracting the performance characterization data collected when the later preset life index value is subtracted from the performance characterization data collected when the previous preset life index value is reached. The difference curves of different performance characterization data are also different. The difference curves corresponding to the I / V curve, EIS, and CV curve are the I / V difference curve, the EIS imaginary impedance difference curve, and the CV difference curve, respectively. After traversing and processing all the performance characterization data collected for all different preset life index values, all the difference curves of the current fuel cell are obtained.

[0060] In this embodiment, the performance characterization data is fitted using a spline function and linear interpolation is performed. For the I / V curve data, the voltage is fitted as a function of the current density using a spline function, and the current density is linearly interpolated. The current density value range is [0mA, 3.1mA], and the linear interpolation interval of the current density is 0.16mA. Figure 2 As shown. For the CV curve data, the current density is fitted as a function of voltage using a spline function, and the voltage is linearly interpolated. The voltage value range is [0.051V, 0.4V], and the linear interpolation interval of the voltage is 0.003V, as shown in Figure 3 As shown in Figure 2. For EIS data, since the frequency points of all EIS data are set to the same during acquisition, there is no need for unified standardization. The EIS imaginary impedance difference curve is obtained using R o2，total EIS imaginary impedance data at 5% increase minus R o2，total The I / V difference curve is calculated using the EIS imaginary impedance data when the growth rate is 0%.lim I / V curve data at 5% attenuation minus I lim The I / V curve data at 0% attenuation was calculated. The CV difference curve was calculated by subtracting the CV curve data at 0% attenuation from the CV curve data at 5% attenuation of the ECSA.

[0061] 3) Extract the preset target life stage features corresponding to each difference curve and establish a feature set; specifically, the features extracted from the I / V difference curve, the EIS imaginary impedance difference curve and the CV difference curve are all single statistical features. The features extracted from the I / V difference curve are the values ​​corresponding to the preset target life index (I lim The value of attenuation of 50%) corresponds to the characteristics related to the test time value. The value extracted from the EIS difference curve is related to the preset target life index value (R o2，total The features extracted from the CV difference curve are related to the test time value corresponding to the preset target life indicator value (the value of ECSA attenuation of 50%).

[0062] The method for extracting a single statistical feature on the I / V difference curve is:

[0063] (1) Subtract the voltage values ​​corresponding to any two current density values ​​on the I / V difference curve and take the absolute value as the candidate two-point feature; (2) Traverse all combinations of two current density values ​​on the I / V difference curve, and a total of (n 2 -n) / 2 candidate two-point features, where n is the number of current density values ​​on the difference curve and n=20; (3) Calculate the correlation between each candidate two-point feature and the fuel cell I lim The Pearson correlation coefficient between the corresponding test time values ​​when the attenuation is 50%, and the candidate two-point features with the largest absolute value of the correlation coefficient are selected as the best two-point features of the I / V difference curve.

[0064] The method for extracting a single statistical feature on the EIS imaginary impedance difference curve is:

[0065] (1) Subtract the imaginary impedance values ​​corresponding to any two frequency values ​​on the EIS imaginary impedance difference curve, and take the absolute value as the candidate two-point feature; (2) Traverse all combinations of two frequency values ​​on the EIS imaginary impedance difference curve, and a total of (n 2 -n) / 2 candidate two-point features, where n is the number of frequency values ​​on the difference curve and n=41; (3) Calculate the correlation between each candidate two-point feature and the fuel cell R o2，total The Pearson correlation coefficient between the test time values ​​corresponding to the 50% increase is used, and the candidate two-point features with the largest absolute value of the correlation coefficient are selected as the best two-point features of the EIS imaginary impedance difference curve.

[0066] The method for extracting a single statistical feature on the CV difference curve is:

[0067] (1) Subtract the current density values ​​corresponding to any two voltage values ​​on the CV difference curve and take the absolute value as the candidate two-point feature; (2) Traverse all combinations of two voltage values ​​on the CV difference curve, and a total of (n 2 -n) / 2 candidate two-point features, where n is the number of voltage values ​​on the difference curve and n=100; (3) calculating the Pearson correlation coefficient between each candidate two-point feature and the corresponding test time value when the fuel cell ECSA decays by 50%, and selecting the candidate two-point feature with the largest absolute value of the correlation coefficient as the best two-point feature of the CV difference curve.

[0068] 4) Input the feature set into the life prediction model, and the model outputs the corresponding prediction result of the life test. The specific life prediction model used is the test time prediction model, which includes I lim Test time prediction model with 50% decay, R o2，total The test time prediction model with a 50% increase and the test time prediction model with a 50% decrease in ECSA are both based on the random forest regression model. lim The test time prediction model for attenuation of 50%. The input of the model is the best two-point feature of the I / V difference curve, and the output of the model is I lim The test time value corresponding to the attenuation reaching 50%. o2，total The test time prediction model with a 50% increase in test time is shown in Figure 1. The input of the model is the best two-point feature of the EIS imaginary impedance difference curve, and the output of the model is R o2，total The test time value when the growth reaches 50%. For the test time prediction model with 50% ECSA decay, the input of the model is the best two-point feature of the CV difference curve, and the output of the model is the test time value when the ECSA decay reaches 50%.

[0069] Finally, it should be noted that the above embodiments and explanations are only used to illustrate the technical solution of the present invention rather than to limit it. Those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope disclosed in the technical solution of the present invention, which should be included in the scope of protection of the claims of the present invention.

Claims

1. A fuel cell accelerated life test method, characterized in that: The following steps are involved: 1) Under the preset life test conditions, collect performance characterization data of different preset life stages during the fuel cell life test; 2) According to the performance characterization data of different preset life stages, the difference curves corresponding to different performance characterization data are calculated, thereby obtaining all the difference curves of the current fuel cell; 3) Extracting the preset target life stage characteristics corresponding to each difference curve and establishing a feature set; 4) The feature set is input into the life prediction model, and the model outputs the corresponding prediction results of the life test.

2. A fuel cell accelerated life test method according to claim 1, characterized in that: The preset life stage includes a preset test time value and a preset life index value.

3. A fuel cell accelerated life test method according to claim 1, characterized in that: The 2) is specifically: All performance characterization data of different preset life stages are standardized to obtain standardized performance characterization data; for each type of standardized performance characterization data, the performance characterization data of the later preset life stage is subtracted from the performance characterization data of the earlier preset life stage to obtain the corresponding difference curve; after traversing and processing all performance characterization data of all different preset life stages, all difference curves of the current fuel cell are obtained.

4. A fuel cell accelerated life test method according to claim 1, characterized in that: In the above 3), the preset target life stage characteristics corresponding to each difference curve include one or more of instantaneous characteristics, statistical characteristics and model parameter characteristics.

5. A fuel cell accelerated life test method according to claim 1, characterized in that: In the above 3), when the preset target life stage is a preset test time value, the preset target life stage feature is a feature related to the life index value corresponding to the preset test time value; When the preset target life stage is a preset life index value, the preset target life stage characteristic is a characteristic related to the test time value corresponding to the preset life index value.

6. A fuel cell accelerated life test method according to claim 1, characterized in that: In the above 4), when the preset target life stage is the preset test time value, the life prediction model used is the life index prediction model, and the life index prediction model outputs the life index value when the fuel cell reaches the preset test time value, and obtains the prediction result of the life test; when the preset target life stage is the preset life index value, the life prediction model used is the test time prediction model, and the test time prediction model outputs the test time value required for the fuel cell to reach the preset life index value, and obtains the prediction result of the life test.

7. A fuel cell accelerated life test method according to claim 1, characterized in that: The different preset life stages are generated through expert experience, filtering method, packaging method or embedding method.

8. A fuel cell accelerated life test system, characterized in that: include: A performance characterization data acquisition unit, used to acquire performance characterization data at different preset life stages during a fuel cell life test; A difference curve generating unit, used for generating difference curve data of corresponding performance characterization data according to different performance characterization data obtained at different preset life stages; A feature extraction unit, used to extract preset target life stage features from the calculated difference curve data and establish a feature set; The test result generating unit is used to input the feature set into the life prediction model, and the model outputs the predicted value of the corresponding life test result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the fuel cell accelerated life test method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the fuel cell accelerated life test method described in any one of claims 1 to 7 are implemented.

Images