Fuel cell impedance spectrum rapid measurement system and measurement method

Abstract

The invention belongs to the related technical field of fuel cells, and particularly relates to a fuel cell impedance spectroscopy rapid measurement system and method, the system is divided into an excitation current injection part, a data acquisition part and an upper computer, the upper computer adopts a DIBS signal with an excellent peak value inhibition capability and a flexible amplitude-frequency design capability as test excitation, and the test excitation part is connected with the data acquisition part; an electronic load is used as an excitation source. The upper computer can control the electronic load to generate a current excitation waveform used for impedance testing based on a DIBS signal, the electronic load injects excitation current into output current of a tested galvanic pile, the upper computer can also read sampling data of the data acquisition part, fast Fourier transform is carried out on the acquired excitation current and response voltage data, and the impedance of the tested galvanic pile is tested. And obtaining an electrochemical impedance spectrum. According to the method, the impedance measurement efficiency is improved, the hardware structure is simple, the method has a good field application prospect, a measured object can be expanded to other electrochemical systems including an electrolytic tank, and the method has good expansibility.

Description

Technical Field

[0001] The present invention belongs to the technical field related to fuel cells, and more specifically, relates to a fast measurement system and method for a fuel cell impedance spectrum. Background Art

[0002] Electrochemical impedance spectroscopy (EIS) measurement technology can characterize the internal state and electrochemical behavior information of fuel cells in a non-invasive manner, and is a common method for fuel cell stack fault diagnosis and health assessment. However, existing EIS detection mainly relies on electrochemical workstations, which "sweep" the wide-band impedance information of the electrochemical system by applying a single-frequency sinusoidal signal to the fuel cell one by one. Although this method can achieve high impedance measurement accuracy, its disadvantage is that the measurement time is very long and it is difficult to obtain real-time information of the electrochemical system, which seriously limits its function for real-time diagnosis and on-site monitoring.

[0003] Toyota's Miral product has developed a DC / DC-based stack impedance measurement module. However, due to the bandwidth limitation of the DC / DC converter, the excitation signal frequency generated by the DC / DC converter is often difficult to reach a high value, thus affecting the measurement range of the impedance spectrum. Ulleberg, F. Zenith, T. Holm. Multi-sine EIS for early detection of PEMFC failure modes. Frontiers in Energy Research, 2022, 10: 855985] Using multi-sine signals as test excitation, the complete EIS with a frequency range of 0.5 Hz to 50 kHz is collected and analyzed within 50 seconds. The test time is difficult to meet the requirements of real-time fault diagnosis.

[0004] Chinese patent CN 109459465 A uses two longest linear shift register sequence (M sequence) signals to excite the high-frequency region and low-frequency region of EIS respectively, which can shorten the injection of excitation signal to 16 seconds and finally shorten the measurement time to 21 seconds. However, it is still not enough for the application of EIS technology in real-time fault diagnosis and needs to be further shortened. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a fuel cell impedance spectrum rapid measurement system and measurement method, the purpose of which is to improve the efficiency of impedance measurement, shorten the impedance spectrum measurement time, and expand the application scenarios of impedance analysis.

[0006] To achieve the above object, according to one aspect of the present invention, a fuel cell impedance spectrum rapid measurement system is provided, comprising: a first electronic load, a second electronic load, a data acquisition unit, and a host computer;

[0007] The first electronic load is connected in series with the fuel cell stack to be tested to form a main loop, which is used to work in a constant current mode to control the static operating point of the fuel cell stack to be tested; the second electronic load is connected in parallel at both ends of the fuel cell stack to be tested to form a secondary loop, which is used to generate a desired current excitation waveform and inject it into the output current of the fuel cell stack to be tested under the control instruction corresponding to the multi-frequency excitation signal designed by the host computer; the multi-frequency excitation signal is a wide-band discrete interval binary sequence signal;

[0008] The data acquisition unit is used to collect the response voltage of the fuel cell stack to be tested to the current excitation waveform, and to sample the excitation current generated by the second electronic load; the host computer is also used to generate an electrochemical impedance spectrum of the fuel cell stack to be tested based on the response voltage and the sampled excitation current.

[0009] Furthermore, the host computer is also used to analyze the noise waveform of the fuel cell stack voltage to be tested under the stable state of the system based on the voltage across the fuel cell stack collected by the data acquisition unit under the stable state of the system, and perform power spectrum analysis on the noise waveform of the stack voltage to generate a multi-frequency excitation signal with an amplitude that meets the signal-to-noise ratio requirements.

[0010] Furthermore, when generating a multi-frequency excitation signal, the host computer uses an AR model to perform a power spectrum estimation operation of the stack voltage noise, wherein the optimal order of the AR model is determined based on an AIC criterion, and parameters other than the order of the AR model are extracted by a Burg algorithm.

[0011] Furthermore, the optimal order of the AR model is 128.

[0012] Furthermore, the working mode of the second electronic load is a list mode. When the host computer controls the second electronic load, it controls the second electronic load through a programmable instrument standard command set according to a designed multi-frequency excitation signal to modulate an excitation current waveform for impedance testing.

[0013] Furthermore, the data acquisition unit includes a voltage-dividing resistor and a data acquisition card; the voltage-dividing resistor is connected in parallel at both ends of the fuel cell stack to be tested to form a voltage-dividing loop to meet the measurement voltage range of the data acquisition card; the data acquisition card is used to read the voltage at both ends of the voltage-dividing resistor and sample the excitation current generated by the second electronic load;

[0014] The host computer is also used to analyze the voltage across the fuel cell stack to be tested according to the voltage across the voltage divider resistor, and further subtract the DC component therefrom to obtain the response voltage of the current excitation waveform.

[0015] Furthermore, the host computer communicates with the second electronic load via USB to achieve transmission of multi-frequency excitation signal control instructions; the data acquisition card communicates with the host computer via Ethernet to achieve transmission of voltage across the voltage divider resistor and excitation current.

[0016] Furthermore, the voltage-dividing resistor is divided into two, and the resistance ratio is 10:1; the data acquisition card acquires the voltage across the two ends of the voltage-dividing resistor with a smaller resistance.

[0017] According to another aspect of the present invention, a method for rapid measurement of a fuel cell impedance spectrum is provided, and the method uses a rapid measurement system for a fuel cell impedance spectrum as described above for measurement.

[0018] In general, compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:

[0019] 1. The present invention proposes a rapid measurement system for the impedance spectrum of a fuel cell, which is divided into an excitation current injection part, a data acquisition part and a host computer. The host computer uses a DIBS signal with excellent peak suppression capability and flexible amplitude-frequency design capability as a test excitation, and uses a common electronic load as an excitation source. The host computer can control the electronic load to generate a desired current excitation waveform based on the test excitation, and the electronic load injects the excitation current into the output current of the electrochemical system under test. The host computer can also read the sampling data of the data acquisition unit, perform a fast Fourier transform on the collected excitation current and response voltage data, and obtain an electrochemical impedance spectrum. The measurement system of the present invention can obtain the broadband impedance information of 0.5 to 1kHz of the fuel cell in one measurement by injecting an alternating current with a duration of only 8.2s, which greatly improves the efficiency of impedance measurement. In addition, it has the advantages of simple hardware structure and has good prospects for field application. In addition, the measurement range of the measurement system can also be extended to other electrochemical systems including lithium batteries, electrolytic cells, etc., with good adaptability and scalability.

[0020] 2. The present invention proposes a rapid measurement system for a fuel cell impedance spectrum. The excitation signal is designed by a host computer based on the power spectrum estimation of the system noise under steady-state conditions, ensuring sufficient frequency component amplitude to reduce noise interference; at the same time, the overall peak value is limited to maintain the linear characteristics of the tested system and avoid strong nonlinear distortion, thereby achieving a more accurate fuel cell impedance spectrum in a single test.

[0021] 3. The present invention also proposes that the host computer uses the AR model to perform the voltage noise power spectrum estimation operation when generating a multi-frequency excitation signal, and the computational complexity is low. Among them, the optimal order of the AR model is determined by the sampling AIC criterion, which can balance the complexity and goodness of fit of the model, so as to select a model that is neither too simple nor too complex, and the AIC criterion is relatively simple, not affected by the specific distribution of the data or the sample size, and has objectivity and consistency. Therefore, the present invention uses the AIC criterion to determine the optimal order of the AR model; in addition, the parameters other than the order of the AR model are extracted by the Burg algorithm. The Burg algorithm solves the AR model parameters based on the recursive process of the data sequence, without taking the autocorrelation function with a large amount of calculation, and has a high frequency resolution, which can achieve high-precision excitation. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a schematic diagram of a fuel cell impedance spectrum rapid measurement system provided by an embodiment of the present invention;

[0023] Figure 2 It is the fluctuation of the output voltage of the fuel cell stack and the power spectrum estimation for the noise provided by the embodiment of the present invention; wherein (a) and (b) are the voltage noise signals when the stack is not started and when it is running at a constant current of 10A, respectively, (c) is the AIC index under different orders, and (d) is the power spectrum obtained by estimating the voltage noise in (b) using the periodogram method and AR models of different orders;

[0024] Figure 3 It is the time domain and frequency domain distribution of the low-frequency and high-frequency signals provided by the embodiment of the present invention; wherein (a) is the designed DIBS signal, (b) is the actual current data obtained by sampling and the corresponding frequency domain distribution, and (c) is the actual stack voltage data obtained by sampling and the corresponding frequency domain distribution;

[0025] Figure 4 1 is the calculation result and analysis of the impedance spectrum provided by the embodiment of the present invention; wherein (a) and (b) are the impedance spectra obtained by repeated measurements when the constant current is 10A and 8A respectively, and (c) and (d) are the deviation and mean square error of repeated measurements when the constant current is 10A and 8A respectively;

[0026] Figure 5 is a diagram of a filtering process for invalid data points in rough impedance data provided by an embodiment of the present invention;

[0027] Figure 6 is used Figure 5 The impedance spectrum of the second-order R-CPE equivalent circuit fitting model obtained by the corresponding impedance data processing method;

[0028] Figure 7This is a specific result diagram of a fuel cell impedance spectrum rapid measurement system provided by an embodiment of the present invention;

[0029] Figure 8 It is an operation flow chart of the fast impedance testing system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0031] Embodiment 1

[0032] A fast measurement system for impedance spectrum of fuel cell, such as Figure 1 As shown, it includes: a first electronic load, a second electronic load, a data acquisition unit, and a host computer;

[0033] The first electronic load is connected in series with the fuel cell stack to be tested to form a main circuit, which is used to work in a constant current mode to control the static operating point of the fuel cell stack to be tested; the second electronic load is connected in parallel at both ends of the fuel cell stack to be tested to form a secondary circuit, which is used to generate a desired current excitation waveform and inject it into the output current of the fuel cell stack to be tested under the control instruction corresponding to the multi-frequency excitation signal designed by the host computer; the multi-frequency excitation signal is a wide-band discrete interval binary sequence signal; the fuel cell to be tested is a proton exchange membrane fuel cell (PEMFC).

[0034] The data acquisition unit is used to collect the response voltage of the fuel cell stack to the current excitation waveform and to sample the excitation current generated by the second electronic load; the host computer is also used to generate the electrochemical impedance spectrum of the fuel cell to be tested based on the response voltage and the sampled excitation current.

[0035] The multi-frequency excitation signal of this embodiment adopts a wide-band discrete-interval binary sequence (DIBS) signal with excellent peak suppression capability and flexible amplitude-frequency design capability.

[0036] As a preferred implementation mode, the host computer is also used to analyze the noise waveform of the fuel cell stack voltage to be tested under the stable state of the system according to the voltage across the fuel cell stack collected by the data acquisition unit under the stable state of the system, and perform power spectrum analysis on the noise waveform of the stack voltage to generate a multi-frequency excitation signal with an amplitude that meets the signal-to-noise ratio requirement.

[0037] Regarding the multi-frequency excitation signal whose amplitude meets the signal-to-noise ratio requirement, it should be noted that the multi-frequency excitation signal here contains multiple frequency components, which can significantly improve the test efficiency. For proton exchange membrane fuel cells, the impedance spectrum at 0.5Hz-1000Hz can reflect sufficient impedance information, so the frequency interval 0.5Hz-1000Hz is selected as the measurement range of the electrochemical impedance spectrum. In addition, since PEMFC is a typical nonlinear time-varying system, it is easier to ensure that the system is in a linear working range during the measurement process by using a smaller disturbance amplitude, avoiding strong nonlinear distortion, but this may cause a decrease in the signal-to-noise ratio (SNR), making it difficult to distinguish the interference caused by noise, and an amplitude that is too high may destroy the quasi-linear requirements for the system, and even affect the normal operation of the system under test. Therefore, the excitation signal needs to have a suitable frequency component amplitude. In this embodiment, the frequency component amplitude is designed to be within 3% of the current amplitude in a stable state.

[0038] As a preferred implementation, the signal can be designed according to the voltage noise power spectrum estimation to ensure sufficient frequency component amplitude to reduce the interference of noise, while limiting the overall peak value to maintain the linear characteristics of the tested system. The voltage noise power spectrum estimation is performed using a 128-order AR model, the AIC criterion is used to determine the optimal order of the AR model, and the Burg algorithm is selected for model parameter extraction.

[0039] That is, the excitation current signal of the test is dynamically generated according to the frequency range of the measurement requirements and the power spectrum estimation of the voltage noise, so as to minimize the disturbance to the system while ensuring the measurement accuracy. Regarding the power spectrum estimation of voltage noise, it is mainly divided into classical power spectrum estimation and parameter model power spectrum estimation. The periodogram method and the autocorrelation method are two representative methods widely used in classical power spectrum estimation. The parameter model power spectrum estimation mainly includes the AR model method, the MA model method and the ARMA model method. The parameter estimation of the AR model is based on linear equations, which makes the calculation relatively simple and direct. In contrast, the parameter estimation of the MA model and the ARMA model involves solving nonlinear equations, and the calculation is relatively complex. Moreover, it can be seen from Wold's theorem that any ARMA model or MA model can be represented by an infinite-order AR model. This means that even if an unsuitable model is selected in practical applications, a reasonable approximation can still be obtained by using a high-order AR model. Therefore, the parameter model selects the AR model to estimate the power spectrum of the electrochemical noise of the battery. The first step is to determine the model order and then solve the corresponding model parameters. The specific steps are as follows:

[0040] (1) The selection of the order of the AR model has an important impact on the accuracy of the power spectrum estimation. An AR model with too low an order may lead to inaccurate power spectrum estimation and omission of important frequency components; while an AR model with too high an order may introduce false spectral peaks, distort the power spectrum estimation, and increase computational complexity. The methods for determining the order of the AR model mainly include the autocorrelation and partial correlation function method, the statistical hypothesis testing method, and the information criterion method. The AIC (Akaike Information Criterion) determines the order of the AR model to balance the complexity and goodness of fit of the model, thereby selecting a model that is neither too simple nor too complex. It is relatively simple, not affected by the specific distribution of the data or the sample size, and has objectivity and consistency. Therefore, preferably, this embodiment uses the AIC criterion to determine the optimal order of the AR model.

[0041] (2) For the parameter extraction of the AR model, the Burg algorithm is preferred here because it solves the AR model parameters based on the recursive process of the data sequence, does not require the use of the autocorrelation function with a large amount of calculation, and has a higher frequency resolution.

[0042] Furthermore, in order to analyze the fluctuation of the output voltage of the fuel cell stack when no alternating current excitation is injected (working at a constant current), the stack is set to run at a constant current of 10 A, and an acquisition card is used to sample the stack terminal voltage. The sampling frequency is set to 2048 Hz and the sampling time is 16 s. Figure 2 (a) and (b) in the figure show the sampling signals of the voltage noise when the battery stack is not started and when it is running at a constant current of 10A. It can be seen that the amplitude of the voltage fluctuation when the battery stack is running is several orders of magnitude higher than the background noise of the measuring instrument. Considering that the order of the AR model has a direct impact on the accuracy and computational efficiency of the Burg algorithm in spectrum estimation, the Burg algorithm is used to fit the AR model of voltage fluctuation from 1 to 256 orders, and the AIC index at different orders is obtained, as shown in Figure 1. Figure 2 As shown in (c) in Figure 1, we can see that the trend of AIC decreasing with the increase of order gradually slows down, and AIC tends to be stable after the order exceeds 150. The periodogram method and AR models of different orders (32, 64, 128, 192) are used to estimate Figure 2 (b) The power spectrum of the voltage noise in Figure 2 As shown in (d) in the figure, it can be seen that there are several peaks in the voltage fluctuation when the battery stack is running at a constant current and the power spectrum of the measured background noise. Their base frequencies are 15Hz, 50Hz, etc., which are likely to be the noise introduced by the operation of electronic loads and other equipment. These frequency points should be avoided when designing the amplitude-frequency distribution of the impedance test excitation signal. Figure 2(d) in it also shows that for the power spectrum estimation of the voltage fluctuation of the battery stack operation, the AR model method can obtain a smoother curve than the periodogram method, which means that a smoother amplitude-frequency distribution can be obtained when designing the test excitation current according to the noise power and the required signal-to-noise ratio. Since DIBS is only an approximation of the expected amplitude-frequency distribution, there is always a certain deviation between the amplitude-frequency of the generated sequence signal and the expected value, and a smoother amplitude-frequency distribution design can make the actual amplitude-frequency line of the signal closer to the expected value. Taking the above into consideration, this embodiment preferably uses a 128-order AR model to estimate the power spectrum of voltage noise.

[0043] After the test excitation current signal is generated, the host computer sends a current control command to the electronic load for current injection, and at the same time starts the current sampling channel of the data acquisition card and synchronizes it with the voltage sampling channel at both ends of the previously opened voltage divider resistor. Since it takes a certain amount of time for the electronic load to respond to the instruction, the sampling time of the acquisition card can be set to be slightly longer than the duration of the excitation current to ensure that the complete excitation and response signals can be sampled. After the test is completed, the host computer determines the moment when the excitation starts in the sampled data according to the preset excitation signal, and intercepts the sampled data according to the designed excitation signal duration. The next step is to calculate the impedance of the system under test at the corresponding test frequency point based on the current excitation and voltage response signals.

[0044] In an implementation of an embodiment, the first electronic load adopts IT9715 electronic load, the second sampling load adopts DL3021 electronic load, and the data acquisition unit is realized by a data acquisition card and a voltage-dividing resistor; the IT9715 electronic load is connected in series with the fuel cell stack to be tested to form a main circuit, which is used to work in a constant current mode to control the static working point of the fuel cell stack to be tested; the DL3021 electronic load is connected in parallel at both ends of the fuel cell stack to be tested to form a secondary circuit, which is used to generate a desired current excitation waveform and inject it into the output current of the fuel cell stack to be tested under the control instruction corresponding to the multi-frequency excitation signal of the host computer; the multi-frequency excitation signal is a wide-band discrete interval binary sequence signal; the voltage-dividing resistor is connected in parallel at both ends of the fuel cell stack to be tested to form a voltage-dividing circuit to meet the measurement voltage range of the data acquisition card; the data acquisition card is used to read the voltage at both ends of the voltage-dividing resistor and sample the excitation current generated by the DL3021 electronic load; the host computer is also used to analyze the voltage at both ends of the fuel cell stack to be tested according to the voltage at both ends of the voltage-dividing resistor as a response voltage to the current excitation waveform, and generate the electrochemical impedance spectrum of the fuel cell stack to be tested in combination with the excitation current sampled by the data acquisition card.

[0045] As a preferred implementation, the DL3021 electronic load works in list mode, the maximum number of steps that can be executed in list mode is 512, and the current value and duration of each step can be set separately. The host computer uses the Standard Commands for Programmable Instruments (SCPI) (i.e., SCPI instructions) to control the DL3021 electronic load to execute the current waveform designed by the host computer according to the programming manual provided by the manufacturer, so as to modulate the excitation current for impedance testing and inject it into the output current of the battery stack.

[0046] In addition, considering the acquisition range of the data acquisition card, the current size in DL3021 can be converted for acquisition. The acquisition card is used to collect the voltage across the resistor with a smaller resistance, and the response voltage of the fuel cell stack is calculated according to the corresponding proportion. Since it takes a certain amount of time for the electronic load to respond to the instruction, the sampling time of the acquisition card is set to be slightly longer than the duration of the excitation current to ensure that the complete excitation and response signals can be sampled.

[0047] Preferably, there are two voltage-dividing resistors with a resistance ratio of 10:1; the data acquisition card acquires the voltage across the two ends of the voltage-dividing resistor with a smaller resistance.

[0048] The output voltage of the fuel cell stack is usually tens of volts, which is greater than the sampling range of the data acquisition card. Two ohm resistors with a resistance of 10:1 are used in parallel with the stack to form a voltage divider circuit to meet the measurement voltage range of the acquisition card.

[0049] Preferably, the host computer communicates with the DL3021 electronic load via USB to transmit the excitation signal control instruction; the data acquisition card communicates with the host computer via Ethernet (LAN) to transmit the voltage across the voltage divider resistor and the sampled excitation current.

[0050] In one example of this embodiment, the VK701NW acquisition card is a 4-channel Ethernet high-speed synchronous acquisition card produced by Shenzhen Micro Fine Art, with 4-channel true differential input, 24-bit resolution to ensure measurement accuracy, a maximum sampling rate of 100k, and a maximum voltage input range of ±10 V. The communication with the data acquisition card is based on the TCP protocol, and the input range, sampling channel, sampling time and sampling frequency of the acquisition card are configured according to the dynamic library function (.dll) provided by the manufacturer to realize the corresponding data sampling function.

[0051] In practical applications, the host computer can be developed using Python in the PySide environment. Its program design includes two main tasks: communicating with the instrument and processing the collected data.

[0052] In one example, the host computer determines the start time of the excitation in the sampled data according to the preset excitation signal, and intercepts the sampled data according to the designed excitation signal duration. Then, the collected excitation current and response voltage data are fast Fourier transformed to obtain the transformed current FI(w) and the transformed voltage FU(w), and then the following formula is used to extract the broadband impedance under the desired harmonic component:

[0053]

[0054] Where n is the expected number of harmonics in the designed broadband signal, FU = F[U(t)], FU = F[I(t)], Z k The impedance test point corresponding to the kth expected harmonic component is used, and the corresponding impedance data processing method is used to improve the robustness of impedance spectrum measurement.

[0055] In order to verify the feasibility of the impedance test system built in this embodiment and the accuracy of EIS measurement, a repeated test experiment was designed. That is, the PEMFC stack is set to work in the same steady-state condition, and its EIS is repeatedly measured many times, so that the discrete degree of the measurement results reflects the accuracy of the measurement. For the EOS1000 stack used in the experiment, the two electronic loads are first adjusted to make the main loop current 10A and the secondary loop current 0.74A. The reason for setting the DC bias in the secondary loop is that the electronic load cannot generate negative current, and the minimum value of the alternating current excitation must be greater than 0. The electronic load in the secondary loop is then used to modulate the excitation current for the impedance test. The DIBS signal used in the test is divided into two sections, low frequency and high frequency, with durations of 8s and 0.2s respectively. The designed frequency distribution is 24 frequency points from 0.5 to 1000Hz, and is distributed in equal proportion.

[0056] Figure 3 (a) shows the design value of the DIBS signal, which is less than 3% of the steady-state operating current of the fuel cell. Similarly, the main loop current was adjusted to 8A for the same test. Figure 3 (b) and (c) in the figure are respectively the current and voltage data and the corresponding frequency domain distribution obtained by injecting two sections of DIBS test excitation into the PEMFC stack. Due to the effective bandwidth of the electronic load and the noise introduced in the modulation current process, the actual current excitation is somewhat distorted compared to the design value, but the results of frequency analysis show that the harmonic distribution of the excitation current is roughly consistent with the design value. In addition, the frequency components corresponding to the response voltage and the excitation current also have a high resolution, indicating that the test has a sufficient signal-to-noise ratio.

[0057] Since the test frequency is basically consistent with the design value, the amplitude and phase of the corresponding harmonic components can be directly extracted from the Fourier transform of the sampled current and voltage signals according to the design frequency point to calculate the impedance spectrum. Under the 10A and 8A working conditions, the measurement was repeated 13 times continuously. The measurement results are as follows: Figure 4 The impedance spectrum of the fuel cell at this operating point can be represented by the average value of these 13 tests. The average curve shows that the shape of the impedance spectrum is mainly determined by the frequency points below 260Hz, while the impedance points at higher frequencies are only slightly different from each other. Figure 4 (c) and (d) in the figure are visual analyses of the discreteness of repeated impedance spectrum measurements. The left ordinate axis is the deviation of a single measurement point relative to the average value, and the right ordinate axis is the mean square error of repeated measurements at each frequency point. It can be seen from the figure that the discreteness of the test points in the low-frequency region is generally greater than that in the high-frequency region. It is particularly noteworthy that at several frequency points between 10 and 100 Hz, the mean square error of the test points is significantly higher than that of the surrounding adjacent frequency points, indicating that there may be abnormal noise interference at these frequency points.

[0058] Furthermore, under ideal conditions, the impedance spectrum is the same when the fuel cell is in the same working state. However, it is impossible to maintain a completely stable working state of the fuel cell in actual measurements. In addition, the instruments used in the measurement will also introduce random noise, so the impedance data obtained from the experimental measurement will inevitably deviate. Keeping the fuel cell operating parameters the same, the impedance spectrum is measured repeatedly, and the impedance data obtained from multiple measurements are plotted on the same graph. The degree of discreteness of the test points can intuitively reflect the interference to the measurement. In actual applications, due to limited test time, of course, the test will not be repeated dozens of times, but efforts will be made to obtain sufficiently accurate impedance spectrum information in one test. The realization of this goal requires corresponding data processing methods. The processing of the original impedance test points includes removing free bad points and correcting some effective deviation points. The specific steps are as follows:

[0059] 1) Bad point filtering: The KK test can be approximated by fitting the RC equivalent circuit model and is only applicable to impedance data with less noise. In actual testing, when shortening the test time and constraining the amplitude of the test excitation signal, the measurement signal-to-noise ratio is often not very high, which means that the impedance data obtained are mostly rough. Considering that the ideal impedance spectrum is usually a very smooth curve, the continuity criterion can be used to identify bad points in the test impedance data. In order to characterize the continuity between impedance points, the Savitzky-Golay filter is first used to obtain a smooth impedance interpolation point Z smooth (w). The SG filter is a filtering method based on the local polynomial least squares fitting method. Each interpolation point consists of N adjacent SGThe data points are obtained using a second-order polynomial estimate. Here is the representation of the real part of the smoothed impedance point, and the calculation of the imaginary part is similar:

[0060] Z Smooth，re (ω j )=a re，1 +a re，2 ln(ω j )+a re，3 ln 2 (ω j )j=1,2,...,n

[0061] Where n is the number of data points. Select N SG =4, coefficient a of the interpolation polynomial re =[a re,1 a re,2 a re,3 ] T Determined by least squares regression

[0062] a re =(X T X) -1 X T Z re

[0063]

[0064]

[0065] is with ω j Four adjacent impedance data points, and Discontinuities at data points are defined as the difference from the smoothly interpolated points:

[0066] ε(ω j )=|Z(ω j )-Z smooth (ω j )|

[0067] To avoid filtering out "normal" data points, only one data point with the largest discontinuity is removed at a time. After a data point is removed, the discontinuity of its nearby points is recalculated, excluding the removed point. The data point with the largest discontinuity is iteratively removed until the overall discontinuity is The drop rate is less than the set threshold, thus achieving the removal of free bad pixels.

[0068]

[0069] Where it represents the number of iterations. Figure 5The filtering process of invalid data points in rough impedance data is demonstrated, and the continuity of data points is gradually enhanced as the bad points are removed.

[0070] 2) The deviation point correction is achieved by assigning different weight values ​​to the data points. The KK test can be used to obtain the fitting residual and weighted mean square error of each point:

[0071]

[0072] Weight adjustment is to modify w(ω j ) value, so that ε is less than the set threshold First, the recommended adjustment value w of the weight is calculated based on the relative "significance" of the KK test residuals s (ω j ):

[0073]

[0074] And through high pass filtering h q =[-0.25 0.5 -0.25] T Avoid deviating data points from affecting nearby normal points. s (ω j ) is updated iteratively to the next KK test:

[0075] w it (ω j ) -1 =ζ·(w s (ω j )) -2 +(1-ζ)·(max(w min ,w it-1 (ω j ))) -1

[0076] In the formula, w it (ω j ) indicates that the frequency after the it-th iteration is ω j The weight value of the impedance point, ζ = 0.5, w min =10 -6 is the lower limit to prevent division by zero.

[0077] In addition, because the fitting effect of the RC equivalent circuit model on the experimental data is unstable, the RC fitting model used in the deviation point correction link is changed to the distribution of relaxation times (DRT) model. The DRT method regards the electrochemical system as a series connection of an ohmic resistor and an infinite number of polarization processes, avoiding the uncertainty introduced by the prior assumptions of the model and being able to approximate the impedance model of any electrochemical system. The DRT analysis is based on the following integral formula:

[0078]

[0079] Among them, Z DRT (ω) is the complex impedance calculated by the DRT model at angular frequency ω, R inf is the series resistance that characterizes a process with negligible relaxation time (infinite characteristic frequency), and τ is the relaxation time. The core problem of the DRT method is to use Z DRT Model fitting experimentally measured impedance Z exp (w), deconvolution is performed to obtain the relaxation time distribution function γ(τ). The solution of γ(τ) is an inverse problem, which can be expressed as a linear combination of weight parameters and corresponding primitive functions, namely:

[0080]

[0081] Where N is the number of relaxation elements. The more the number, the higher the accuracy, but the more calculations are required. j is the weight parameter of the basis function for fitting estimation; φ j (lnτ) is the radial basis function.

[0082] Therefore, the solution of γ(τ) can be transformed into minimizing the fitting model Z DRT (ω) and test data Z exp (w) Error between:

[0083]

[0084] Due to the influence of noise in the experimental data, directly solving Loss often results in ill-posed optimization results (underfitting, overfitting, and difficulty in improving accuracy, etc.). For this reason, a regularization method is often used to add a penalty term to the equation to smooth the noisy data. The EISART software package developed by Li Hangyue uses a regularization calculation method, which can automatically select primitive functions and regularization parameters according to the data type, and can achieve high-precision DRT model fitting. The present invention selects to use the EISART open source tool to achieve equivalent EIS model fitting for impedance measurement data. EISART can not only find the fitting model of the data point, but also output the data point confidence reference in the range of 0 to 1. The measurement confidence reference can not only reflect the degree of deviation of the data point, but also be added to the operation of impedance spectrum feature extraction to improve the accuracy of feature parameter identification.

[0085] Furthermore, the impedance data processing method was applied to the PEMFC impedance spectrum under steady-state conditions and the impedance spectrum of the fitting model (2nd order R-CPE) obtained from each set of test data was as follows: Figure 6 Compared to Figure 4 For the original test points, the impedance spectra of each fitting model are quite close to the average value of the test points. If the average value of repeated measurements is regarded as the real impedance of the battery stack, then with the help of the impedance data processing method, even if some impedance test points are disturbed by noise and have a large deviation, a relatively accurate impedance spectrum can still be obtained through a single test.

[0086] Embodiment 2

[0087] A method for quickly measuring a fuel cell impedance spectrum is provided, and a fuel cell impedance spectrum quickly measuring system as described in the first embodiment is used for measurement.

[0088] In actual measurement, the following steps can be followed:

[0089] (1) Build an impedance spectrum fast test system and initialize each component; the fast impedance spectrum measurement system includes components such as fuel cell stack, IT9715 electronic load, DL3021 electronic load, data acquisition card, host computer and voltage divider resistor, etc., and its structure is as follows: Figure 7 shown.

[0090] (2) The host computer determines whether the system is in a stable state for impedance testing by real-time monitoring of the voltage of the fuel cell stack, and generates an excitation signal with appropriate amplitude and spectrum distribution based on the power spectrum analysis of the collected voltage waveform.

[0091] (3) The host computer uses USB to communicate with the DL3021 programmable electronic load, controls the electronic load to generate the desired current excitation waveform, and injects it into the output current of the electrochemical system under test. At the same time, the sampled data in the acquisition card cache is read through Ethernet (LAN) communication. Since it takes a certain amount of time for the electronic load to respond to instructions, the sampling time of the acquisition card is set to be slightly longer than the duration of the excitation current to ensure that the complete excitation and response signals can be sampled.

[0092] (4) According to the actual excitation signal generated by the electronic load, the excitation start time in the sampled data is determined, and the sampled data is intercepted according to the duration of the excitation signal. Then, the collected excitation current and response voltage data are fast Fourier transformed to obtain the electrochemical impedance spectrum, and the corresponding impedance data processing method is used to improve the robustness of the impedance spectrum measurement.

[0093] During the test, the IT9715 electronic load works in constant current mode to control the static working point of the battery stack. The fuel cell stack is air-cooled and self-humidifying. The hydrogen required for its operation is supplied by a high-pressure hydrogen cylinder that is depressurized to about 0.5 bar by a pressure reducing valve. The host computer is developed in Python in the PySide environment. Its program design includes two main tasks: communicating with the instrument and processing the collected data.

[0094] Figure 8 The operation process of the fast impedance test system is demonstrated. After the program enables the EIS measurement module, the voltage sampling channel of the acquisition card is started to continuously monitor the voltage of the fuel cell, and the power spectrum of the voltage noise is estimated to determine whether the stability conditions of the EIS test are met. When the operator manually issues a test request through the graphical user interface (GUI) and the voltage fluctuation is less than the set threshold (determined by the test signal peak constraint and the test signal-to-noise ratio), the module enters the test state. The test excitation current signal is dynamically generated according to the frequency range of the measurement requirements and the power spectrum estimation of the voltage noise, minimizing the disturbance to the system while ensuring the measurement accuracy.

[0095] In general, the present invention relates to a fuel cell impedance spectrum rapid measurement system and method. The constructed impedance spectrum rapid test system consists of an excitation current injection part, a data acquisition part and a host computer. A wideband discrete interval binary sequence (DIBS) signal is used as a test excitation. The host computer determines whether the system is in a stable state where impedance testing can be implemented by real-time monitoring of the voltage of the fuel cell stack, and generates an excitation signal with an amplitude that meets the signal-to-noise ratio requirement based on the power spectrum analysis of the collected voltage waveform. Then, a USB is used to communicate with a programmable electronic load to control the electronic load to generate a desired current excitation waveform and inject it into the output current of the electrochemical system under test. At the same time, the host computer reads the sampled data in the acquisition card cache through Ethernet (LAN) communication. The collected excitation current and response voltage data are subjected to fast Fourier transform to obtain an electrochemical impedance spectrum, and the corresponding impedance data processing method is used to improve the robustness of the impedance spectrum measurement. The present invention adopts DIBS signal with excellent peak suppression capability and flexible amplitude-frequency design capability as test excitation, uses common electronic load as excitation source, and injects alternating current with a duration of only 8.2s. The broadband impedance information of 0.5-1kHz of the battery can be obtained in one measurement, which greatly improves the efficiency of impedance measurement. In addition, the advantages of simple hardware structure and other advantages have good prospects for field application. In addition, the measurement range of the measurement system can also be extended to other electrochemical systems including lithium batteries, electrolytic cells, etc., with good adaptability and scalability.

[0096] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A fuel cell impedance spectrum rapid measurement system, characterized in that: include: A first electronic load, a second electronic load, a data acquisition unit, and a host computer; The first electronic load is connected in series with the fuel cell stack to be tested to form a main loop, which is used to work in a constant current mode to control the static operating point of the fuel cell stack to be tested; the second electronic load is connected in parallel at both ends of the fuel cell stack to be tested to form a secondary loop, which is used to generate a desired current excitation waveform and inject it into the output current of the fuel cell stack to be tested under the control instruction corresponding to the multi-frequency excitation signal designed by the host computer; the multi-frequency excitation signal is a wide-band discrete interval binary sequence signal; The data acquisition unit is used to collect the response voltage of the fuel cell stack to be tested to the current excitation waveform, and to sample the excitation current generated by the second electronic load; the host computer is also used to generate an electrochemical impedance spectrum of the fuel cell stack to be tested based on the response voltage and the sampled excitation current.

2. A fuel cell impedance spectrum rapid measurement system according to claim 1, characterized in that: The host computer is also used to analyze the noise waveform of the fuel cell stack voltage to be tested in the stable state of the system according to the voltage across the fuel cell stack collected by the data acquisition unit in the stable state of the system, and perform power spectrum analysis on the noise waveform of the stack voltage to generate a multi-frequency excitation signal with an amplitude that meets the signal-to-noise ratio requirement.

3. A fuel cell impedance spectrum rapid measurement system according to claim 2, characterized in that: When generating a multi-frequency excitation signal, the host computer uses an AR model to perform a power spectrum estimation operation of the stack voltage noise, wherein the optimal order of the AR model is determined based on an AIC criterion, and parameters other than the order of the AR model are extracted through a Burg algorithm.

4. A fuel cell impedance spectrum rapid measurement system according to claim 3, characterized in that: The optimal order of the AR model is 128.

5. A fuel cell impedance spectrum rapid measurement system according to claim 1, characterized in that: The working mode of the second electronic load is the list mode. When controlling the second electronic load, the host computer controls the second electronic load through a programmable instrument standard command set according to a designed multi-frequency excitation signal to modulate an excitation current waveform for impedance testing.

6. A fuel cell impedance spectrum rapid measurement system according to claim 1, characterized in that: The data acquisition unit includes a voltage-dividing resistor and a data acquisition card; the voltage-dividing resistor is connected in parallel at both ends of the fuel cell stack to be tested to form a voltage-dividing loop to meet the measurement voltage range of the data acquisition card; the data acquisition card is used to read the voltage at both ends of the voltage-dividing resistor and sample the excitation current generated by the second electronic load; The host computer is also used to analyze the voltage across the fuel cell stack to be tested according to the voltage across the voltage divider resistor, and further subtract the DC component therefrom to obtain the response voltage of the current excitation waveform.

7. A fuel cell impedance spectrum rapid measurement system according to claim 6, characterized in that: The host computer communicates with the second electronic load via USB to achieve the transmission of multi-frequency excitation signal control instructions; the data acquisition card communicates with the host computer via Ethernet to achieve the transmission of the voltage across the voltage divider resistor and the excitation current.

8. A fuel cell impedance spectrum rapid measurement system according to claim 6, characterized in that: The voltage-dividing resistor is divided into two, and the resistance ratio is 10:1; the data acquisition card acquires the voltage at both ends of the voltage-dividing resistor with a smaller resistance.

9. A method for rapid measurement of fuel cell impedance spectrum, characterized in that: The measurement is performed using a fuel cell impedance spectrum rapid measurement system as described in any one of claims 1 to 8.

Images