Fuel cell fault diagnosis method and device based on sensor signal and medium

Abstract

The invention discloses a fuel cell fault diagnosis method and device based on a sensor signal and a medium, belongs to the technical field of fuel cells, and is used for solving the technical problem that the diagnosis accuracy is difficult to improve due to the fact that existing fuel cell fault diagnosis is limited to information of the sensor signal in a two-dimensional space. The method comprises: acquiring a sensor signal of a fuel cell; performing time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image; feature points of the time-frequency spectrum image are obtained through feature extraction; the method comprises the following steps: acquiring a time-frequency spectrum image, copying the time-frequency spectrum image to obtain a first time-frequency spectrum image and a second time-frequency spectrum image, placing the first time-frequency spectrum image on a first plane, placing the second time-frequency spectrum image on a second plane, connecting mutually matched feature points to obtain a feature landscape, and determining the fault type of the fuel cell based on the feature landscape. According to the method, the limitation of observing the state information of the fuel cell from a low-dimensional angle is overcome, and the fault diagnosis accuracy of the fuel cell is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fuel cells, and in particular to a fuel cell fault diagnosis method, device and medium based on sensor signals. Background Art

[0002] Proton exchange membrane fuel cells (PEMFCs) are widely used in the transportation industry for their high efficiency and pollution-free operation. For example, PEMFCs can be used in the powertrains of automobiles and aircraft. The next generation of high power density fuel cells is expected to play a greater role in powertrain applications. However, the degradation of fuel cells under typical operating conditions makes it difficult to meet the transient power requirements of automobiles. In addition to the limitations of dynamic response capabilities, fuel cells also face challenges in durability. For example, corrosion currents at startup, temporary relative air shortages under load changes, and water state fluctuations at idle can significantly affect the service life of fuel cells. Therefore, the development of fast and accurate fault diagnosis methods is crucial to extend the service life of fuel cells because it can provide rich and effective status information for subsequent fault recovery measures.

[0003] In the prior art, the method for diagnosing the fault of a fuel cell based on the sensor signals such as the voltage signal and the current signal of the fuel cell generally diagnoses the fault of the fuel cell based on the one-dimensional sensor signal, or obtains a two-dimensional image through the sensor signal and diagnoses the fault of the fuel cell based on the two-dimensional image. That is, the exploration of sensor signals in the prior art is limited to two-dimensional space, and no high-dimensional space above two-dimensional space is explored. Due to the inherent limitations of low-dimensional spaces such as one-dimensional space and two-dimensional space, the amount of information about the working state of the fuel cell obtained in the low-dimensional space is limited. It is difficult to obtain more comprehensive and reasonable information about the state of the fuel cell through the one-dimensional sensor signal itself or the two-dimensional image obtained based on the sensor signal, which makes it difficult to improve the accuracy of fault diagnosis. Summary of the invention

[0004] The present invention provides a fuel cell fault diagnosis method, device and medium based on sensor signals, so as to solve the technical problem that the prior art performs fuel cell fault diagnosis based on one-dimensional sensor signals or two-dimensional images, is limited to two-dimensional space, and the status information obtained about the fuel cell is limited, making it difficult to improve the accuracy of fault diagnosis.

[0005] In a first aspect, the present invention provides a fuel cell fault diagnosis method based on a sensor signal, comprising:

[0006] Acquire sensor signals of the fuel cell;

[0007] Performing time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image of the sensor signal;

[0008] Obtaining feature points of the time-frequency spectrum image by feature extraction;

[0009] Copying the time-spectrum image to obtain a first time-spectrum image and a second time-spectrum image, placing the first time-spectrum image on a first plane, placing the second time-spectrum image on a second plane, connecting mutually matching feature points in the first time-spectrum image and the second time-spectrum image, and determining a line segment obtained by the connection as a feature landscape, wherein the first plane and the second plane are parallel to each other and do not overlap;

[0010] Based on the characteristic landscape, a fault type of the fuel cell is determined.

[0011] In a feasible implementation manner, obtaining a sensor signal of a fuel cell includes:

[0012] The voltage response signal of the fuel cell after the interference signal is injected is obtained.

[0013] In a feasible implementation manner, performing time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image of the sensor signal includes:

[0014] Performing a generalized S transform on the sensor signal to obtain a generalized S transform time-frequency spectrum of the sensor signal;

[0015] Based on the generalized S-transform time-frequency spectrum, a time-frequency spectrum image of the sensor signal is generated.

[0016] In a feasible implementation manner, extracting feature points of the time-frequency spectrum image by feature extraction includes:

[0017] The scale-invariant features of the time-frequency spectrum image are extracted through scale-invariant feature transformation, and the scale-invariant features are determined as feature points of the time-frequency spectrum image.

[0018] In a feasible implementation manner, determining the fault type of the fuel cell based on the characteristic landscape includes:

[0019] Projecting the characteristic landscape onto a third plane to obtain a projection image, wherein the third plane is perpendicular to the first plane;

[0020] Based on the projection image, a fault type of the fuel cell is determined.

[0021] In a feasible implementation manner, determining the fault type of the fuel cell based on the projection image includes:

[0022] Based on the number of projections of the corresponding characteristic landscape at each position in the projection image, a barcode image is generated, wherein the barcode image is composed of black rectangles and white rectangles of different widths, including: for the position where the projection of the characteristic landscape exists in the projection image, a black rectangle in the barcode image corresponds, and the width of the black rectangle is proportional to the number of projections of the characteristic landscape in the corresponding position in the projection image; for the area in the projection image where the projection of the characteristic landscape does not exist, a white rectangle in the barcode image corresponds, and the width of the white rectangle is proportional to the width of the area;

[0023] Based on the barcode image, a fault type of the fuel cell is determined.

[0024] In a feasible implementation manner, determining the fault type of the fuel cell based on the barcode image includes:

[0025] The barcode image is input into a classifier model, and the fault type of the fuel cell is determined based on a classifier result of the classifier model.

[0026] In a feasible implementation manner, determining the fault type of the fuel cell based on the barcode image includes:

[0027] Calculating the ratio of the area of ​​the black rectangle to the total area in the barcode image, and determining the ratio as the black-bar-space ratio; and / or,

[0028] Calculating the ratio of the area of ​​the white rectangle to the total area in the barcode image, and determining the ratio as the white-bar-space ratio;

[0029] Based on the black bar empty ratio and / or the white bar empty ratio, a fault type of the fuel cell is determined.

[0030] In a second aspect, the present invention provides a fuel cell fault diagnosis device based on sensor signals, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor so that the at least one processor can execute the fuel cell fault diagnosis method based on sensor signals described in any of the above embodiments.

[0031] In a third aspect, the present invention provides a non-volatile computer storage medium, which is a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium stores at least one program, each of which includes instructions, and when the instructions are executed by a terminal, the terminal executes the fuel cell fault diagnosis method based on sensor signals described in any of the above embodiments.

[0032] The fuel cell fault diagnosis method, device and medium based on sensor signals provided by the present invention have the following beneficial technical effects compared with the prior art:

[0033] (1) After acquiring a one-dimensional sensor signal, the present invention obtains a two-dimensional time-frequency spectrum image through time-frequency transformation; the time-frequency spectrum image is copied, the feature points in the two time-frequency spectrum images are placed on different planes respectively, and the two images are matched and connected to obtain a feature landscape in a three-dimensional space. The working state information of the fuel cell in the time-frequency spectrum image of the sensor signal is obtained from the perspective of the three-dimensional space, which takes a step towards a higher-dimensional space, obtains more comprehensive information in the time-frequency spectrum image, overcomes the limitation of observing the state information of the fuel cell from a low-dimensional perspective, and improves the accuracy of fuel cell fault diagnosis.

[0034] (2) The present invention generates a barcode image with high information density based on the projection of the characteristic landscape in the three-dimensional space onto the third plane. The barcode image can accurately and quickly track the working state of the fuel cell and enhance the distinction between various fault types, including water flooding fault, membrane dry fault, hydrogen starvation fault and oxygen starvation fault, thereby reducing the error rate of fault type determination.

[0035] (3) The voltage signal of the fuel cell in the present invention includes a normal voltage signal component and a fault signal component. The fault signal component in the voltage response signal obtained after the interference signal is injected is amplified, which is conducive to determining the fault type of the fuel cell based on the voltage response signal. For fuel cells with different fault types, the voltage response signal is unknown and contains information related to the working state of the fuel cell; the interference signal is the same for fuel cells with different fault types and is known information; therefore, based on the idea of ​​heavy response and weak disturbance (Emphasizes-Response Subdues-Disturbing, ERSD), the present invention uses the voltage response signal as the basis for determining the fuel cell fault type, and the interference signal does not participate in the calculation, which reduces information redundancy and unnecessary calculation time. On the premise that the fuel cell fault type can be effectively determined, the speed and efficiency of determining the fuel cell fault type are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0037] Figure 1 A flow chart of a fuel cell fault diagnosis method based on sensor signals provided by the present invention;

[0038] Figure 2A schematic diagram of a time-frequency spectrum image of a voltage response signal provided by the present invention;

[0039] Figure 3 A schematic diagram of a characteristic landscape provided by the present invention;

[0040] Figure 4 A schematic diagram of a barcode image corresponding to a fuel cell under different fault types provided by the present invention;

[0041] Figure 5 A schematic diagram of the black bar empty ratio corresponding to a fuel cell under different water fault types provided by the present invention;

[0042] Figure 6 A schematic diagram of the black bar space ratio corresponding to a fuel cell under different gas fault types provided by the present invention;

[0043] Figure 7 A schematic diagram of the structure of a fuel cell fault diagnosis device based on sensor signals provided by the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0045] The technical solution proposed by the present invention is described in detail below with reference to the accompanying drawings.

[0046] Figure 1 The flowchart of the fuel cell fault diagnosis method based on sensor signals provided by the present invention is as follows. Figure 1 As shown, the method includes the following execution steps:

[0047] S101 . Obtain a sensor signal of a fuel cell.

[0048] The sensor signal of the fuel cell in the present invention may be a voltage signal, a current signal or other signal collected by a sensor and which may reflect the working state of the fuel cell.

[0049] As a possible implementation, S101 acquires a sensor signal of a fuel cell, including: acquiring a voltage response signal of the fuel cell after an interference signal is injected into the fuel cell.

[0050] The voltage signal of the fuel cell in the present invention includes a normal voltage signal component and a fault signal component. The fault signal component in the voltage response signal obtained after injecting the interference signal is amplified, which is conducive to determining the fault type of the fuel cell based on the voltage response signal.

[0051] S102 : performing time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image of the sensor signal.

[0052] In the present invention, the sensor signal is a one-dimensional signal, and a two-dimensional time-frequency spectrum related to time and frequency can be obtained through time-frequency conversion, and a two-dimensional time-frequency spectrum image can be obtained based on the time-frequency spectrum.

[0053] As a possible implementation method, S102 performs a time-frequency transform on the sensor signal to obtain a time-frequency spectrum image of the sensor signal, including: performing a generalized S transform on the sensor signal to obtain a generalized S transform time-frequency spectrum of the sensor signal; and generating a time-frequency spectrum image of the sensor signal based on the generalized S transform time-frequency spectrum.

[0054] In some embodiments, the product of the Fast Fourier Transform (FFT) of the sensor signal and the FFT of the Gaussian window function is calculated to obtain V ST (k), as shown below.

[0055] V ST (k) = V(k)·G(k;σ(k))

[0056] Where k represents the frequency, V(k) represents the FFT of the sensor signal, G(k; σ(k)) is the FFT of the Gaussian window function, and σ(k) represents the scale parameter associated with the frequency k.

[0057] Then, by multiplying the above product V ST (k) Perform an inverse fast Fourier transform (IFFT) to obtain the generalized S-transform time-frequency spectrum V of the sensor signal ST (n,k), as shown in the following formula.

[0058] V ST (n, k) = IFFT(V ST (k))

[0059] Here, n represents time.

[0060] Exemplarily, the sensor signal is a voltage response signal of the fuel cell after the interference signal is injected. The voltage response signal is subjected to a generalized S transform to obtain a generalized S transform time-frequency spectrum of the voltage response signal, and a time-frequency spectrum image of the voltage response signal is generated, such as Figure 2Shown is a schematic diagram of a time-frequency spectrum image of a voltage response signal provided by the present invention.

[0061] In addition to the generalized S transform, other time-frequency transform methods can also be used, for example, short-time Fourier transform, continuous wavelet transform, real wavelet transform, wavelet synchronous compression transform, wavelet synchronous extraction transform, vertical second-order synchronous compression transform, multi-scale synchronous compression transform, local maximum synchronous compression transform, transient extraction transform, Hilbert transform and other time-frequency transform methods.

[0062] S103 , obtaining feature points of the time-frequency spectrum image through feature extraction.

[0063] In order to improve processing efficiency and extract information from the image, the present invention can perform feature extraction on the time-frequency spectrum image to obtain key points of significance and stability in the time-frequency spectrum image, also known as feature points of the time-frequency spectrum image.

[0064] The scale invariant feature obtained by the Scale Invariant Feature Transform (SIFT) remains invariant to rotation, scale scaling, brightness change, etc., and is a very stable local feature. As a possible implementation method, S103 extracts feature points of the time-frequency spectrum image through feature extraction, including: extracting the scale invariant features of the time-frequency spectrum image through the scale invariant feature transform, and determining the scale invariant features as the feature points of the time-frequency spectrum image.

[0065] S104, copying the time-spectrum image to obtain a first time-spectrum image and a second time-spectrum image, placing the first time-spectrum image on a first plane, placing the second time-spectrum image on a second plane, connecting mutually matching feature points in the first time-spectrum image and the second time-spectrum image, and determining a line segment obtained by the connection as a feature landscape.

[0066] The first plane and the second plane are parallel to each other and do not overlap.

[0067] In some embodiments, D1 and D2 are used to represent feature points in two identical time-frequency spectrum images, respectively, and the Euclidean distance d(D1, D2) between the feature points of the two images is calculated, as shown in the following formula.

[0068]

[0069] Among them, d 1i is the value of D1 in the i-th dimension, d 2i is the value of D2 in the i-th dimension.

[0070] For a feature point D1 in one of the images, find the feature point D1 in the other image that is closest to the feature point. nn The second closest feature point Dnnn .

[0071] Then, at the nearest feature point D nn The second closest feature point D nnn Apply a ratio test between them to calculate the feature point D1 in one image and the closest feature point D in the other image. nn The first distance between the feature point D1 in one image and the second closest feature point D in the other image nnn If the ratio of the first distance to the second distance is less than a preset ratio threshold, it is determined that the match is valid, and the feature point D1 in one image is matched with the feature point D in the other image. nn Determined as a set of matching points.

[0072] It is understandable that it is necessary to ensure that each feature point is matched only once to obtain a set of matching points of two identical time-frequency spectrum images.

[0073] For example, Figure 3 A schematic diagram of a characteristic landscape provided by the present invention, such as Figure 3 As shown, the feature points of one of the two identical time-frequency spectrum images are placed on a first plane, and the feature points of the other image are placed on a second plane. The first plane and the second plane are different and both are parallel to the XY plane. The line segment obtained by connecting the mutually matching feature points is determined as the feature landscape.

[0074] S105: Determine the fault type of the fuel cell based on the characteristic landscape.

[0075] After acquiring a one-dimensional sensor signal, the present invention obtains a two-dimensional time-frequency spectrum image through time-frequency transformation; the time-frequency spectrum image is copied, the feature points in the two time-frequency spectrum images are placed on different planes respectively, the two images are matched and connected to obtain a feature landscape in a three-dimensional space, and the working state information of the fuel cell in the time-frequency spectrum image of the sensor signal is obtained from the perspective of the three-dimensional space, which takes a step towards a higher-dimensional space, obtains more comprehensive information in the time-frequency spectrum image, overcomes the limitation of observing the state information of the fuel cell from a low-dimensional perspective, and improves the accuracy of fuel cell fault diagnosis.

[0076] As a possible implementation method, S105 determines the fault type of the fuel cell based on the characteristic landscape, including:

[0077] The characteristic landscape is projected onto a third plane to obtain a projection image, where the third plane is perpendicular to the first plane; based on the projection image, the fault type of the fuel cell is determined.

[0078] As in the above example, the first plane and the second plane are planes parallel to the XY plane, and the third plane may be a plane parallel to the XZ plane, or a plane parallel to the YZ plane.

[0079] As a possible implementation method, based on the projection image, the fault type of the fuel cell is determined, including:

[0080] Based on the number of projections of the corresponding characteristic landscape at each position in the projection image, a barcode image is generated, and the barcode image is composed of black rectangles and white rectangles of different widths, including: for the position where the projection of the characteristic landscape exists in the projection image, it corresponds to a black rectangle in the barcode image, and the width of the black rectangle is proportional to the number of projections of the characteristic landscape in the corresponding position in the projection image; for the area in the projection image where the projection of the characteristic landscape does not exist, it corresponds to a white rectangle in the barcode image, and the width of the white rectangle is proportional to the width of the area; based on the barcode image, the fault type of the fuel cell is determined.

[0081] In this way, based on the projection of the characteristic landscape in the three-dimensional space onto the third plane, a barcode image with high information density is generated. The barcode image can accurately and quickly track the working status of the fuel cell and enhance the distinction between various fault types, including water flooding fault, membrane dry fault, hydrogen starvation fault and oxygen starvation fault, thereby reducing the error rate of fault type determination.

[0082] As a possible implementation, the fault type of the fuel cell is determined based on the barcode image, including: inputting the barcode image into a classifier model, and determining the fault type of the fuel cell based on a classifier result of the classifier model.

[0083] For example, Figure 4 A schematic diagram of a barcode image corresponding to a fuel cell under different fault types provided by the present invention, Figure 4 The barcode images corresponding to the fuel cells under water flooding fault, membrane dry fault, hydrogen starvation fault, and oxygen starvation fault are shown. It can be seen that the barcode images corresponding to the fuel cells under the same fault type are similar, and the barcode images corresponding to the fuel cells under different fault types are different. In other words, the fuel cell fault type can be determined through the barcode image.

[0084] As a possible implementation method, based on the barcode image, determining the fault type of the fuel cell includes: calculating the ratio of the area of ​​the black rectangle in the barcode image to the total area, and determining the ratio as the black bar-space ratio; and / or calculating the ratio of the area of ​​the white rectangle in the barcode image to the total area, and determining the ratio as the white bar-space ratio; based on the black bar-space ratio and / or the white bar-space ratio, determining the fault type of the fuel cell.

[0085] For example, Figure 5 A schematic diagram of the black bar empty ratio corresponding to a fuel cell under different water fault types provided by the present invention, Figure 5 The black bar to space ratio of the fuel cell under the flooding fault and the membrane dry fault is shown. It can be seen that the black bar to space ratio corresponding to the flooding fault is significantly different from the black bar to space ratio corresponding to the membrane dry fault, that is, the flooding fault and the membrane dry fault can be effectively distinguished by the black bar to space ratio; Figure 6 A schematic diagram of the black bar space ratio corresponding to a fuel cell under different gas fault types provided by the present invention, Figure 6 The black bar space ratio of the fuel cell under hydrogen starvation fault and oxygen starvation fault is shown. It can be seen that the black bar space ratio corresponding to the hydrogen starvation fault and the black bar space ratio corresponding to the oxygen starvation fault are significantly different, that is, the hydrogen starvation fault and the oxygen starvation fault can be effectively distinguished by the black bar space ratio.

[0086] Exemplarily, the fuel cell fault diagnosis method based on sensor signals provided by the present invention can be implemented by three core components: component 1 is used to obtain the voltage response signal of the fuel cell after the interference signal is injected, and perform a generalized S transform on the voltage response signal to obtain a time-frequency spectrum image of the voltage response signal; component 2 is used to obtain a characteristic landscape based on the time-frequency spectrum image of the voltage response signal and generate a barcode image; component 3 uses a barcode reader to calculate the black bar-space ratio or the white bar-space ratio of the barcode image, and determines the fault type of the fuel cell based on the black bar-hole ratio or the white bar-space ratio.

[0087] Specifically, component 1 is responsible for injecting interference signals, and then using a rolling window to combine the voltage response signals collected by the sensor at a fixed frequency. Each new data point forms a window with the previous 19 data points, which rolls forward and slides point by point. It takes 0.845 seconds to slide a data point. For each data window, a generalized S transform is performed to obtain a time-frequency spectrum image, such as Figure 2 Compared with the traditional architecture that calculates the electrochemical impedance spectrum based on the interference signal and the response signal, component 1 determines the working state and fault type of the fuel cell based on the electrochemical impedance spectrum, reduces the information redundancy in the data processing process, and only takes 0.007 seconds to calculate the time-frequency spectrum image of each window. It can be seen that the above method greatly reduces the data processing time.

[0088] Component 2 is responsible for extracting scale-invariant features from the time-frequency spectrum image obtained by component 1 and generating a feature landscape in three-dimensional space, as shown in 3. Component 2 also includes a barcode generator, which obtains a barcode image with high-dimensional features and high information density based on the projection of the feature landscape on the third plane, as shown in Figure 4As shown. Compared with the prior art method of fuel cell fault diagnosis based on one-dimensional signals or two-dimensional images, component 2 overcomes the inherent limitations of low-dimensional features. It only takes 1.06 seconds for component 2 to generate a barcode image based on a time-frequency spectrum image.

[0089] The barcode reader of component 3 is responsible for using virtual photoelectric recognition to scan the barcode image obtained by component 2, determine the black bar hole ratio or white bar empty ratio of the barcode image, and determine the fault type of the fuel cell through a machine learning algorithm. The time for component 3 to determine the fuel cell fault type based on the barcode image is only 0.017 seconds.

[0090] Corresponding to the above-mentioned embodiments, the present invention also provides a fuel cell fault diagnosis device based on sensor signals. Figure 7 A structural schematic diagram of a fuel cell fault diagnosis device based on sensor signals provided in an embodiment of the present invention, the device may include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, so that the at least one processor can execute the fuel cell fault diagnosis method based on sensor signals as described in the above embodiment.

[0091] In a possible implementation of the present invention, the aforementioned at least one processor is capable of executing the following steps: acquiring a sensor signal of a fuel cell; performing a time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image of the sensor signal; obtaining feature points of the time-frequency spectrum image by feature extraction; copying the time-frequency spectrum image to obtain a first time-frequency spectrum image and a second time-frequency spectrum image, placing the first time-frequency spectrum image on a first plane, placing the second time-frequency spectrum image on a second plane, connecting mutually matching feature points in the first time-frequency spectrum image and the second time-frequency spectrum image, determining a line segment obtained by the connection as a feature landscape, wherein the first plane and the second plane are parallel to each other and do not overlap; and determining the fault type of the fuel cell based on the feature landscape.

[0092] In a specific implementation, the present invention can also be provided as a non-volatile computer storage medium, wherein the storage medium is a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium stores at least one program, each program includes instructions, and when the instructions are executed by the terminal, the terminal executes the fuel cell fault diagnosis method based on sensor signals as described in the above embodiment.

[0093] In a possible implementation of the present invention, the aforementioned terminal executes, obtains a sensor signal of the fuel cell; performs time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image of the sensor signal; obtains feature points of the time-frequency spectrum image by feature extraction; copies the time-frequency spectrum image to obtain a first time-frequency spectrum image and a second time-frequency spectrum image, places the first time-frequency spectrum image on a first plane, places the second time-frequency spectrum image on a second plane, connects mutually matching feature points in the first time-frequency spectrum image and the second time-frequency spectrum image, determines a line segment obtained by the connection as a feature landscape, and the first plane and the second plane are parallel to each other and do not overlap; based on the feature landscape, determines the fault type of the fuel cell.

[0094] The various embodiments of the present invention are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0095] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0096] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A fuel cell fault diagnosis method based on sensor signals, characterized in that: The method comprises: Acquire sensor signals of the fuel cell; Performing time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image of the sensor signal; Obtaining feature points of the time-frequency spectrum image by feature extraction; Copying the time-spectrum image to obtain a first time-spectrum image and a second time-spectrum image, placing the first time-spectrum image on a first plane, placing the second time-spectrum image on a second plane, connecting mutually matching feature points in the first time-spectrum image and the second time-spectrum image, and determining a line segment obtained by the connection as a feature landscape, wherein the first plane and the second plane are parallel to each other and do not overlap; Based on the characteristic landscape, a fault type of the fuel cell is determined.

2. The fuel cell fault diagnosis method based on sensor signals according to claim 1, characterized in that: Obtain fuel cell sensor signals, including: The voltage response signal of the fuel cell after the interference signal is injected is obtained.

3. The fuel cell fault diagnosis method based on sensor signals according to claim 1, characterized in that: Performing a time-frequency transformation on the sensor signal to obtain a time-frequency spectrum image of the sensor signal includes: Performing a generalized S transform on the sensor signal to obtain a generalized S transform time-frequency spectrum of the sensor signal; Based on the generalized S-transform time-frequency spectrum, a time-frequency spectrum image of the sensor signal is generated.

4. The fuel cell fault diagnosis method based on sensor signals according to claim 1, characterized in that: Extracting feature points of the time-frequency spectrum image by feature extraction includes: The scale-invariant features of the time-frequency spectrum image are extracted through scale-invariant feature transformation, and the scale-invariant features are determined as feature points of the time-frequency spectrum image.

5. The fuel cell fault diagnosis method based on sensor signals according to claim 1, characterized in that: Determining the fault type of the fuel cell based on the characteristic landscape includes: Projecting the characteristic landscape onto a third plane to obtain a projection image, wherein the third plane is perpendicular to the first plane; Based on the projection image, a fault type of the fuel cell is determined.

6. The fuel cell fault diagnosis method based on sensor signals according to claim 5, characterized in that: Determining a fault type of the fuel cell based on the projection image includes: Based on the number of projections of the corresponding characteristic landscape at each position in the projection image, a barcode image is generated, wherein the barcode image is composed of black rectangles and white rectangles of different widths, including: for the position where the projection of the characteristic landscape exists in the projection image, a black rectangle in the barcode image corresponds, and the width of the black rectangle is proportional to the number of projections of the characteristic landscape in the corresponding position in the projection image; for the area in the projection image where the projection of the characteristic landscape does not exist, a white rectangle in the barcode image corresponds, and the width of the white rectangle is proportional to the width of the area; Based on the barcode image, a fault type of the fuel cell is determined.

7. The fuel cell fault diagnosis method based on sensor signals according to claim 6, characterized in that: Determining a fault type of the fuel cell based on the barcode image includes: The barcode image is input into a classifier model, and the fault type of the fuel cell is determined based on a classifier result of the classifier model.

8. The fuel cell fault diagnosis method based on sensor signals according to claim 6, characterized in that: Determining a fault type of the fuel cell based on the barcode image includes: Calculating the ratio of the area of ​​the black rectangle to the total area in the barcode image, and determining the ratio as the black-bar-space ratio; and / or, Calculating the ratio of the area of ​​the white rectangle to the total area in the barcode image, and determining the ratio as the white-bar-space ratio; Based on the black bar empty ratio and / or the white bar empty ratio, a fault type of the fuel cell is determined.

9. A fuel cell fault diagnosis device based on sensor signals, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, so that the at least one processor can execute the fuel cell fault diagnosis method based on sensor signals according to any one of claims 1-8.

10. A non-volatile computer storage medium, characterized in that: The storage medium is a non-volatile computer-readable storage medium, which stores at least one program, each of which includes instructions, which, when executed by a terminal, enable the terminal to execute a fuel cell fault diagnosis method based on sensor signals according to any one of claims 1-8.

Images