A method and device for predicting stack failure, a storage medium and an equipment

Abstract

The application discloses a kind of pre-judgment method, device, storage medium and equipment of electric pile failure, according to the weight value of difference, the weight value of consistency, overall single cell voltage matrix, difference data point, consistency data point are calculated and target sample distance is determined;According to target sample distance, local density is calculated;The target local density of difference data point and consistency data point is calculated;According to local density, target local density, local outlier factor is calculated;When local outlier factor is greater than preset threshold, electric pile voltage is early warning state;According to difference data point, consistency data point, clustering index, target clustering index are constructed;When clustering index is not preset threshold, target clustering index is preset threshold, early warning state is changed to failure state, compared with prior art, it is not necessary to artificially operate to change the pressure difference of yin and yang two sides, based on clustering index and target clustering index, current electric pile condition is known, the pre-judgment of electric pile performance failure when real vehicle online operation is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fuel cells, in particular to a method and device for predicting stack failure, a storage medium and equipment. BACKGROUND

[0002] A proton exchange membrane hydrogen fuel cell can directly convert the chemical energy of hydrogen fuel into electrical energy and heat energy through an electrochemical reaction. Due to the characteristic that a fuel cell can maintain high power density in a low-temperature environment, a fuel cell vehicle is suitable for use in extreme environments such as high-cold, high-temperature and high-altitude. However, complex operating environments can cause mechanical failure or even damage to the proton exchange membrane, which in turn can cause the phenomenon that hydrogen fuel leaks from the anode to the cathode in the fuel cell stack to cause a chemical reaction, resulting in damage to the membrane and catalyst in the membrane hole area due to the temperature rise. This phenomenon can cause a large amount of hydrogen leakage in the stack, which can cause safety hazards and trigger the hydrogen sensor detection alarm, resulting in the fuel cell vehicle being unable to start.

[0003] However, the existing methods for detecting hydrogen leakage in a fuel cell stack all require a professional engineer to manually change the pressure difference between the anode and the cathode to detect the hydrogen leakage phenomenon in the fuel cell stack, or to stop the stack on a professional test bench for open-circuit voltage detection after returning to the factory. Therefore, it is impossible to predict the performance failure of the stack during online operation of the vehicle.

[0004] Therefore, how to predict the performance failure of the stack during online operation of the vehicle has become a problem to be solved in the field. SUMMARY

[0005] The present application provides a method and device for predicting stack failure, a storage medium and equipment, which aims to predict the performance failure of the stack during online operation of the vehicle.

[0006] In order to achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0007] A method for predicting stack failure, comprising:

[0008] According to the difference weight value, the consistency weight value and the overall single cell voltage matrix, difference data points and consistency data points are calculated; the difference weight value and the consistency weight value are obtained by pre-processing based on the covariance matrix; and the overall voltage matrix is established based on each single cell voltage in advance;

[0009] According to the difference data points and the consistency data points, a target sample distance is determined;

[0010] According to the target sample distance, a local density is calculated;

[0011] periodically calculate all local densities of the difference data points and the consistency data points to obtain a target local density;

[0012] calculate a local outlier factor according to the local density and the target local density;

[0013] confirm that the stack voltage is in a pre-warning state when the local outlier factor is greater than a preset threshold value;

[0014] construct a clustering index and a target clustering index according to the difference data points and the consistency data points;

[0015] change the pre-warning state to a fault state when the clustering index is not the preset threshold value and the target clustering index is the preset threshold value.

[0016] Optionally, the process of pre-establishing the overall voltage matrix based on each single cell voltage includes:

[0017] store all the single cell voltages in a database when all the single cell voltages sent by a preset device are received;

[0018] obtain each single cell voltage in a preset voltage range in a first preset time period from the database;

[0019] establish the overall single cell voltage matrix according to each single cell voltage.

[0020] Optionally, the process of pre-processing the difference weight value and the consistency weight value based on the covariance matrix includes:

[0021] construct each single cell voltage matrix according to the overall single cell voltage matrix;

[0022] decentralize each single cell voltage matrix to obtain a decentralized single cell voltage matrix;

[0023] establish the covariance matrix according to the decentralized single cell voltage matrix;

[0024] singular value decompose the covariance matrix to obtain the difference weight value and the consistency weight value.

[0025] Optionally, the target sample distance determined according to the difference data points and the consistency data points includes:

[0026] calculate sample distances between the difference data points and the consistency data points in a second preset time period to obtain each sample distance;

[0027] Select the largest sample distance from the various sample distances, and identify it as the target sample distance.

[0028] Optionally, the constructing a clustering index and a target clustering index according to the difference data points and the consistency data points comprises:

[0029] Based on the difference data points and the consistency data points in a third preset time period, a data set is constructed;

[0030] Based on the data set, a clustering index is established;

[0031] When the clustering index is not the preset threshold value, based on the difference data points and the consistency data points in a fourth preset time period, a target data set is constructed;

[0032] Based on the target data set, a target clustering index is established.

[0033] Optionally, it further comprises:

[0034] When the clustering index is the preset threshold value, the early warning state is changed to a normal operation state.

[0035] Optionally, it further comprises:

[0036] When the target clustering index is not the preset threshold value, prompt information is sent to a user.

[0037] A device for pre-judging a stack fault, comprising:

[0038] A first calculation unit is configured to calculate difference data points and consistency data points based on a difference weight value, a consistency weight value, and a total cell voltage matrix; the difference weight value and the consistency weight value are obtained by pre-processing a covariance matrix; and the total voltage matrix is established based on each cell voltage in advance.

[0039] A determination unit is configured to determine a target sample distance based on the difference data points and the consistency data points.

[0040] A second calculation unit is configured to calculate a local density based on the target sample distance.

[0041] A third calculation unit is configured to periodically calculate all local densities of the difference data points and the consistency data points to obtain a target local density.

[0042] A fourth calculation unit is configured to calculate a local outlier factor based on the local density and the target local density.

[0043] A confirmation unit is configured to confirm that a stack voltage is in an early warning state when the local outlier factor is greater than a preset threshold value.

[0044] a construction unit, configured to construct a clustering index and a target clustering index according to the difference data points and the consistency data points;

[0045] a change unit, configured to change the early warning state to a fault state when the clustering index is not the preset threshold value and the target clustering index is the preset threshold value.

[0046] A computer readable storage medium, comprising a stored program, wherein the program is executed by a processor to perform the method for predicting a stack fault.

[0047] A device for predicting a stack fault, comprising a processor, a memory and a bus; the processor is connected with the memory through the bus;

[0048] The memory is configured to store a program, and the processor is configured to execute the program, wherein the program is executed by the processor to perform the method for predicting a stack fault.

[0049] The technical scheme provided in the application calculates difference data points, consistency data points and determines a target sample distance according to the weight value of difference, the weight value of consistency and the overall single cell voltage matrix; calculates a local density according to the target sample distance; calculates a target local density of the difference data points and the consistency data points; calculates a local outlier factor according to the local density and the target local density; when the local outlier factor is greater than a preset threshold value, the stack voltage is in an early warning state; constructs a clustering index and a target clustering index according to the difference data points and the consistency data points; when the clustering index is not the preset threshold value and the target clustering index is the preset threshold value, changes the early warning state to a fault state. Compared with the prior art, it is not necessary to manually change the pressure difference between the positive side and the negative side, and the current stack condition is known based on the clustering index and the target clustering index, so that the prediction of the stack performance fault during the online operation of the real vehicle is realized. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Figure 1a A flowchart of a method for predicting a stack fault provided by an embodiment of the present application;

[0052] Figure 1b A flowchart of a method for predicting a stack fault provided by an embodiment of the present application;

[0053] Figure 1c A schematic diagram of a fuel cell stack state provided in an embodiment of this application;

[0054] Figure 2 A flowchart illustrating another method for predicting fuel cell stack faults provided in this application embodiment;

[0055] Figure 3 This is a schematic diagram of the architecture of a fuel cell stack fault prediction device provided in an embodiment of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0057] like Figure 1a , Figure 1b The diagram shown is a flowchart of a method for predicting fuel cell stack faults according to an embodiment of this application, applied to a host computer, and includes:

[0058] S101: When all individual unit voltages are received from the preset device, store all individual unit voltages in the database.

[0059] The preset equipment includes, but is not limited to: vehicle-mounted remote information processor and individual cell voltage indicator stack individual cell voltage.

[0060] It should be noted that, through a preset communication method (such as CAN communication), the fuel cell control unit (FCU) reads the voltage data of the N fuel cell stack cells and periodically sends the cell voltage data to the host computer via Ethernet through the on-board telematics processor.

[0061] S102: Retrieve the voltage of each individual unit within the preset voltage range during the first preset time period from the database.

[0062] The preset voltage range includes, but is not limited to, the range from minimum current density to maximum current density.

[0063] Optionally, the first preset time period can be set according to the actual situation, and no specific restrictions are imposed here.

[0064] S103: Establish the overall individual unit voltage matrix based on the individual unit voltages.

[0065] The specific implementation process of establishing the overall cell voltage matrix according to the individual cell voltage is: establishing the overall cell voltage matrix by using a matrix formula, and the specific form of the matrix formula is shown in formula (1).

[0066]

[0067] In formula (1), U is the overall cell voltage matrix, U t,N is the voltage value of the nth cell at the tth moment, t is any moment t from 1 to T, the minimum current density value corresponding to the cell voltage value at any moment t is I tn,min , and the maximum current density value is I tn,max .

[0068] S104: According to the overall cell voltage matrix, each individual cell voltage matrix is constructed.

[0069] The specific form of each individual cell voltage matrix is shown in formula (2).

[0070]

[0071] In formula (2), U n is the voltage value of the nth cell of the fuel cell stack at 1 to T moments within a predetermined time period.

[0072] S105: For each individual cell voltage matrix, the individual cell voltage matrix is decentered to obtain a decentered individual cell voltage matrix.

[0073] The specific implementation process of decentering the individual cell voltage matrix is: decentering the individual cell voltage matrix by using a decentering formula, and the specific form of the decentering formula is shown in formula (3).

[0074]

[0075] In formula (3), U' n is the decentered individual cell voltage matrix, U' t,n is the decentered individual cell voltage value of the nth cell at the tth moment.

[0076] S106: According to the decentered individual cell voltage matrix, a covariance matrix is established.

[0077] It should be noted that the specific form of the covariance matrix is shown in formula (4).

[0078]

[0079] Wherein,

[0080]

[0081] In formula (4), Var(U N ) is the covariance of the fuel cell stack nth single cell in the interval of 1 to t, Cov(U N , U n ) is the covariance of the fuel cell stack nth single cell and the nth single cell in the interval of 1 to t, the single cell voltage covariance matrix Q can embody the change rule of any single cell voltage in the interval of 1 to T and the change degree rule between any two single cell voltages, which can reflect the single cell voltage change rule in the interval of 1 to T and the current density I tn,min to I tn,max interval, embody the different degrees of fuel cell stack voltage change when fuel cell performance failure occurs (including but not limited to leakage problem).

[0082] It should be noted that the performance failure of the fuel cell stack will cause the degree of change of the single cell voltage of all single cells of the fuel cell stack to change significantly over time, and the change relationship between each single cell will also change, therefore, the covariance matrix is established to calculate the covariance between each single cell and other single cells.

[0083] S107: singular value decomposition is performed on the covariance matrix to obtain a weight value of difference and a weight value of consistency.

[0084] Wherein, the difference indicates the performance difference of the fuel cell stack, and the consistency indicates the performance consistency of the fuel cell stack.

[0085] It should be noted that the specific implementation process of singular value decomposition on the covariance is: singular value decomposition is performed on the covariance matrix by using a singular value decomposition formula, and the specific form of the singular value decomposition formula is shown in formula (5).

[0086]

[0087] In formula (5), u and u T are orthogonal matrices corresponding to two new principal elements of the fuel cell stack N single cell voltages, u x,n is the weight value of the difference of the fuel cell stack nth single cell voltage, u y,n is the weight value of the consistency of the fuel cell stack nth single cell voltage, and are singular values reduced by T times, is the importance of difference, is the importance of consistency, and T is the total time length of the fuel cell stack single cell voltage sample.

[0088] S108: Calculate the difference data points and the consistency data points according to the weight value of difference, the weight value of consistency and the overall single cell voltage matrix.

[0089] The difference data points indicate the data points of the difference of the fuel cell stack, and the consistency data points indicate the data points of the consistency of the fuel cell stack.

[0090] It should be noted that the specific implementation process of calculating the difference data points and the consistency data points according to the weight value of difference, the weight value of consistency and the overall single cell voltage matrix is: calculating by using the data point calculation formula, and the specific form of the data point calculation formula is shown in formula (6).

[0091]

[0092] In formula (6), X new is the difference data points, Y new is the consistency data points, X t is the voltage sample point of the difference of the fuel cell stack, and Y t is the voltage sample point of the consistency of the fuel cell stack.

[0093] S109: Calculate the sample distance between the difference data points and the consistency data points in the second preset time period to obtain each sample distance.

[0094] Optionally, the second preset time period can be from t-1 time to 1 time.

[0095] It should be noted that the specific implementation process of calculating the sample distance between the difference data points and the consistency data points in the first preset time period is: calculating by using the sample distance formula, and the specific form of the sample distance formula is shown in formula (7).

[0096]

[0097] In formula (7), d is the sample distance, X t-1 is the voltage sample point of the difference data points at t-1 time, X1 is the voltage sample point of the difference data points at 1 time, Y t-1 is the voltage sample point of the consistency data points at t-1 time, and Y1 is the voltage sample point of the consistency data points at 1 time.

[0098] S110: Select the largest sample distance from each sample distance and mark it as the target sample distance.

[0099] S111: Calculate the local density according to the target sample distance.

[0100] It should be noted that the specific implementation process of calculating the local density according to the target sample distance is: calculating by using the local density formula, and the specific form of the local density formula is shown in formula (8).

[0101]

[0102] In formula (8), δ t-1 is the local density, d t-1 is the maximum sample distance (i.e. the target sample distance) from the t-1 time difference data point and the consistency data point to the 1 time.

[0103] S112: periodically calculating all local densities of the difference data point and the consistency data point to obtain a target local density.

[0104] It should be noted that the specific implementation process of periodically calculating all local densities of the difference data point and the consistency data point is: calculating by using the target local density formula, and the specific form of the target local density formula is shown in formula (9).

[0105]

[0106] In formula (9), δ i is the local density at i time.

[0107] S113: calculating a local outlier factor according to the local density and the target local density.

[0108] It should be noted that the specific implementation process of calculating the local outlier factor according to the local density and the target local density is: calculating by using the local outlier factor formula, and the specific form of the local outlier factor formula is shown in formula (10).

[0109]

[0110] In formula (10), α t is the local outlier factor.

[0111] S114: judging whether the local outlier factor is not greater than a preset threshold.

[0112] If the local outlier factor is not greater than the preset threshold, S115 is executed, otherwise S116 is executed.

[0113] Optionally, the preset threshold can be 1.

[0114] Specifically, assuming that the local outlier factor is 0.5, it is judged whether the local outlier factor is not greater than the preset threshold. Obviously, the local outlier factor is not greater than the preset threshold, and therefore S115 is continued to be executed.

[0115] S115: Confirm that the fuel cell stack voltage is in normal operating condition.

[0116] S116: Confirm that the fuel cell stack voltage is in a warning state.

[0117] It should be noted that, due to the randomness of the operating conditions of fuel cell vehicles and the probability of individual anomalies in voltage signal transmission, the warning status may be due to individual phenomena caused by operational abnormalities, rather than stack performance failures. Therefore, in order to avoid the situation where noise data deviates slightly from the data center but is mistakenly identified as abnormal data, a secondary judgment is required.

[0118] S117: Construct a dataset based on the differential and consistent data points within the third preset time period.

[0119] Optionally, the third preset time period can be from time 1 to time t+5×P.

[0120] It should be noted that the specific form of the dataset is as shown in formula (11).

[0121] N ε (U j )={U j ∈U 1,t+5×P |d(U i U j )≤ε} (11)

[0122] Where, |N ε (U j )|≥P

[0123] In formula (11), U i Let U be the difference data point and the consistency data point at time i. j Let N be the difference and consistency data points at time j, ε be the center of the circle, and N be the number of points. ε (U i (From time 1 to) t+5×P The set of samples centered at ε, representing the difference and consistency of data points within a given time period. ε (U j From the first moment to t+5×P The set of samples centered at ε, representing the difference and consistency of data points within a given time period, U i For U j In the ε-neighborhood of U j For any sample point outside of U, any point belonging to U i ε-neighborhood but not belonging to U j The ε-neighborhood of dissimilar and consistent data points U k They are connected by density.

[0124] S118: establishing a clustering index based on the data set.

[0125] The specific form of the clustering index is shown in formula (12).

[0126]

[0127] In formula (12), U k is the difference data point and the consistency data point at the kth moment, connection normal is the clustering index.

[0128] S119: determining whether the clustering index is a preset threshold.

[0129] If the clustering index is the preset threshold, S120 is executed, otherwise S121 is executed.

[0130] Specifically, assuming that the clustering index is 0 and the preset threshold is 1, it is determined whether the clustering index is the preset threshold. Obviously, the clustering index is not the preset threshold, and therefore S121 is continuously executed.

[0131] S120: changing the early warning state to a normal operation state.

[0132] It should be noted that when the clustering index is the preset threshold, it indicates that the difference data point and the consistency data point are connected in density, and therefore the early warning state is changed to the normal operation state.

[0133] S121: constructing a target data set based on the difference data point and the consistency data point in a fourth preset time period.

[0134] Optionally, the fourth preset time period is from the tth moment to the t+5×P moment.

[0135] When the clustering index is not the preset threshold, it indicates that the difference data point and the consistency data point are not connected in density, and therefore the target data set is constructed based on the difference data point and the consistency data point in the fourth preset time period.

[0136] It should be noted that the specific form of the target data set is shown in formula (13).

[0137] N’ ε (U j )={U j ∈U t,t+5×P |d(U i ,U j )≤ε} (13)

[0138] Wherein, |N’ ε (U j) ≥ P

[0139] In formula (13), N ε ’(U i ) is a sample set of the difference data points and the consistency data points with ε as the center in the i-th moment to the t+5×P-th moment, N ε ’(U j ) is a sample set of the difference data points and the consistency data points with ε as the center in the j-th moment to the t+5×P-th moment.

[0140] S122: establishing a target clustering index based on the target data set.

[0141] In formula (14), connection warning is the target clustering index.

[0142]

[0143] In formula (14), connection warning is the target clustering index.

[0144] S123: judging whether the target clustering index is a preset threshold.

[0145] If the target clustering index is the preset threshold, S124 is executed, otherwise S125 is executed.

[0146] Optionally, whether the stack is abnormal can be judged through a fault index, the fault index being constructed based on the target clustering index and the clustering index, wherein the specific form of the fault index is shown in formula (15).

[0147]

[0148] In formula (15), z is the fault index, when z=0, it means that no stack fault occurs, if z=1, it means that a stack performance fault occurs which does not affect normal use, but continuous no treatment will cause irreversible damage to the stack, therefore, the warning state is changed to the fault state, if z=2, it appears irregular distribution, which is a vehicle-mounted telematics processor signal transmission problem, wherein the warning state, the normal state and the fault state can be shown as formula (16). Figure 1c

[0149] S124: changing the warning state to the fault state.

[0150] S125: sending prompt information to the user.

[0151] In formula (16), the prompt information indicates information prompting the user to check the preset equipment.

[0152] In summary, when the clustering index is not the preset threshold value, the target data set is constructed based on the difference data points and the consistency data points in the second preset time period, and the target clustering index is established based on the target data set. When the target clustering index is the preset threshold value, the warning state is changed to the fault state. Compared with the prior art, it is not necessary to manually change the pressure difference between the positive side and the negative side by a professional engineer. By judging whether the clustering index and the target clustering index are the preset threshold value, the current stack condition can be known, so that the prediction of the stack performance fault during the online running of the real vehicle is realized.

[0153] As shown in Figure 2 FIG. 2 is a flowchart of another method for predicting a stack fault provided by an embodiment of the present application, including the following steps:

[0154] S201: The difference data points and the consistency data points are calculated according to the weight value of the difference, the weight value of the consistency, and the overall single cell voltage matrix.

[0155] The weight value of the difference and the weight value of the consistency are obtained by pre-processing based on the covariance matrix, and the overall voltage matrix is established based on each single cell voltage.

[0156] S202: The target sample distance is determined according to the difference data points and the consistency data points.

[0157] S203: The local density is calculated according to the target sample distance.

[0158] S204: The target local density is obtained by periodically calculating all the local densities of the difference data points and the consistency data points.

[0159] S205: The local outlier factor is calculated according to the local density and the target local density.

[0160] S206: When the local outlier factor is greater than the preset threshold value, it is determined that the stack voltage is in a warning state.

[0161] S207: The clustering index and the target clustering index are constructed according to the difference data points and the consistency data points.

[0162] S208: When the clustering index is not the preset threshold value and the target clustering index is the preset threshold value, the warning state is changed to the fault state.

[0163] In summary, when the clustering index is not the preset threshold value, the target data set is constructed based on the difference data points and the consistency data points in the second preset time period, and the target clustering index is established based on the target data set. When the target clustering index is the preset threshold value, the warning state is changed to the fault state. Compared with the prior art, the pressure difference between the positive side and the negative side does not need to be manually changed by a professional engineer. By judging whether the clustering index and the target clustering index are the preset threshold value, the current stack condition can be known. Therefore, the performance fault of the stack during online running of the real vehicle is predicted.

[0164] As shown in Figure 3 FIG. 1 is a schematic diagram of an architecture of a stack fault prediction device provided by an embodiment of the present application, which comprises:

[0165] The first calculation unit 100 is configured to calculate the difference data points and the consistency data points based on the weight value of difference, the weight value of consistency and the overall single cell voltage matrix. The weight value of difference and the weight value of consistency are obtained by pre-processing the covariance matrix. The overall single cell voltage matrix is established based on each single cell voltage.

[0166] The first calculation unit 100 is specifically configured to: store all the single cell voltages in the database when receiving all the single cell voltages sent by the preset device; obtain each single cell voltage of the preset voltage interval in the first preset time period from the database; and establish the overall single cell voltage matrix based on each single cell voltage.

[0167] The first calculation unit 100 is specifically configured to: construct each single cell voltage matrix based on the overall single cell voltage matrix; perform decentralization processing on the single cell voltage matrix to obtain the decentralized single cell voltage matrix for each single cell voltage matrix; establish the covariance matrix based on the decentralized single cell voltage matrix; and perform singular value decomposition on the covariance matrix to obtain the weight value of difference and the weight value of consistency.

[0168] The determination unit 200 is configured to determine the target sample distance based on the difference data points and the consistency data points.

[0169] The determination unit 200 is specifically configured to: calculate the sample distance between the difference data points and the consistency data points in the second preset time period to obtain each sample distance; and select the maximum sample distance from each sample distance and mark it as the target sample distance.

[0170] The second calculation unit 300 is configured to calculate the local density based on the target sample distance.

[0171] The third calculation unit 400 is configured to periodically calculate all the local densities of the difference data points and the consistency data points to obtain the target local density.

[0172] The fourth calculation unit 500 is configured to calculate a local outlier factor according to the local density and the target local density.

[0173] The confirmation unit 600 is configured to confirm that the stack voltage is in the early warning state when the local outlier factor is greater than a preset threshold.

[0174] The construction unit 700 is configured to construct a clustering index and a target clustering index according to the difference data points and the consistency data points.

[0175] The construction unit 700 is specifically configured to: construct a data set based on the difference data points and the consistency data points in a third preset time period; establish the clustering index based on the data set; when the clustering index is not the preset threshold, construct a target data set based on the difference data points and the consistency data points in a fourth preset time period; and establish the target clustering index based on the target data set.

[0176] The change unit 800 is configured to change the early warning state to a fault state when the clustering index is not the preset threshold and the target clustering index is the preset threshold.

[0177] The change unit 800 is further configured to change the early warning state to a normal operation state when the clustering index is the preset threshold.

[0178] The change unit 800 is further configured to send prompt information to a user when the target clustering index is not the preset threshold.

[0179] In summary, when the clustering index is not the preset threshold, the target data set is constructed based on the difference data points and the consistency data points in the second preset time period, the target clustering index is established based on the target data set, and the early warning state is changed to the fault state when the target clustering index is the preset threshold. Compared with the prior art, it is not necessary to manually change the pressure difference between the positive side and the negative side by a professional engineer. By judging whether the clustering index and the target clustering index are the preset threshold, the current stack condition can be known, and thus the prediction of the stack performance fault during the online operation of the real vehicle is realized.

[0180] The application further provides a computer readable storage medium, which comprises a stored program, wherein the program executes the stack fault prediction method provided by the application.

[0181] The application further provides a stack fault prediction device, which comprises a processor, a memory and a bus. The processor is connected with the memory through the bus. The memory is used for storing a program, and the processor is used for running the program. When the program is run, the stack fault prediction method provided by the application is executed, which comprises the following steps:

[0182] According to the weight value of difference, the weight value of consistency, the overall single voltage matrix, the difference data point and the consistency data point are calculated; the weight value of difference and the weight value of consistency are obtained by pre-processing based on the covariance matrix; the overall single voltage matrix is established based on each single voltage;

[0183] According to the difference data point and the consistency data point, the target sample distance is determined;

[0184] According to the target sample distance, the local density is calculated;

[0185] The local density of the difference data point and the consistency data point is periodically calculated to obtain the target local density;

[0186] According to the local density and the target local density, the local outlier factor is calculated;

[0187] When the local outlier factor is greater than a preset threshold, it is determined that the stack voltage is in a warning state;

[0188] According to the difference data point and the consistency data point, the clustering index and the target clustering index are constructed;

[0189] When the clustering index is not the preset threshold and the target clustering index is the preset threshold, the warning state is changed to a fault state.

[0190] Optionally, the process of establishing the overall single voltage matrix based on each single voltage includes:

[0191] When all the single voltages sent by the preset device are received, all the single voltages are stored in a database;

[0192] Each single voltage in a preset voltage interval in a first preset time period is obtained from the database;

[0193] According to each single voltage, the overall single voltage matrix is established.

[0194] Optionally, the process of obtaining the weight value of difference and the weight value of consistency based on the covariance matrix includes:

[0195] According to the overall single voltage matrix, each single voltage matrix is constructed;

[0196] For each single voltage matrix, the single voltage matrix is subjected to decentralization processing to obtain a decentralized single voltage matrix;

[0197] According to the decentralized single voltage matrix, the covariance matrix is established;

[0198] perform singular value decomposition on the covariance matrix to obtain the weight value of the difference and the weight value of the consistency.

[0199] Optionally, the method further comprises:

[0200] calculating sample distances between the difference data points and the consistency data points in a second preset time period to obtain each sample distance;

[0201] selecting a maximum sample distance from the sample distances and marking the maximum sample distance as a target sample distance.

[0202] Optionally, the method further comprises:

[0203] constructing a data set based on the difference data points and the consistency data points in a third preset time period;

[0204] establishing a clustering index based on the data set;

[0205] when the clustering index is not the preset threshold, constructing a target data set based on the difference data points and the consistency data points in a fourth preset time period;

[0206] establishing a target clustering index based on the target data set.

[0207] Optionally, the method further comprises:

[0208] when the clustering index is the preset threshold, changing the early warning state to a normal operation state.

[0209] Optionally, the method further comprises:

[0210] when the target clustering index is not the preset threshold, sending a prompt information to a user.

[0211] If the functions of the method are realized in the form of software function units and sold or used as independent products, the functions can be stored in a readable storage medium of a computing device. Based on this understanding, the part of the prior art that the embodiments of the present application contribute to or part of the technical solutions can be embodied in the form of a software product stored in a storage medium, including a plurality of instructions for causing a computing device (which can be a personal computer, a server, a mobile computing device, or a network device) to execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, and various storage medium that can store program codes.

[0212] The various embodiments described in this specification are presented for the purpose of illustration and description. Each of the embodiments described in this specification is presented individually for ease of understanding, and the same or similar elements in each embodiment are cross-referenced to each other.

[0213] The above description of disclosed embodiments is intended to be illustrative and not restrictive. Many modifications of these embodiments by one having ordinary skill in the art are intended to be within the scope of the application. The general principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not to be limited to the embodiments described herein but is to be accorded the widest scope consistent with the principles and novel features described herein.

Claims

1. A method for predicting fuel cell stack faults, characterized in that, include: When all individual unit voltages are received from the preset device, all individual unit voltages are stored in the database; The unit voltage indicates the unit voltage of the fuel cell stack; Obtain the voltage of each individual unit within a preset voltage range during a first preset time period from the database; Based on the individual unit voltages, establish the overall unit voltage matrix; Based on the overall individual unit voltage matrix, construct the individual unit voltage matrix; For each of the individual voltage matrices, the individual voltage matrices are decentered to obtain the decentered individual voltage matrices. Based on the decentralized individual voltage matrix, a covariance matrix is ​​established; Singular value decomposition is performed on the covariance matrix to obtain the weight values ​​for difference and consistency; where difference indicates the performance difference of the fuel cell stack, and consistency indicates the performance consistency of the fuel cell stack. Based on the weight values ​​of difference, consistency, and the overall individual voltage matrix, the difference data points and consistency data points are calculated. Calculate the sample distance between the differential data points and the consistent data points within the second preset time period to obtain each sample distance; select the largest sample distance from each sample distance and mark it as the target sample distance; The local density is calculated based on the target sample distance; Periodically calculate all local densities of the differential data points and the consistent data points within a second preset time period to obtain the target local density; The local outlier factor is calculated based on the local density and the target local density. When the local outlier factor is greater than a preset threshold, the stack voltage is confirmed to be in a warning state. A dataset is constructed based on the differential data points and the consistent data points within the third preset time period; Establish clustering metrics based on the dataset; When the clustering index is not at the preset threshold, a target dataset is constructed based on the differential data points and the consistent data points within the fourth preset time period. Establish target clustering indicators based on the target dataset; When the clustering index is not at the preset threshold and the target clustering index is at the preset threshold, the warning status is changed to a fault status.

2. The method for predicting fuel cell stack faults according to claim 1, characterized in that, Also includes: When the clustering index reaches the preset threshold, the warning status is changed to normal operation status.

3. The method for predicting fuel cell stack faults according to claim 1, characterized in that, Also includes: When the target clustering index is not at the preset threshold, a prompt message is sent to the user.

4. A device for predicting fuel cell stack faults, used to implement the method for predicting fuel cell stack faults according to any one of claims 1-3, characterized in that, include: The first calculation unit is used to calculate the difference data points and the consistency data points based on the weight values ​​of the differences, the weight values ​​of the consistency, and the overall individual voltage matrix. The determining unit is used to calculate the sample distance between the differential data points and the consistent data points within the second preset time period, and obtain each sample distance; select the largest sample distance from each sample distance and mark it as the target sample distance; The second calculation unit is used to calculate the local density based on the target sample distance; The third calculation unit is used to periodically calculate all local densities of the differential data points and the consistent data points within a second preset time period to obtain the target local density. The fourth calculation unit is used to calculate the local outlier factor based on the local density and the target local density. The confirmation unit is used to confirm that the stack voltage is in a warning state when the local outlier factor is greater than a preset threshold. The construction unit is used to construct a dataset based on the differential data points and the consistent data points within a third preset time period; Establish clustering metrics based on the dataset; When the clustering index is not at the preset threshold, a target dataset is constructed based on the differential data points and the consistent data points within the fourth preset time period. Establish target clustering indicators based on the target dataset; The modification unit is used to change the warning state to a fault state when the clustering index is not the preset threshold and the target clustering index is the preset threshold.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program is executed by a processor to perform the method for predicting stack faults according to any one of claims 1-3.

6. A device for predicting fuel cell stack faults, characterized in that, include: Processor, memory, and bus; The processor and the memory are connected via the bus; The memory is used to store a program, and the processor is used to run the program, wherein the program is executed by the processor to perform the method for predicting stack faults according to any one of claims 1-3.

Images