Electric vehicle power abnormity fault diagnosis method and device

Abstract

The invention provides an electric vehicle power abnormity fault diagnosis method and device, and relates to vehicle fault diagnosis, and the method comprises the steps: obtaining power data of a to-be-diagnosed electric vehicle; inputting the power data into a pre-trained fault diagnosis model, and outputting a fault diagnosis result of the to-be-diagnosed electric vehicle; wherein the fault diagnosis model is obtained by training a plurality of preset kernel extreme learning machines through a historical data training set, the historical data training set comprises important fault features and a historical sample set, and the important fault features are obtained by performing importance evaluation and updating on original fault features extracted from the historical sample set; determining whether the fault diagnosis result is smaller than a preset detection threshold value, if yes, determining that the fault diagnosis result meets the validity requirement and outputting the fault diagnosis result, and if not, performing fault diagnosis on the power data again. The scheme can effectively detect the power fault of the electric vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile fault diagnosis, and in particular to a method and device for diagnosing abnormal power fault of an electric vehicle. Background Art

[0002] The advancement of electric vehicles (EVs) has promoted the realization of the global carbon neutrality strategy. The number of electric vehicles has increased 1,300 times in the past 10 years. However, the complex electric vehicle systems have unique failure modes such as power failures, which pose huge safety hazards to the safe operation of electric vehicles.

[0003] In the related technologies, the power fault diagnosis of electric vehicles mainly relies on user complaints and accident investigations. This method is not only difficult to meet the actual diagnostic needs of the massively growing number of electric vehicles, but also has a large lag, which greatly reduces the operating safety of electric vehicles.

[0004] Based on this, there is an urgent need for an electric vehicle power abnormality fault diagnosis method and device to solve the above technical problems. Summary of the invention

[0005] The embodiment of the present invention provides a method and device for diagnosing abnormal power failure of an electric vehicle, which can effectively detect power failure of the electric vehicle and improve the safety of vehicle operation.

[0006] In a first aspect, an embodiment of the present invention provides a method for diagnosing abnormal power failure of an electric vehicle, comprising:

[0007] Acquire power data of the electric vehicle to be diagnosed;

[0008] The power data is input into a pre-trained fault diagnosis model, and a fault diagnosis result of the electric vehicle to be diagnosed is output; wherein the fault diagnosis model is obtained by training a preset multiple core extreme learning machine with a historical data training set, the historical data training set includes important fault features and a historical sample set, and the important fault features are obtained by updating the importance evaluation of the original fault features extracted from the historical sample set;

[0009] Determine whether the fault diagnosis result is less than a preset detection threshold. If so, determine that the fault diagnosis result meets the validity requirement and output the fault diagnosis result. Otherwise, re-diagnose the power data.

[0010] In a second aspect, an embodiment of the present invention further provides an electric vehicle power abnormality fault diagnosis device, comprising:

[0011] An acquisition module, used for acquiring power data of the electric vehicle to be diagnosed;

[0012] A diagnosis module, used for inputting the power data into a pre-trained fault diagnosis model, and outputting a fault diagnosis result of the electric vehicle to be diagnosed; wherein the fault diagnosis model is obtained by training a preset multiple core extreme learning machine through a historical data training set, the historical data training set includes important fault features and a historical sample set, and the important fault features are obtained by updating the importance evaluation of the original fault features extracted from the historical sample set;

[0013] The detection module is used to determine whether the fault diagnosis result is less than a preset detection threshold. If so, it is determined that the fault diagnosis result meets the validity requirements and outputs the fault diagnosis result. Otherwise, the power data is re-diagnosed for fault.

[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in any embodiment of this specification is implemented.

[0015] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, enables the computer to execute the method described in any embodiment of this specification.

[0016] The embodiment of the present invention provides a method and device for diagnosing abnormal power faults of electric vehicles. First, a data set containing power fault features of electric vehicles and original power data of electric vehicles is constructed. Then, the fault features of the power fault feature library are sorted by importance using an importance selection algorithm, and important fault features that are strongly related to power faults of new energy vehicles are extracted. Subsequently, a preset kernel extreme learning machine is trained using the important fault features to obtain a fault diagnosis model. Finally, the fault diagnosis model is calibrated according to a preset detection threshold to reduce the false alarm rate of diagnosis. This method can effectively detect power faults of electric vehicles, improve the safety of vehicle operation, and provide a technical basis for power fault diagnosis of new energy vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 This is a flow chart of a method for diagnosing abnormal power failure of an electric vehicle provided by an embodiment of the present invention;

[0019] Figure 2is a hardware architecture diagram of an electronic device provided by an embodiment of the present invention;

[0020] Figure 3 It is a structural diagram of an electric vehicle power abnormality fault diagnosis device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in 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.

[0022] As mentioned above, the existing electric vehicle power fault diagnosis mainly relies on user complaints and accident investigations. With the increasing number of electric vehicles in my country, the traditional electric vehicle power fault diagnosis method can no longer meet the requirements of large-scale product supervision.

[0023] Based on this, the concept of the present invention is to construct a vehicle power fault database through multiple approaches, and use the database to train multiple kernel extreme learning machines to obtain a fault diagnosis model for fault diagnosis, and further improve the diagnosis accuracy according to the detection threshold of key indicators.

[0024] The specific implementation of the above concept is described below.

[0025] Please refer to Figure 1 The embodiment of the present invention provides a method for diagnosing abnormal power failure of an electric vehicle, the method comprising:

[0026] Step 100, obtaining power data of the electric vehicle to be diagnosed;

[0027] Step 102, inputting the power data into a pre-trained fault diagnosis model, and outputting a fault diagnosis result of the electric vehicle to be diagnosed; wherein the fault diagnosis model is obtained by training a preset multiple core extreme learning machine through a historical data training set, the historical data training set includes important fault features and a historical sample set, and the important fault features are obtained by updating the importance evaluation of the original fault features extracted from the historical sample set;

[0028] Step 104, determining whether the fault diagnosis result is less than a preset detection threshold, if so, determining that the fault diagnosis result meets the validity requirement and outputting the fault diagnosis result, otherwise re-diagnosing the power data.

[0029] In the embodiment of the present invention, a data set including electric vehicle power fault features and electric vehicle power raw data is first constructed, and then the fault features of the power fault feature library are sorted by importance using an importance selection algorithm to extract important fault features that are strongly related to new energy vehicle power faults, and then the preset kernel extreme learning machine is trained using the important fault features to obtain a fault diagnosis model, and finally the fault diagnosis model is calibrated according to a preset detection threshold to reduce the false alarm rate of diagnosis. This method can effectively detect power faults of electric vehicles, improve vehicle operation safety, and provide a technical basis for power fault diagnosis of new energy vehicles.

[0030] Described below Figure 1 How the various steps are performed.

[0031] First, with respect to step 100 , power data of the electric vehicle to be diagnosed is obtained.

[0032] Power failure is a failure mode unique to electric vehicles, including system power loss and abnormal acceleration. System power loss refers to insufficient power supply when the driver attempts to operate the accelerator pedal, which manifests as slowed acceleration or even no response. In extreme cases, power loss may increase the potential risk of traffic accidents. Abnormal acceleration refers to the vehicle accelerating incorrectly when the driver operates the deceleration pedal. Both power loss and sudden acceleration pose great risks to vehicle safety.

[0033] In order to detect vehicle power failure, it is necessary to obtain the power data of the electric vehicle as shown in Table 1 below:

[0034] Table 1

[0035]

[0036]

[0037] Then, for step 102, the power data is input into a pre-trained fault diagnosis model, and a fault diagnosis result of the electric vehicle to be diagnosed is output.

[0038] In an embodiment of the present invention, a fault diagnosis model is obtained by training a preset plurality of core extreme learning machines using a historical data training set. The historical data training set includes important fault features and a historical sample set. The important fault features are updated by performing importance evaluation on the original fault features extracted from the historical sample set.

[0039] Specifically, the process of establishing the historical sample set includes utilizing the new energy vehicle cloud monitoring platform and user-related complaints, performing threshold detection on the faulty vehicle's own data, and extracting the electric vehicle power anomaly-related data as shown in Table 1 in the above steps.

[0040] Furthermore, by considering the differences between the normal vehicle data and the faulty vehicle data in the sample set in multiple dimensions including pedals, motors, batteries, insulation resistance, etc., the power fault feature extraction based on the faulty vehicle's own historical data and the normal vehicle cluster data is achieved, and the original fault feature library that can characterize the new energy vehicle power fault and can cause the new energy vehicle power fault is obtained. The feature library contains the data information shown in Table 2.

[0041] Table 2

[0042]

[0043]

[0044] Among them, for the simulated motor torque output M in Table 2 T,sim The existing fault vehicle information can be used in combination with the cloud-based motor power modeling to calculate the specific calculation formula:

[0045]

[0046] Where, P is the motor output power; M n is the motor speed; η is the motor efficiency; M V is the motor bus voltage; M I is the motor bus current.

[0047] Furthermore, a fault feature importance evaluation database is constructed by combining normal vehicle data and faulty vehicle data. The database content is shown in Table 3 below, which includes normal vehicle data, faulty vehicle data, and the time-varying values ​​of the 16 features included in Table 2. In this table, 1 frame of data represents a sample, and N represents the total number of frames of data.

[0048] Table 3

[0049]

[0050]

[0051] In the above table, category 0 represents that the sample has no power failure, and category 1 represents a power failure.

[0052] Since Tables 2 and 3 above contain a large number of original fault features, and in actual application scenarios, vehicle power failures are often reflected in several more important fault features, that is, once the data corresponding to these fault features do not match the normal data, it usually indicates that the vehicle has a power failure. Therefore, in order to improve the diagnostic accuracy and speed of the fault, it is necessary to update the importance evaluation of the original fault features to obtain important fault features.

[0053] In an embodiment of the present invention, important fault features are updated in the following manner: determining the same sample neighbors and different sample neighbors of each sample in the historical sample set according to the Euclidean distance; iteratively calculating the initial weight values ​​of the original fault features according to the same sample neighbors and the different sample neighbors to obtain the first weights of all original fault features that meet the iteration requirements; calculating the cumulative result of the normalized first weights, and determining that the fault features in the cumulative result that meet the preset weight threshold are important fault features.

[0054] Specifically, first determine the same sample neighbors and different sample neighbors of each sample in Table 3 above: the same sample neighbors and different sample neighbors can be determined by calculating the Euclidean distance, that is, in the same type of samples, the k samples with the closest Euclidean distance to Sample are regarded as the same sample neighbors; in different types of samples, the k samples with the closest Euclidean distance to Sample are regarded as different sample neighbors. The Euclidean distance calculation formula is:

[0055]

[0056] Where X, Y represent points in 16 dimensions, and D(X, Y) represents the Euclidean distance.

[0057] Furthermore, the first weight value of each original fault feature is calculated using the following formula:

[0058]

[0059] Where, T(C) is the first weight value; T init (C) is the initial weight of the original fault feature C or the previous iteration weight; Sample is the sample; H i is the nearest neighbor of the same sample of Sample; M i (C) is the nearest neighbor of different samples of Sample; p(Sample) is the probability that a sample belongs to its category; p(class(Sample)) is the probability of a category in the entire historical sample set.

[0060] Each sample is iterated and updated L times through the above formula until all samples N are iterated, and the weight value of the updated original fault feature is obtained as shown in Table 4 below. The specific number of iterations can be adjusted according to actual needs and will not be described in detail here.

[0061] Table 4

[0062]

[0063] Among them C i represents the i-th feature, T(C i ) represents C i The corresponding weight.

[0064] Further, according to the magnitude relationship of the first weights, the normalized results are accumulated and calculated in descending order to obtain an accumulated result, wherein the normalized calculation formula is:

[0065]

[0066] Finally, the original fault feature corresponding to the first weight value whose accumulated result is greater than 0.9 for the first time is determined to be an important fault feature, that is, the feature whose weighted result accounts for more than 90% is taken as the important fault feature.

[0067] After determining the important fault features, the important fault features and historical sample sets can be input into the nuclear extreme learning machine as historical data training sets for training. The training process includes the following steps: optimizing the nuclear extreme learning machine training according to the Lagrange multiplier to obtain a nuclear extreme learning machine that eliminates randomness and minimizes output deviation; inputting the historical sample set into the optimized nuclear extreme learning machine and outputting the fault classification result; initializing the weights of all samples in the historical data sample set, and training the nuclear extreme learning machine according to the weighting results and fault classification results to obtain a fault diagnosis model.

[0068] Specifically, since the extreme learning machine (ELM) is a feedforward neural network, it usually has only one hidden layer, and its initial parameters are randomly generated. The parameters of the hidden layer do not need to be adjusted, and the model is trained by minimizing the error of the output layer, thereby achieving a very fast training speed. ELM has satisfactory performance when processing large-scale data sets and can perform classification tasks efficiently. Despite its encouraging advantages such as fast training speed, low complexity, and overcoming local minima or overfitting, ELM requires more hidden neurons when processing complex nonlinear classification tasks, resulting in a highly complex network structure.

[0069] Therefore, the embodiment of the present invention introduces the kernel function into the ELM to obtain the kernel extreme learning machine (kELM), which eliminates the requirement of explicitly defining the mapping function or setting the number of hidden neurons, thereby saving time for optimizing the number of hidden neurons. The main contribution of the kernel method is to convert low-dimensional, linearly inseparable samples into a high-dimensional, non-linearly separable feature space. Compared with the traditional ELM algorithm, the kELM uses kernel mapping instead of random mapping, which effectively improves the generalization and stability problems caused by the random allocation of hidden neurons.

[0070] Furthermore, in order to eliminate the randomness generated during the training process and minimize the output result deviation, two objectives are considered to achieve the training optimization of the kernel extreme learning machine, which correspond to minimizing the output result deviation and optimizing the weight configuration, as shown in the following formula:

[0071]

[0072] In the above formula, β represents the model weight, δ i represents the deviation between the output and the true value, and N represents the number of samples.

[0073] Since the constraints involved in the function include the error between the output and the true value, as well as the regularization of the model parameters, in order to simultaneously meet these constraints and minimize the objective function during the optimization process, it is necessary to introduce the Lagrange multiplier into the above formula to become:

[0074]

[0075] In the above formula, a ij represents the Lagrange multiplier, M represents the number of neurons in the output layer, β j is the weight vector associated with the jth output neuron (or output dimension), t ij is the true target value or label of the model, x i is the input sample, h(x i ) is x i The value after mapping by the kernel function, ∈ ij represents the error term.

[0076] By solving the partial derivatives of the Lagrangian function with respect to various variables (such as weight β, bias term δ, etc.), the conditions for minimizing the Lagrangian function can be obtained.

[0077]

[0078] In the above formula, c represents the regularization parameter, whose value is between [-1,1], and H represents the hidden layer output matrix.

[0079] Since the decision function satisfies

[0080] d(x)=h(x)·β

[0081] Then the output function can be written as:

[0082]

[0083] The kernel function h(x) is written as:

[0084] h(x)=exp(-||xz|| / 2σ 2 )

[0085] Where x represents the new input sample point, z represents a sample point in the training set, and the kernel function measures the relationship between the new input x and the training sample z in the high-dimensional space by calculating the similarity between x and z; σ represents the kernel width parameter.

[0086] Since a single extreme learning machine is a weak classifier, although it can realize the diagnosis and classification of faults, the classification effect is poor. Therefore, the embodiment of the present invention proposes to introduce an Adaboost algorithm containing multiple core extreme learning machines, and improve the performance of the model by combining multiple weak classifiers into a strong classifier. The weight of each weak classifier is adjusted according to the error rate of the previous round of classifiers. The misclassified samples will get a higher weight, so that they will receive more attention in the next round of training to further improve, and finally improve the accuracy of fault diagnosis.

[0087] Specifically, firstly, all samples in the historical data sample set are weighted and initialized:

[0088]

[0089] Where D l (i) is the initial sample weight of the i-th sample; N is the total number of samples.

[0090] Furthermore, the following steps are performed to train the kernel extreme learning machine according to the weighting results and the fault classification results to obtain a fault diagnosis model:

[0091] S1, calculate the error rate ε of the historical sample set and the fault classification result of the t-th core extreme learning machine according to the sample weight t :

[0092]

[0093] Among them, y i represents the historical normal sample, h t (x i ) represents the optimization output of the t-th kernel extreme learning machine; I(y i ≠h t (x i )) is the indicator function, which is 1 when the classification is wrong, otherwise it is 0; D l (i) is the initial sample weight of the i-th sample; N is the total number of samples;

[0094] S2, calculate the initial classifier weight α of the t-th kernel extreme learning machine according to the error rate t :

[0095]

[0096] S3, update the initial sample weight of the t-th kernel extreme learning machine according to the first weight to obtain the sample update weight D l+1 (i)

[0097]

[0098] Among them, Z t represents the normalization factor;

[0099] S4, taking the sample update weight as the initial sample weight, repeating S1 to S3 until the error rate of the kernel extreme learning machine meets the preset standard and the training is completed;

[0100] S5, performing weighted processing on the classifier weights and fault classification results of all the trained nuclear extreme learning machines to obtain the final output of the fault diagnosis model.

[0101] It is worth noting that, as mentioned above, a single kernel extreme learning machine is a weak classifier. Therefore, for multiple kernel extreme learning machines, even if the input initial sample weights are exactly the same, the classification results of each kernel extreme learning machine are not exactly the same.

[0102] For example, suppose there are 5 kernel extreme learning machines and 10 samples, where the samples include 5 normal data and 5 fault data numbered 1-10. After all the weighted initialized sample data are input into the 5 kernel extreme learning machines in turn, the diagnostic identification result of each machine is obtained. At this time, S1 is executed to calculate the error rate, that is, to compare the difference between the identified fault data of each machine and the actual fault data in the sample to obtain the error rate of each machine. Then S2 is executed to calculate the weight of each machine according to the error rate. The lower the error rate, the higher the weight of the machine, that is, when the final diagnostic result is output, the higher the credibility of the machine. Then S3 is executed to readjust the weight of each sample according to the machine weight for the next round of machine training. That is, for the first machine, the results of fault diagnosis may be 1-4 and 8-10 as normal data, and 5-7 as fault data. At this time, 5 and 8-10 are samples of incorrect diagnosis. After S3 updates the sample weights, these four sample data will receive more attention in the next round of training, that is, they will be used as key sample data to train the core extreme learning machine, thereby improving the machine's diagnostic accuracy for fault data.

[0103] Whether the training of each kernel extreme learning machine is completed depends on whether the error rate of the machine meets the preset requirements. This standard can be adjusted by technicians in this field according to actual needs and will not be elaborated here. After all kernel extreme learning machines have completed training, S5 is executed, that is, weighted processing is performed by combining the classification results of all machines and the final weights of the corresponding machines as the final output of the fault diagnosis model.

[0104] For step 104, determine whether the fault diagnosis result is less than a preset detection threshold. If so, determine that the fault diagnosis result meets the validity requirement and output the fault diagnosis result. Otherwise, re-diagnose the power data.

[0105] Although the data-driven method has a certain generalization ability and can achieve good classification performance on untrained data, in engineering practice, even a very small number of false alarms will have a significant impact on after-sales work, especially for millions of electric vehicles. Therefore, regardless of whether the above-mentioned fault diagnosis model can verify the detection of power anomalies, the final threshold detection should be performed on key indicators to further improve the accuracy. In an embodiment of the present invention, motor speed, pedal-vehicle speed consistency and voltage inconsistency are selected as the basis for the effectiveness of the fault judgement result, and the threshold method is used to evaluate the effectiveness of the judgment result. The detection threshold is shown in Table 5 below:

[0106] Table 5

[0107]

[0108] It is worth noting that the validity basis of the judgment result of the above-mentioned threshold-based fault judgement device includes but is not limited to motor speed, pedal-vehicle speed consistency and voltage inconsistency, and may also include the ratio of vehicle speed to pedal stroke, the ratio of motor torque to pedal stroke, etc.

[0109] like Figure 2 , Figure 3 As shown, an embodiment of the present invention provides an electric vehicle power abnormality fault diagnosis device. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. From the hardware level, Figure 2 As shown, it is a hardware architecture diagram of an electronic device in which an abnormal power fault diagnosis device for an electric vehicle provided by an embodiment of the present invention is located. Figure 2 In addition to the processor, memory, network interface, and non-volatile memory shown, the electronic device in the embodiment may also include other hardware, such as a forwarding chip responsible for processing messages, etc. Taking software implementation as an example, Figure 3 As shown, as a device in a logical sense, the CPU of the electronic device in which it is located reads the corresponding computer program in the non-volatile memory into the internal memory and runs it. This embodiment provides an electric vehicle power abnormality fault diagnosis device, including:

[0110] The acquisition module 300 is used to acquire the power data of the electric vehicle to be diagnosed;

[0111] The diagnosis module 302 is used to input the power data into a pre-trained fault diagnosis model and output a fault diagnosis result of the electric vehicle to be diagnosed; wherein the fault diagnosis model is obtained by training a preset multiple core extreme learning machine through a historical data training set, and the historical data training set includes important fault features and a historical sample set, and the important fault features are obtained by updating the importance evaluation of the original fault features extracted from the historical sample set;

[0112] The detection module 304 is used to determine whether the fault diagnosis result is less than a preset detection threshold. If so, it is determined that the fault diagnosis result meets the validity requirement and outputs the fault diagnosis result. Otherwise, the power data is re-diagnosed.

[0113] In an embodiment of the present invention, important fault features are updated specifically in the following manner: determining the same sample neighbors and different sample neighbors of each sample in the historical sample set according to the Euclidean distance; iteratively calculating the initial weight values ​​of the original fault features according to the same sample neighbors and the different sample neighbors to obtain the first weights of all original fault features that meet the iteration requirements; calculating the cumulative result of the normalized first weights, and determining that the fault features in the cumulative result that meet the preset weight threshold are important fault features.

[0114] In the embodiment of the present invention, the first weight value is calculated by the following formula:

[0115]

[0116] Where, T init (C) is the previous iteration weight of the original fault feature C; Sample is the sample; H i is the nearest neighbor of the same sample of Sample; M i (C) is the nearest neighbor of different samples of Sample; p(Sample) is the probability that a sample belongs to its category; p(class(Sample)) is the probability of a category in the entire historical sample set.

[0117] In an embodiment of the present invention, the calculation of the accumulated result of the first weight after normalization, and determining that the fault feature in the accumulated result that meets the preset weight threshold is an important fault feature, includes: according to the size relationship of the first weight, the normalized results are accumulated in order from large to small to obtain the accumulated result; and determining that the original fault feature corresponding to the first weight value greater than 0.9 for the first time in the accumulated result is an important fault feature.

[0118] In an embodiment of the present invention, the fault diagnosis model is specifically trained in the following manner: the kernel extreme learning machine is trained and optimized according to the Lagrange multiplier to obtain a kernel extreme learning machine that eliminates randomness and minimizes output deviation; the historical sample set is input into the optimized kernel extreme learning machine, and the fault classification result is obtained as output; all samples in the historical data sample set are weighted and initialized, and the kernel extreme learning machine is trained according to the weighted results and the fault classification results to obtain the fault diagnosis model.

[0119] In the embodiment of the present invention, the method of training the kernel extreme learning machine according to the weighting result and the fault classification result to obtain the fault diagnosis model includes:

[0120] S1, calculate the error rate ε of the historical sample set and the fault classification result of the t-th core extreme learning machine according to the sample weight t :

[0121]

[0122] Among them, y i represents the historical normal sample, h t (x i ) represents the optimization output of the t-th kernel extreme learning machine; I(y i ≠h t (x i )) is the indicator function, which is 1 when the classification is wrong, otherwise it is 0; D l (i) is the initial sample weight of the i-th sample; N is the total number of samples;

[0123] S2, calculate the initial classifier weight α of the t-th kernel extreme learning machine according to the error rate t :

[0124]

[0125] S3, update the initial sample weight of the t-th kernel extreme learning machine according to the first weight to obtain the sample update weight D l+1 (i)

[0126]

[0127] Among them, Z t represents the normalization factor;

[0128] S4, taking the sample update weight as the initial sample weight, repeating S1 to S3 until the error rate of the kernel extreme learning machine meets the preset standard and the training is completed;

[0129] S5, performing weighted processing on the classifier weights and fault classification results of all the trained nuclear extreme learning machines to obtain the final output of the fault diagnosis model.

[0130] In the embodiment of the present invention, the historical sample set includes historical normal power data and historical fault power data.

[0131] It is to be understood that the structure illustrated in the embodiment of the present invention does not constitute a specific limitation on an electric vehicle power abnormality fault diagnosis device. In other embodiments of the present invention, an electric vehicle power abnormality fault diagnosis device may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0132] The information interaction, execution process and other contents between the modules in the above-mentioned device are based on the same concept as the embodiment of the method of the present invention. For the specific contents, please refer to the description in the embodiment of the method of the present invention, and no further description is given here.

[0133] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, an electric vehicle power abnormality fault diagnosis method in any embodiment of the present invention is implemented.

[0134] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the processor executes an electric vehicle power abnormality fault diagnosis method in any embodiment of the present invention.

[0135] Specifically, a system or device equipped with a storage medium can be provided, on which software program code that implements the functions of any of the above-mentioned embodiments is stored, and a computer (or CPU or MPU) of the system or device can be enabled to read and execute the program code stored in the storage medium.

[0136] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present invention.

[0137] The storage medium embodiments for providing the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded from a server computer by a communication network.

[0138] In addition, it should be clear that the functions of any of the above embodiments can be implemented not only by executing the program code read by the computer, but also by enabling an operating system operating on the computer to complete part or all of the actual operations based on instructions from the program code.

[0139] In addition, it can be understood that the program code read from the storage medium is written to a memory provided in an expansion board inserted into the computer or to a memory provided in an expansion module connected to the computer, and then based on the instructions of the program code, a CPU installed on the expansion board or expansion module is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above-mentioned embodiments.

[0140] It should be noted that, in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical factors in the process, method, article or device including the elements.

[0141] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various media that can store program codes.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for diagnosing abnormal power failure of an electric vehicle, characterized in that: include: Acquire power data of the electric vehicle to be diagnosed; The power data is input into a pre-trained fault diagnosis model, and a fault diagnosis result of the electric vehicle to be diagnosed is output; wherein the fault diagnosis model is obtained by training a preset multiple core extreme learning machine with a historical data training set, the historical data training set includes important fault features and a historical sample set, and the important fault features are obtained by updating the importance evaluation of the original fault features extracted from the historical sample set; Determine whether the fault diagnosis result is less than a preset detection threshold. If so, determine that the fault diagnosis result meets the validity requirement and output the fault diagnosis result. Otherwise, re-diagnose the power data.

2. The method according to claim 1, characterized in that The important fault characteristics are updated in the following way: Determine the same sample nearest neighbors and different sample nearest neighbors of each sample in the historical sample set according to the Euclidean distance; Iteratively calculate the initial weight value of the original fault feature according to the same sample nearest neighbor and the different sample nearest neighbor to obtain the first weight of all the original fault features that meet the iteration requirement; An accumulation result of the normalized first weights is calculated, and a fault feature in the accumulation result that meets a preset weight threshold is determined as an important fault feature.

3. The method according to claim 2, characterized in that The first weight value is calculated by the following formula: Where, T init (C) is the previous iteration weight of the original fault feature C; Sample is the sample; H i is the nearest neighbor of the same sample of Sample; M i (C) is the nearest neighbor of different samples of Sample; p(Sample) is the probability that a sample belongs to its category; p(class(Sample)) is the probability of a category in the entire historical sample set.

4. The method according to claim 3, characterized in that The calculating the accumulated result of the normalized first weights, and determining the fault features in the accumulated result that meet a preset weight threshold as important fault features, includes: According to the magnitude relationship of the first weights, the normalized results are accumulated and calculated in descending order to obtain an accumulated result; It is determined that the original fault feature corresponding to the first weight value in which the accumulated result is greater than 0.9 for the first time is an important fault feature.

5. The method according to claim 1, characterized in that The fault diagnosis model is specifically trained in the following way: The kernel extreme learning machine is trained and optimized according to Lagrange multipliers, and a kernel extreme learning machine is obtained that eliminates randomness and minimizes output deviation. The historical sample set is input into the optimized kernel extreme learning machine, and the fault classification result is output; All samples in the historical data sample set are weighted and initialized, and the nuclear extreme learning machine is trained according to the weighting results and fault classification results to obtain the fault diagnosis model.

6. The method according to claim 4, characterized in that The method of training the nuclear extreme learning machine according to the weighting result and the fault classification result to obtain the fault diagnosis model includes: S1, calculate the error rate ε of the historical sample set and the fault classification result of the t-th core extreme learning machine according to the sample weight t : Among them, y i represents the historical normal sample, h t (x i ) represents the optimization output of the t-th kernel extreme learning machine; I(y i ≠h t (x i )) is the indicator function, which is 1 when the classification is wrong, otherwise it is 0; D l (i) is the initial sample weight of the i-th sample; N is the total number of samples; S2, calculate the initial classifier weight α of the t-th kernel extreme learning machine according to the error rate t : S3, update the initial sample weight of the t-th kernel extreme learning machine according to the first weight to obtain the sample update weight D l+1 (i) Among them, Z t represents the normalization factor; S4, taking the sample update weight as the initial sample weight, repeating S1 to S3 until the error rate of the kernel extreme learning machine meets the preset standard and the training is completed; S5, performing weighted processing on the classifier weights and fault classification results of all the trained nuclear extreme learning machines to obtain the final output of the fault diagnosis model.

7. The method according to claim 1, characterized in that The historical sample set includes historical normal power data and historical fault power data.

8. An electric vehicle power abnormality fault diagnosis device, characterized in that: include: An acquisition module, used for acquiring power data of the electric vehicle to be diagnosed; A diagnosis module, used for inputting the power data into a pre-trained fault diagnosis model, and outputting a fault diagnosis result of the electric vehicle to be diagnosed; wherein the fault diagnosis model is obtained by training a preset multiple core extreme learning machine through a historical data training set, the historical data training set includes important fault features and a historical sample set, and the important fault features are obtained by updating the importance evaluation of the original fault features extracted from the historical sample set; The detection module is used to determine whether the fault diagnosis result is less than a preset detection threshold. If so, it is determined that the fault diagnosis result meets the validity requirements and outputs the fault diagnosis result. Otherwise, the power data is re-diagnosed for fault.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1 to 7.

Images