Heavy-load group train battery fault monitoring method, system and equipment

Abstract

The invention discloses a heavy-load group train battery fault monitoring method, which comprises the following steps of: obtaining historical data of a normal battery and preprocessing the historical data to obtain normalized historical data; a LightGBM-ELM auto-encoder model is constructed, and the auto-encoder model is trained based on the normalized historical data; obtaining real-time data of the vehicle-mounted battery, and carrying out feature standardization on the real-time data to obtain normalized real-time data; inputting the normalized real-time data into the trained auto-encoder model to obtain reconstruction data corresponding to the real-time data, and judging whether a reconstruction error between the real-time data and the corresponding reconstruction data reaches an abnormal standard or not based on an abnormal judgment threshold; according to the method, whether the difference between the real-time data and the reconstruction data output by the auto-encoder model reaches the abnormity judgment threshold value is judged through the trained auto-encoder model, the abnormal data which is not consistent with the normal battery is quickly identified, and the real-time detection of the battery fault is realized.

Description

Technical Field

[0001] The present invention relates to the field of battery management technology, and in particular to a method, system and device for monitoring battery failure in a heavy-load group train. Background Art

[0002] At present, during the operation of heavy-load trains, the power supply for on-board auxiliary devices is usually provided by on-board batteries. As energy storage devices, on-board batteries face many safety hazards during long-term use, especially the risk of fire and explosion, which pose a serious threat to train operation. Therefore, how to detect potential safety risks in time before the battery fails irreversibly has become an important issue that needs to be urgently solved in the field of safe operation of heavy-load trains.

[0003] The patent with publication number CN207809424U discloses an intelligent monitoring and alarm device for heavy-duty trains. By collecting battery voltage data and sending it to the CPU control analysis circuit, it realizes preliminary monitoring of the battery status. However, the device does not provide a specific on-board battery abnormality detection strategy, and cannot comprehensively evaluate the internal health status of the battery, so its monitoring effect is limited; the patent with publication number CN118534396A discloses a new energy vehicle battery fault detection method. By considering the impact of external environmental factors on various battery parameters, the battery value is calibrated and compared with the preset health indicator threshold to determine whether the battery is faulty. However, this method is highly dependent on manually read data for display and judgment, which is not only inefficient, but also difficult to capture small changes in battery performance over time or sudden failures.

[0004] In summary, those skilled in the art are in urgent need of a method, system and device for monitoring battery failure in heavy-load group trains. Summary of the invention

[0005] In response to the above-mentioned problems in the prior art, the purpose of the present invention is to provide a battery fault monitoring method, system and equipment for heavy-load group trains. By constructing a LightGBM-ELM autoencoder model to learn the normal battery voltage law, and comparing the reconstruction error between the input data and the reconstructed data of the autoencoder model to see if it meets the battery abnormality standard, the fault detection of the on-board battery can be achieved.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme:

[0007] In a first aspect, an embodiment of the present invention provides a method for monitoring battery failure of a heavy-load group train, comprising:

[0008] S1: Obtain historical data of normal batteries, perform data preprocessing and feature standardization on them to obtain normalized historical data;

[0009] S2: Build a LightGBM-ELM autoencoder model, and train the LightGBM-ELM autoencoder model based on the normalized historical data, so that the trained autoencoder model outputs reconstructed data that conforms to normal battery characteristics;

[0010] The LightGBM-ELM autoencoder model includes an encoder built based on the LightGBM model and a decoder built based on the ELM model;

[0011] S3: Obtain real-time data of the vehicle battery, perform data cleaning and feature standardization on it to obtain standardized real-time data;

[0012] S4: Setting the abnormality determination threshold;

[0013] S5: Input the normalized real-time data into the trained LightGBM-ELM autoencoder model to obtain the reconstructed data corresponding to the real-time data, and based on the abnormality judgment threshold, determine whether the reconstruction error between the real-time data and its corresponding reconstructed data meets the abnormality standard.

[0014] Optionally, S2 includes the following specific steps:

[0015] S101: Acquire historical data of normal batteries, and perform data synchronization and data cleaning on the historical data;

[0016] The data synchronization is to add a timestamp to each historical data record and align the historical data at different times;

[0017] S102: extract the mean and standard deviation of historical data;

[0018] S103: normalizing the historical data;

[0019] S104: Perform Z-Score standardization on the normalized historical data to generate standardized historical data.

[0020] Optionally, the function expression of the normalization process in S103 is:

[0021]

[0022] In the formula, x' is the normalized historical data; x is the historical data; min(x) is the minimum value of the historical data, and max(x) is the maximum value of the historical data;

[0023] The function expression of Z-Score standardization in S104 is:

[0024]

[0025] Where X is the normalized historical data; μ is the mean of the historical data, and σ is the standard deviation of the historical data.

[0026] Optionally, S2 includes the following specific steps:

[0027] S201: Build LightGBM-ELM autoencoder model;

[0028] S202: The encoder built based on the LightGBM model processes the normalized historical data. Each decision tree in the LightGBM model is split layer by layer according to the characteristics of the historical data until the historical data reaches the leaf node of the decision tree. The leaf node index of the historical data in each decision tree is recorded to form a multi-dimensional feature vector as a low-dimensional feature representation of the normalized historical data.

[0029] S203: randomly initialize the weights and biases of the decoder built based on the ELM model, perform nonlinear transformation on the low-dimensional feature representation through the initialized decoder, calculate the hidden layer output, and after obtaining the output weight with the goal of minimizing the loss function, map the hidden layer output back to the original data space to generate reconstructed data corresponding to the normalized historical data;

[0030] S204: Calculate a reconstruction error based on the historical data and its corresponding reconstructed data, and adjust the parameters of the autoencoder model based on the reconstruction error, and continue iterative training until the reconstruction error between the historical data output by the autoencoder model and its corresponding reconstructed data is less than a preset error threshold.

[0031] Optionally, the encoder constructed based on the LightGBM model in S201 has a function expression as follows:

[0032]

[0033] In the formula, x i To normalize historical data, is the encoder output; T is the total number of decision trees; F T (x i ) is the prediction result of T decision trees, F0(x i ) is the initial model output; η is the learning rate or iterative step size, f t is the t-th decision tree for x i The prediction result; m is the complexity of the decision tree; depth is the maximum depth of the decision tree;

[0034] The expression of the low-dimensional feature representation in S202 is:

[0035] Z=v(x i )=[v1(x i),v2(x i ),…,vT(x i )]

[0036] Where Z is a low-dimensional representation; v(x i ) is a multidimensional feature vector; vT(x i ) is the leaf node index of the normalized historical data in the Tth decision tree.

[0037] Optionally, the function expression for reconstructing the data in S203 is:

[0038]

[0039] In the formula, is the reconstructed data, g is the activation function, W is the input weight, b is the randomly generated bias, and β represents the output weight;

[0040] The expression of the loss function in S203 is:

[0041]

[0042] Where L is the loss function, is the reconstructed data corresponding to the normalized historical data, and n is the number of samples of the normalized historical data;

[0043] The function expression of the reconstruction error in S204 is:

[0044]

[0045] In the formula, e i is the reconstruction error between the historical data and its corresponding reconstructed data.

[0046] Optionally, in S5, it is determined whether the reconstruction error between the real-time data and the corresponding reconstructed data reaches an abnormal standard, as follows:

[0047]

[0048] In the formula, e d is the reconstruction error between the real-time data and its corresponding reconstructed data; S is the abnormality judgment threshold.

[0049] In a second aspect, an embodiment of the present invention provides a battery failure detection system for a heavy-load group train, comprising:

[0050] A historical data acquisition module is used to obtain historical data of normal batteries, perform data preprocessing and feature standardization on them, so as to obtain normalized historical data;

[0051] Detection model construction module: used to build the LightGBM-ELM autoencoder model and train the LightGBM-ELM autoencoder model based on normalized historical data so that the trained autoencoder model outputs reconstructed data that conforms to normal battery characteristics;

[0052] Real-time data acquisition module, used to obtain real-time data of vehicle batteries, perform data cleaning and feature standardization to obtain standardized real-time data;

[0053] A threshold setting module, used to set an abnormality determination threshold;

[0054] The anomaly detection module is used to input the normalized real-time data into the trained LightGBM-ELM autoencoder model to obtain the reconstructed data corresponding to the real-time data, and based on the anomaly judgment threshold, determine whether the reconstruction error between the real-time data and its corresponding reconstructed data meets the anomaly standard.

[0055] In the third aspect, an embodiment of the present invention provides a battery fault detection device for a heavy-load group train, comprising at least one control processor and a memory for communicating with the at least one control processor; the memory stores instructions that can be executed by the at least one control processor, and the instructions are executed by the at least one control processor so that the at least one control processor can execute the heavy-load group train battery fault detection method as described above.

[0056] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the heavy-load group train battery failure detection method as described above.

[0057] The beneficial effects brought by the embodiments provided by the present invention include:

[0058] After constructing the LightGBM-ELM autoencoder model through the LightGBM algorithm and the ELM algorithm, the present invention uses the normal battery historical data to train the autoencoder model, and collects real-time data to input into the trained autoencoder model. By judging whether the difference between the real-time data and the reconstructed data output by the autoencoder model reaches the abnormality judgment threshold, the abnormal data that does not match the normal battery can be quickly identified, thereby realizing real-time detection of battery failures and improving the safety and reliability of train operation.

[0059] The present invention can continuously monitor the vehicle battery through the trained LightGBM-ELM autoencoder model, and can send out early warning signals at the early stage of battery failure, so that maintenance personnel have enough time to take measures to intervene and avoid further expansion of the failure or more serious consequences. At the same time, through automated monitoring methods, the frequency and difficulty of manual inspections can be reduced, thereby reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] 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 only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0061] Figure 1 A flow chart of a method for monitoring battery failure in a heavy-load group train according to an embodiment of the present specification is shown;

[0062] Figure 2 A flowchart of normalization processing of historical data in a method for monitoring battery failure of a heavy-load group train according to an embodiment of the present specification is shown;

[0063] Figure 3 A flowchart of real-time data normalization processing in a method for monitoring battery failure in a heavy-load group train according to an embodiment of this specification is shown;

[0064] Figure 4 A structural block diagram of a battery fault monitoring system for a heavy-load group train according to an embodiment of the present specification is shown. DETAILED DESCRIPTION

[0065] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0066] However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, the description of the known technology is omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0067] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The term "comprising" used herein indicates the existence of features, steps, operations, but does not exclude the existence or addition of one or more other features. It should be noted that all terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used here should be interpreted as having meanings consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0068] Example 1

[0069] Figure 1 A flow chart of a method for monitoring battery failure in a heavy-load group train according to an embodiment of the present specification is shown;

[0070] like Figure 1 As shown, this embodiment provides a method for monitoring battery failure of a heavy-load group train, comprising:

[0071] S1: Obtain historical data of normal batteries, perform data preprocessing and feature standardization on them to obtain normalized historical data;

[0072] Exemplarily, the historical data includes battery voltage; wherein the sampling frequency of normal battery voltage is 5 times per second.

[0073] Figure 2 A flowchart of normalization processing of historical data in a method for monitoring battery failure of a heavy-load group train according to an embodiment of this specification is shown;

[0074] like Figure 2 As shown, in some embodiments, S1 includes the following specific steps:

[0075] S101: Synchronize and clean historical data;

[0076] Specifically, data synchronization refers to adding a timestamp to each historical data record and aligning the historical data at different times; data cleaning refers to removing invalid values ​​in the original data.

[0077] S102: extract the mean and standard deviation of historical data;

[0078] S103: Perform Min-Max normalization processing on historical data;

[0079] S104: Perform Z-Score standardization on the normalized historical data, convert the historical data into a standard normal distribution with a mean of 0 and a standard deviation of 1, and generate standardized historical data.

[0080] Exemplarily, the function expression of the normalization process in S103 is:

[0081]

[0082] In the formula, x' is the normalized historical data; x is the historical data; min(x) is the minimum value of the historical data, and max(x) is the maximum value of the historical data;

[0083] Exemplarily, the function expression of Z-Score standardization in S104 is:

[0084]

[0085] Where X is the normalized historical data; μ is the mean of the historical data, and σ is the standard deviation of the historical data.

[0086] S2: Build a LightGBM-ELM autoencoder model, and train the LightGBM-ELM autoencoder model based on the normalized historical data, so that the trained autoencoder model outputs reconstructed data that conforms to normal battery characteristics;

[0087] In some embodiments, S2 includes the following specific steps:

[0088] S201: Build LightGBM-ELM autoencoder model;

[0089] Exemplarily, the LightGBM-ELM autoencoder model includes an encoder constructed based on the LightGBM model and a decoder constructed based on the ELM model;

[0090] It should be noted that the LightGBM (Light Gradient Boosting Machine) model in this embodiment is a highly influential and widely used gradient boosting decision tree (Gradient Boosting Decision Tree, GBDT) algorithm, which is mainly used to solve machine learning tasks such as classification, regression and sorting; the ELM (Extreme Learning Machine) model refers to an extreme learning machine model, which belongs to an artificial neural network learning algorithm and is mainly used to solve various machine learning problems, including but not limited to classification, regression, clustering and feature learning.

[0091] S202: The encoder built based on the LightGBM model processes the normalized historical data. Each decision tree in the LightGBM model is split layer by layer according to the characteristics of the historical data until the historical data reaches the leaf node of the decision tree. The leaf node index of the historical data in each decision tree is recorded to form a multi-dimensional feature vector as a low-dimensional feature representation of the normalized historical data.

[0092] The specific steps of obtaining low-dimensional feature representation and iteratively training encoder output by the encoder built based on the LightGBM model in this embodiment are as follows:

[0093] First, determine an initial model output, which is usually the average of the actual target values ​​in the input data, to provide a benchmark for subsequent iterations so that subsequent tree models can be optimized on this basis;

[0094] In the LightGBM model, when each input data passes through a decision tree, the tree will split layer by layer according to the characteristics of the input data until the input data reaches a leaf node of the tree and obtains the index corresponding to the leaf node. The index of this leaf node is unique and identifies the position of this sample in the tree; at the same time, the prediction result of each tree is based on the prediction of the target value at the position corresponding to the leaf node index;

[0095] Through multiple rounds of iterative training, each round of iteration builds a new decision tree based on the difference between the current model prediction result and the actual target value, and adds the decision trees to the model in turn, taking into account the maximum depth and complexity limits of the decision tree; as the iteration proceeds, the prediction result of each tree is multiplied by a learning rate to guide the learning of the next tree, and then accumulated to the current model prediction to update the model. This process is repeated until the predetermined number of iterations is reached or the model performance is no longer significantly improved;

[0096] Finally, the prediction results of all trees are merged to form the final prediction of the input data. At the same time, the combination of the leaf node indexes of the input data in all trees forms a T-dimensional vector, which is the low-dimensional feature representation of the input data.

[0097] For example, the encoder built based on the LightGBM model has the following function expression:

[0098]

[0099] In the formula, x i To normalize historical data, is the encoder output; T is the total number of decision trees; F T (x i ) is the prediction result of T decision trees, F0(x i ) is the initial model output; η is the learning rate or iterative step size, f t is the t-th decision tree for x i The prediction result; m is the complexity of the decision tree; depth is the maximum depth of the decision tree;

[0100] Exemplarily, the expression of low-dimensional feature representation is:

[0101] Z=v(x i )=[v1(x i ),v2(x i ),…,vT(x i )]

[0102] Where Z is a low-dimensional feature representation; v(x i ) is a T-dimensional feature vector; vT(x i ) is the leaf node index of the normalized historical data in the Tth tree.

[0103] S203: randomly initialize the weights and biases of the decoder built based on the ELM model, perform nonlinear transformation on the low-dimensional feature representation through the initialized decoder, calculate the hidden layer output, and after obtaining the output weight with the goal of minimizing the loss function, map the hidden layer output back to the original data space to generate reconstructed data corresponding to the normalized historical data;

[0104] Exemplarily, the function expression of reconstructing data in S203 is:

[0105]

[0106] In the formula, is the reconstructed data, g is the activation function, W is the input weight, b is the randomly generated bias, and β represents the output weight;

[0107] Exemplarily, the expression of the loss function in S203 is:

[0108]

[0109] Where L is the loss function, is the reconstructed data corresponding to the normalized historical data, and n is the number of samples of the normalized historical data;

[0110] S204: Calculate a reconstruction error based on the historical data and its corresponding reconstructed data, and adjust the parameters of the autoencoder model based on the reconstruction error, and continue iterative training until the reconstruction error between the historical data output by the autoencoder model and its corresponding reconstructed data is less than a preset error threshold.

[0111] In some embodiments, the parameters of the autoencoder model include encoder parameters, decoder parameters, and the number of iterative training of the autoencoder model;

[0112] The encoder parameters include the number of decision trees, the maximum depth of the decision tree, the complexity of the decision tree, and the learning rate; the decoder parameters include the choice of activation function and the number of neurons;

[0113] Exemplarily, the function expression of the reconstruction error in S204 is:

[0114]

[0115] In the formula, e i is the reconstruction error between the historical data and its corresponding reconstructed data.

[0116] In some embodiments, the preset error threshold can be set according to the training accuracy requirement.

[0117] S3: Obtain real-time data of the vehicle battery, perform data cleaning and feature standardization on it to obtain standardized real-time data;

[0118] Figure 3 A flowchart of real-time data normalization processing in a method for monitoring battery failure in a heavy-load group train according to an embodiment of this specification is shown;

[0119] Exemplarily, the real-time data includes battery voltage; wherein the sampling frequency of the battery voltage is 5 times per second.

[0120] like Figure 3 As shown, in some embodiments, S3 includes the following specific steps:

[0121] S301: collecting the road terminal voltage of the vehicle battery in real time to obtain real-time data of the vehicle battery;

[0122] S302: Clean the real-time data to remove invalid values ​​in the real-time data;

[0123] S303: extracting the mean and standard deviation of real-time data;

[0124] S304: Perform Min-Max normalization processing on the real-time data;

[0125] S305: Perform Z-Score standardization on the normalized real-time data, convert the real-time data into a standard normal distribution with a mean of 0 and a standard deviation of 1, and generate standardized real-time data.

[0126] S4: Setting the abnormality determination threshold;

[0127] The abnormality determination threshold in this embodiment is to first observe the normal fluctuation range of normal battery charging data by analyzing historical data, and then analyze real-time data, taking into account the possible influence of vehicle battery usage time, charging times and temperature and humidity of the environment where the battery is located, and then set the abnormality determination threshold, and determine that when the reconstruction error between the battery data and its corresponding reconstruction data exceeds this range, it is considered abnormal. At the same time, the setting of the abnormality determination threshold can also be set only based on historical data, that is, the step of setting the abnormality determination threshold can be set before step S3 in this embodiment.

[0128] It should be noted that, in some embodiments, the abnormality determination threshold can also be selectively adjusted according to actual conditions to adapt to the requirements of different battery types and charging environments, thereby improving the accuracy and reliability of battery abnormality detection.

[0129] S5: Input the normalized real-time data into the trained LightGBM-ELM autoencoder model to obtain the reconstructed data corresponding to the real-time data, and based on the abnormality judgment threshold, determine whether the reconstruction error between the real-time data and its corresponding reconstructed data meets the abnormality standard.

[0130] Exemplarily, in S5, it is determined whether the reconstruction error between the real-time data and the corresponding reconstructed data reaches the abnormal standard, as follows:

[0131]

[0132]

[0133] In the formula, e d is the reconstruction error between the real-time data and its corresponding reconstructed data; S is the abnormality judgment threshold; x d To normalize real-time data, The reconstructed data corresponding to the normalized real-time data.

[0134] In this embodiment, if the reconstruction error between the real-time data and the reconstructed real-time data is greater than the abnormality determination threshold, the vehicle battery is determined to be abnormal, the vehicle battery is marked as abnormal, and the user is reminded to replace the battery in time.

[0135] In some embodiments, the historical data also includes battery temperature, state of charge, number of charges, and number of battery cycles;

[0136] In some embodiments, the real-time data also includes battery temperature, state of charge, number of charges, and number of battery cycles.

[0137] Example 2

[0138] Figure 4The structure diagram of the battery fault monitoring system for heavy-load group trains according to the embodiment of this specification is shown.

[0139] like Figure 4 As shown, this embodiment provides a heavy-load group train battery fault monitoring system 100, including:

[0140] A historical data acquisition module 101 is used to obtain historical data of normal batteries, perform data preprocessing and feature standardization on the data, so as to obtain normalized historical data;

[0141] A detection model building module 102 is used to build a LightGBM-ELM autoencoder model, and train the LightGBM-ELM autoencoder model based on normalized historical data, so that the trained autoencoder model outputs reconstructed data that conforms to normal battery characteristics;

[0142] A real-time data acquisition module 103 is used to obtain real-time data of the vehicle battery, perform data cleaning and feature standardization on the data, so as to obtain standardized real-time data;

[0143] A threshold setting module 104, used to set an abnormality determination threshold;

[0144] The anomaly detection module 105 is used to input the normalized real-time data into the trained LightGBM-ELM autoencoder model to obtain the reconstructed data corresponding to the real-time data, and based on the anomaly judgment threshold, determine whether the reconstruction error between the real-time data and its corresponding reconstructed data reaches the anomaly standard.

[0145] Example 3

[0146] This embodiment provides a heavy-load group train battery failure monitoring device, including at least one control processor and a memory for communication connection with the at least one control processor;

[0147] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0148] The non-transient software program and instructions required to implement the above-mentioned embodiment of a heavy-load group train battery fault monitoring method are stored in the memory. When executed by the processor, the above-mentioned embodiment of a heavy-load group train battery fault monitoring method is executed, for example, the above-mentioned Figure 1 The method comprises steps S1 to S5.

[0149] The system embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0150] Example 4

[0151] This embodiment provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute a heavy-load group train battery failure monitoring method as described in Example 1.

[0152] It should be noted that the computer-readable storage medium of the present embodiment may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0153] In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device, or device. In this embodiment, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable storage medium other than a computer-readable storage medium, which may send, propagate, or transmit a program used by or in combination with an instruction execution system, device, or device. The computer program contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0154] The computer readable storage medium can be written in one or more programming languages ​​or a combination thereof to execute the computer program of the present embodiment, and the programming language includes an object-oriented programming language, such as Python, Java, C++, and also includes a conventional procedural programming language, such as C language or similar programming languages. The program can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0155] In summary, the LightGBM-ELM autoencoder model in this embodiment, based on the unsupervised learning method of the autoencoder, can automatically extract key features from the preprocessed historical data, and can learn the characteristics of normal batteries from a large amount of historical data in a short time. These features can fully reflect the normal behavior of the battery. This feature learning ability not only improves the generalization ability of the model, but also enables the model to more accurately identify abnormal situations that deviate from the normal mode; the LightGBM-ELM autoencoder model in this embodiment, by combining the LightGBM model to process large-scale data sets and the ELM model to quickly learn features and train quickly, can quickly detect whether the on-board battery is faulty. The LightGBM-ELM autoencoder model performs well in prediction accuracy, robustness and real-time performance, and has significant advantages in processing nonlinear relationships and reducing the risk of overfitting; the LightGBM-ELM autoencoder model proposed in this embodiment has lower maintenance costs than traditional battery anomaly detection methods, and the model can automatically detect anomalies without manual intervention or frequent parameter adjustment. In addition, due to the high accuracy and reliability of the model, the increase in maintenance costs caused by false positives or false negatives can be significantly reduced.

[0156] The above description is only a preferred embodiment of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention.

Claims

1. A method for monitoring battery failure of a heavy-load group train, characterized in that: include: S1: Obtain historical data of normal batteries, perform data preprocessing and feature standardization on them to obtain normalized historical data; S2: Build a LightGBM-ELM autoencoder model, and train the LightGBM-ELM autoencoder model based on the normalized historical data, so that the trained autoencoder model outputs reconstructed data that conforms to normal battery characteristics; The LightGBM-ELM autoencoder model includes an encoder built based on the LightGBM model and a decoder built based on the ELM model; S3: Obtain real-time data of the vehicle battery, perform data cleaning and feature standardization on it to obtain standardized real-time data; S4: Setting the abnormality determination threshold; S5: Input the normalized real-time data into the trained LightGBM-ELM autoencoder model to obtain the reconstructed data corresponding to the real-time data, and based on the abnormality judgment threshold, determine whether the reconstruction error between the real-time data and its corresponding reconstructed data meets the abnormality standard.

2. The method according to claim 1, characterized in that The S1 comprises the following specific steps: S101: Acquire historical data of normal batteries, and perform data synchronization and data cleaning on the historical data; The data synchronization is to add a timestamp to each historical data record and align the historical data at different times; S102: extract the mean and standard deviation of historical data; S103: normalizing the historical data; S104: Perform Z-Score standardization on the normalized historical data to generate standardized historical data.

3. The method according to claim 2, characterized in that The function expression of the normalization process in S103 is: In the formula, x' is the normalized historical data; x is the historical data; min(x) is the minimum value of the historical data, and max(x) is the maximum value of the historical data; The function expression of Z-Score standardization in S104 is: Where X is the normalized historical data; μ is the mean of the historical data, and σ is the standard deviation of the historical data.

4. The method according to claim 1, characterized in that: The S2 comprises the following specific steps: S201: Build LightGBM-ELM autoencoder model; S202: The encoder built based on the LightGBM model processes the normalized historical data. Each decision tree in the LightGBM model is split layer by layer according to the characteristics of the historical data until the historical data reaches the leaf node of the decision tree. The leaf node index of the historical data in each decision tree is recorded to form a multi-dimensional feature vector as a low-dimensional feature representation of the normalized historical data. S203: randomly initialize the weights and biases of the decoder built based on the ELM model, perform nonlinear transformation on the low-dimensional feature representation through the initialized decoder, calculate the hidden layer output, and after obtaining the output weight with the goal of minimizing the loss function, map the hidden layer output back to the original data space to generate reconstructed data corresponding to the normalized historical data; S204: Calculate a reconstruction error based on the historical data and its corresponding reconstructed data, and adjust the parameters of the autoencoder model based on the reconstruction error, and continue iterative training until the reconstruction error between the historical data output by the autoencoder model and its corresponding reconstructed data is less than a preset error threshold.

5. The method according to claim 4, characterized in that The encoder constructed based on the LightGBM model in S201 has a function expression as follows: In the formula, x i To normalize historical data, is the encoder output; T is the total number of decision trees; F T (x i ) is the prediction result of T decision trees, F0(x i ) is the initial model output; η is the learning rate or iterative step size, f t is the t-th decision tree for x i The prediction result; m is the complexity of the decision tree; depth is the maximum depth of the decision tree; The expression of the low-dimensional feature representation in S202 is: Z=v(x i )=[v1(x i ),v2(x i ),…,vT(x i )] Where Z is a low-dimensional feature representation; v(x i ) is a multidimensional feature vector; vT(x i ) is the leaf node index of the normalized historical data in the Tth decision tree.

6. The method according to claim 4, characterized in that The function expression of reconstructing the data in S203 is: In the formula, is the reconstructed data, g is the activation function, W is the input weight, b is the randomly generated bias, and β represents the output weight; The expression of the loss function in S203 is: Where L is the loss function, is the reconstructed data corresponding to the normalized historical data, and n is the number of samples of the normalized historical data; The function expression of the reconstruction error in S204 is: In the formula, e i is the reconstruction error between the historical data and its corresponding reconstructed data.

7. The method according to claim 1, characterized in that In S5, it is determined whether the reconstruction error between the real-time data and its corresponding reconstructed data reaches the abnormal standard, as shown in the following formula: In the formula, e d is the reconstruction error between the real-time data and its corresponding reconstructed data; S is the abnormality judgment threshold.

8. A battery fault detection system for a heavy-load group train, characterized in that: include: A historical data acquisition module is used to obtain historical data of normal batteries, perform data preprocessing and feature standardization on them, so as to obtain normalized historical data; Detection model construction module: used to build the LightGBM-ELM autoencoder model and train the LightGBM-ELM autoencoder model based on normalized historical data so that the trained autoencoder model outputs reconstructed data that conforms to normal battery characteristics; Real-time data acquisition module, used to obtain real-time data of vehicle batteries, perform data cleaning and feature standardization to obtain standardized real-time data; A threshold setting module, used to set an abnormality determination threshold; The anomaly detection module is used to input the normalized real-time data into the trained LightGBM-ELM autoencoder model to obtain the reconstructed data corresponding to the real-time data, and based on the anomaly judgment threshold, determine whether the reconstruction error between the real-time data and its corresponding reconstructed data meets the anomaly standard.

9. A battery fault detection device for a heavy-load group train, characterized in that: It includes at least one control processor and a memory for communicating with the at least one control processor; the memory stores instructions that can be executed by the at least one control processor, and the instructions are executed by the at least one control processor to enable the at least one control processor to execute the heavy-load group train battery fault detection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the heavy-load group train battery failure detection method as described in any one of claims 1 to 7.

Images