Energy storage battery weak supervision learning model training method, device and related equipment

By employing a weakly supervised learning model training method and utilizing multilayer perceptron and clustering analysis techniques, the annotation cost of energy storage battery fault diagnosis is reduced, the accuracy and adaptability of diagnosis are improved, and the problems of high cost and time consumption in traditional methods are solved.

CN118625155BActive Publication Date: 2024-11-29XIANGJIANG LAB +3
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411104544.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-29
Estimated Expiration
2044-08-13

AI Technical Summary

Technical Problem

Traditional energy storage battery fault diagnosis methods rely on supervised machine learning, which requires a large amount of labeled data, resulting in high diagnosis costs and time-consuming model training, making it difficult to adapt to new fault conditions.

Method used

We employ a weakly supervised learning model training method. We use a multilayer perceptron to compress and cluster battery operation data, and use the K nearest neighbors of the cluster centers to label the data, thereby obtaining a small amount of training data with label information and gradually optimizing the model.

Benefits of technology

It reduces the labeling cost of fault diagnosis for energy storage batteries, improves the accuracy of fault diagnosis and the ability to adapt to new faults, and reduces model training time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118625155B_ABST
    Figure CN118625155B_ABST
Patent Text Reader

Abstract

The present invention discloses a training method, device, computer equipment and storage medium for a weakly supervised learning model of an energy storage battery. The method comprises: obtaining battery operation data of a preset time period, wherein the battery operation data is weakly supervised data; based on a multi-layer perceptron of an initial weakly supervised learning model of an energy storage battery, compressing the battery operation data to obtain a latent space; clustering the battery operation data according to the latent space to obtain a cluster center, wherein the cluster center includes a label; labeling K nearest neighbors of the cluster center, and determining training data according to the obtained label results, wherein K is a positive integer; based on the training data, training the initial weakly supervised learning model of the energy storage battery until an objective function meets a preset condition, thereby obtaining a weakly supervised learning model of the energy storage battery. The present invention is used to reduce the diagnostic cost of energy storage battery fault diagnosis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of energy storage battery fault detection, and in particular to a weakly supervised learning model training method and device for energy storage batteries and related equipment. Background Art

[0002] As the world's energy demand grows and the reliance on renewable energy increases, energy storage battery packs, as new energy storage devices, are gaining more and more attention. Energy storage battery packs are composed of energy storage batteries, which can store electrical energy and release it when needed, providing a stable power supply for the power system. They are used in new energy storage systems, electric vehicles, drones and other fields. However, energy storage battery packs are prone to various faults during use, such as damage to battery cells, increased internal resistance of batteries, abnormal battery temperature, etc. These faults not only affect the performance and life of the battery pack, but may also cause equipment damage and safety accidents. Therefore, the research and development of energy storage battery fault diagnosis technology is of great significance.

[0003] Traditional energy storage battery fault diagnosis methods are usually based on supervised machine learning technology, which requires a large amount of labeled data to train the model. However, in actual application, labeled data is difficult to obtain and the labeling cost is high, resulting in high cost for energy storage battery fault diagnosis. In addition, every time the equipment has new operating data, the traditional method also needs to retrain the model to adapt to the new fault situation. This method requires a lot of model training costs and time consumption to a large extent, resulting in high cost of fault diagnosis.

[0004] Therefore, the existing energy storage battery fault diagnosis method has the technical problem of high diagnosis cost. Summary of the invention

[0005] The embodiments of the present invention provide a method, device, computer equipment and storage medium for training a weakly supervised learning model of an energy storage battery, so as to reduce the diagnostic cost of energy storage battery fault diagnosis.

[0006] In order to solve the above technical problems, the present application embodiment provides a weakly supervised learning model training method for energy storage batteries, comprising:

[0007] Acquire battery operation data for a preset time period, wherein the battery operation data is weakly supervised data;

[0008] Based on a multi-layer perceptron of a weakly supervised learning model of an initial energy storage battery, data compression is performed on the battery operation data to obtain a latent space, wherein the latent space is used to characterize the characteristics of the battery operation data;

[0009] Clustering the battery operation data according to the latent space to obtain cluster centers, wherein the cluster centers include labels;

[0010] Labeling the K nearest neighbor points of the cluster center, and determining the training data according to the obtained labeling results, where K is a positive integer;

[0011] Based on the training data, the initial energy storage battery weak supervision learning model is trained until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model.

[0012] In order to solve the above technical problems, the embodiment of the present application also provides a weakly supervised learning model training device for energy storage batteries, comprising:

[0013] A battery operation data acquisition module is used to acquire battery operation data for a preset time period, wherein the battery operation data is weakly supervised data;

[0014] A latent space determination module, used for compressing the battery operation data based on a multi-layer perceptron of a weakly supervised learning model of an initial energy storage battery to obtain a latent space, wherein the latent space is used to characterize the characteristics of the battery operation data;

[0015] A cluster center determination module, used to cluster the battery operation data according to the latent space to obtain a cluster center, wherein the cluster center includes a label;

[0016] A training data determination module, used to label the K nearest neighbor points of the cluster center and determine the training data according to the obtained label results, where K is a positive integer;

[0017] The training module is used to train the initial energy storage battery weak supervision learning model based on the training data until the objective function meets the preset conditions to obtain the energy storage battery weak supervision learning model.

[0018] In order to solve the above technical problems, an embodiment of the present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned energy storage battery weakly supervised learning model training method when executing the computer program.

[0019] In order to solve the above technical problems, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned energy storage battery weak supervision learning model training method are implemented.

[0020] The energy storage battery weak supervision learning model training method, device, computer equipment and storage medium provided by the embodiment of the present invention obtains the battery operation data of a preset time period, and the battery operation data is weak supervision data; based on the multi-layer perceptron of the initial energy storage battery weak supervision learning model, the battery operation data is compressed to obtain a latent space, and the latent space is used to characterize the characteristics of the battery operation data; according to the latent space, the battery operation data is clustered to obtain a cluster center, and the cluster center includes a label; the K nearest neighbors of the cluster center are labeled, and the training data is determined according to the obtained label result; based on the training data, the initial energy storage battery weak supervision learning model is trained until the objective function meets the preset conditions, and the energy storage battery weak supervision learning model is obtained. Through the multi-layer perceptron and cluster analysis technology, the distribution characteristics and characterization of the battery operation data are learned, and the training data with a small amount of label information is obtained by labeling the battery operation data, which reduces the cost of annotation, thereby reducing the diagnostic cost of energy storage battery fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0022] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0023] Figure 2 It is a flowchart of an embodiment of the energy storage battery weak supervision learning model training method of the present application;

[0024] Figure 3 It is a structural schematic diagram of an embodiment of a weakly supervised learning model training device for energy storage batteries according to the present application;

[0025] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0027] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] See also Figure 1 ,like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0030] Users can use terminal devices 101 , 102 , 103 to interact with server 105 via network 104 to receive or send messages, etc.

[0031] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Eperts Group Audio Layer III), MP4 (Moving Picture Eperts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0032] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0033] It should be noted that the energy storage battery weak supervision learning model training method provided in the embodiment of the present application is executed by a server, and accordingly, the energy storage battery weak supervision learning model training device is arranged in the server.

[0034] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. According to the implementation requirements, there may be any number of terminal devices, networks and servers. The terminal devices 101, 102, 103 in the embodiment of the present application may specifically correspond to application systems in actual production.

[0035] See also Figure 2 , Figure 2 A weakly supervised learning model training method for energy storage batteries provided by an embodiment of the present invention is shown. The method is applied in Figure 1 The server in is used as an example to illustrate the details as follows:

[0036] S201. Obtain battery operation data for a preset time period, where the battery operation data is weakly supervised data.

[0037] In step S201, the battery operation data includes but is not limited to battery capacity, discharge power, temperature, charge state, and operation log.

[0038] The above-mentioned preset time period can be set manually. No specific restrictions are made here. For example, when updating the weak supervision learning model of the energy storage battery at time T, the preset time period is the T-1 time period, that is, from the time when data collection starts to the T-1 time.

[0039] Weakly supervised data refers to data that is unlabeled or has few labels, that is, the data in the battery operation data is unlabeled data or has a small amount of labels.

[0040] The label data in the battery operation data can be used to mark the single cell fault conditions of the battery operation data through known fault log files and expert knowledge.

[0041] It should be understood that after obtaining the battery operation data of the preset time period, the battery operation data needs to be pre-processed, including data cleaning, normalization processing, etc., which can be adjusted according to actual conditions.

[0042] S202. Based on the multi-layer perceptron of the initial energy storage battery weak supervision learning model, the battery operation data is compressed to obtain a latent space, and the latent space is used to characterize the characteristics of the battery operation data.

[0043] In step S202, the initial energy storage battery weakly supervised learning model refers to a weakly supervised learning model constructed based on a multi-layer perceptron and an objective function.

[0044] A multilayer perceptron is a forward-structured artificial neural network that includes an input layer, an output layer, and a hidden layer, and is used to map a set of input vectors to a set of output vectors.

[0045] The latent space refers to the low-dimensional latent subspace obtained by compressing the battery operation data through a multi-layer perceptron. The low-dimensional latent subspace is defined as an abstract multi-dimensional space that encodes a meaningful internal representation of external observation events. Similar samples in the external world are close to each other in the latent space. It contains all the important information required to represent the original data.

[0046] Furthermore, the latent space is obtained according to the following formula (1):

[0047] (1)

[0048] in, It means at time t-1 The potential space output by the k-th layer perceptron, k is the number of layers of the multi-layer perceptron, f is the activation function, is the battery operation data at time t-1, is the weight parameter of the k-th layer perceptron, It refers to the bias of the k-th layer perceptron.

[0049] Preferably, the multi-layer perceptron of the present application is a 4-layer perceptron.

[0050] By compressing the battery operation data, a latent space is obtained to facilitate subsequent learning of the distribution characteristics and representation of the data in the latent space, thereby improving the accuracy of fault type labeling for the battery operation data.

[0051] S203 . Cluster the battery operation data according to the latent space to obtain cluster centers, where the cluster centers include labels.

[0052] In step S203, the above clustering includes but is not limited to a distance-based clustering algorithm, a density-based clustering algorithm, and a distribution-based clustering algorithm.

[0053] The specific clustering algorithm can be adjusted according to the actual situation. The specific clustering process depends on the specific algorithm and is not limited here.

[0054] The labels of the cluster center include normal battery labels and faulty battery labels.

[0055] By learning the distribution characteristics and representation of battery operation data in the latent space, the cluster center is obtained to improve the accuracy of fault type labeling of battery operation data.

[0056] S204, labeling the K nearest neighbor points of the cluster center, and determining the training data according to the obtained labeling results, where K is a positive integer.

[0057] In step S204, specifically, for each cluster center, the K nearest neighbor points of the cluster center are selected, and the data corresponding to the K nearest neighbor points are labeled, that is, the label of the cluster center is used as the label of the data corresponding to the K nearest neighbor points, and the labeled battery operation data is used as training data.

[0058] The methods for obtaining K nearest neighbor points include but are not limited to similarity matching methods, label propagation methods, and semi-supervised learning methods. Among them, the label propagation method refers to a method for predicting the label information of unlabeled nodes from the label information of labeled nodes. Each node label is propagated to adjacent nodes according to the similarity. At each step of node propagation, each node updates its own label according to the labels of adjacent nodes. The greater the similarity with the node, the greater the influence of its adjacent nodes on its labeling.

[0059] The above training data refers to the labeled battery operation data.

[0060] By labeling the K nearest neighbor points of the cluster center, more labeled data are obtained, which reduces the cost of labeling and thus reduces the diagnostic cost of energy storage battery fault diagnosis.

[0061] S205. Based on the training data, the initial energy storage battery weak supervision learning model is trained until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model.

[0062] In step S205 , the objective function is a function used to determine the end of training.

[0063] The preset conditions include but are not limited to the objective function value reaching the preset objective function value, and the objective function value remains unchanged, which can be adjusted according to actual conditions.

[0064] In this embodiment, the distribution characteristics and representations of battery operation data are learned through multi-layer perceptron and cluster analysis technology, and the battery operation data is labeled to obtain training data with a small amount of label information, thereby reducing the labeling cost and thus reducing the diagnostic cost of energy storage battery fault diagnosis.

[0065] In some optional implementations of this embodiment, step S205 includes S2051 to S2052:

[0066] S2051. Reconstruct data in the latent space through a reverse multi-layer perceptron to obtain reconstructed data.

[0067] S2052. Based on the reconstructed data and the training data, the initial energy storage battery weak supervision learning model is trained until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model.

[0068] In step S2051, the reverse multilayer perceptron refers to the process of calculating the gradient from the output layer to the input layer and updating the parameters. The back propagation process is the opposite process of the multilayer perceptron extracting the latent space, that is, the process is opposite.

[0069] The above-mentioned data reconstruction refers to decoding the data in the latent space through the reverse multi-layer perceptron mechanism to generate the reconstructed data corresponding to the latent space.

[0070] The latent space refers to the features obtained after data compression of the battery operation data.

[0071] In step S2052, the objective function is calculated according to the following formula (2):

[0072] (2)

[0073] Where L is the objective function, refers to the weight coefficient of the i-th training data, refers to the i-th training data, refers to the i-th reconstructed data, i refers to the number of training data, the value range of i is (1, N), i is a positive integer, j refers to the number of layers of the multilayer perceptron, the value range of j is (1, M), It refers to the output data of the i-th training data after the j-th layer perceptron. It refers to the output data of the i-th reconstructed data after the j-th layer of reverse perceptron, β refers to the threshold parameter, refers to the low-dimensional data of the i-th cluster center in the latent subspace, refers to the kth cluster center, the value range of k is (1, C), C refers to the total number of cluster centers, and e refers to the exponential function.

[0074] It should be understood that when the label of the i-th training data is a faulty device, The value of is 0. When the label of the i-th training data is a non-faulty device, The value of is 1. Used to calculate the reconstruction error between training data and reconstructed data, Used to calculate the relative error between the data output by the j-th layer perceptron and the j-th layer reverse perceptron after the i-th training data passes through.

[0075] In this embodiment, the initial energy storage battery weak supervision learning model is trained by acquiring training data with a small amount of label information and reconstructed data after the inverse perceptron reconstructs the data, and a weak supervision learning model of the energy storage battery is obtained, thereby reducing the diagnostic cost of energy storage battery fault diagnosis.

[0076] In some optional implementations of this embodiment, after step S205, the method further includes:

[0077] S206: Use the battery operation data of the next preset time period as the newly added data.

[0078] S207. Determine the potential space of new data based on the weakly supervised learning model of the energy storage battery.

[0079] S208. Based on the latent space of the newly added data, update the weakly supervised learning model of the energy storage battery.

[0080] In step S206, when the preset time period is time T-1, the next preset time period is time T. That is, the battery operation data at time T is collected as the newly added data, and the battery operation data is weakly supervised data.

[0081] It should be understood that after obtaining the new data, it is necessary to pre-process the new data, including data cleaning, normalization, etc., which can be adjusted according to the actual situation.

[0082] In step S207, specifically, the newly added data is input into the weakly supervised learning model of the energy storage battery, and based on the multi-layer perceptron of the weakly supervised learning model of the energy storage battery, the newly added data is compressed to determine the potential space of the newly added data.

[0083] In step S208, based on the continuous learning update strategy, the weakly supervised learning model of the energy storage battery is updated according to the latent space of the newly added data.

[0084] The continuous learning update strategy refers to updating the latent space and cluster centers of the training data according to the newly added data, and using the updated latent space and cluster centers as feedback information, optimizing and adjusting the parameters in the weakly supervised learning model of the energy storage battery using the reverse perceptron, and updating the strategy of the weakly supervised learning model of the energy storage battery.

[0085] The above step S208 specifically includes steps S2081 to S2084:

[0086] S2081. Calculate the distance between the new data and the cluster center according to the latent space of the new data, and determine the neighboring clusters of the new data.

[0087] S2082. Calculate the distance between clusters in the nearest neighbor cluster to obtain the nearest neighbor distance, and determine the cluster radius of each cluster.

[0088] S2083. Determine the target cluster based on the nearest neighbor distance and the cluster radius.

[0089] S2084. Based on the target cluster, update the cluster center of the weakly supervised learning model of the energy storage battery.

[0090] In step S2081, the cluster center refers to the cluster center in the weakly supervised learning model of the energy storage battery. That is, the distance between each new sample in the new data and each cluster center of the weakly supervised learning model of the energy storage battery is calculated, and the preset number of clusters closest to the new sample are found as neighboring clusters.

[0091] Preferably, the preset number of clusters is two clusters.

[0092] In step S2082, specifically, the distance between the cluster centers corresponding to two clusters in the nearest neighbor cluster is calculated to obtain the nearest neighbor distance, the distance between the samples in each cluster and the cluster center of the cluster is calculated, and the maximum distance result is used as the cluster radius of the cluster.

[0093] In step S2083, specifically, for each new sample in the new data, the distance between the new sample and all cluster centers in the weakly supervised learning model of the energy storage battery is calculated to obtain a first distance result. When the first distance result satisfies the cluster radius that is less than a certain cluster and the first distance result is within the nearest neighbor distance of the cluster, the new sample is divided into the cluster and the cluster is used as the target cluster.

[0094] In step S2084, specifically, after determining the target cluster of the newly added data, the newly added data is added to the target cluster, the cluster is updated, and the cluster center of the weakly supervised learning model of the energy storage battery is updated.

[0095] It should be understood that after obtaining the weakly supervised learning model of the energy storage battery, the preset abnormal evaluation indicators and the degree of equipment importance can be used to measure the battery failure condition, and the battery failure condition can be pushed to the operation and maintenance personnel according to the degree of battery failure, to assist the operation and maintenance personnel in formulating scientific maintenance decision plans.

[0096] In this embodiment, weakly supervised learning technology and continuous learning methods are used to quickly learn and update the weakly supervised learning model of the energy storage battery from weakly supervised data and new data, so as to achieve accurate fault detection of the energy storage battery group, which can effectively reduce the cost of data annotation and improve the accurate assessment of the health status of the energy storage battery.

[0097] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0098] Figure 3 The schematic diagram of the weak supervision learning model training device for energy storage batteries is shown in one-to-one correspondence with the weak supervision learning model training method for energy storage batteries in the above embodiment. Figure 3 As shown, the energy storage battery weak supervision learning model training device includes a battery operation data acquisition module 31, a potential space determination module 32, a cluster center determination module 33, a training data determination module 34 and a training module 35. The functional modules are described in detail as follows:

[0099] The battery operation data acquisition module 31 is used to acquire the battery operation data of a preset time period, and the battery operation data is weakly supervised data.

[0100] The latent space determination module 32 is used to compress the battery operation data based on the multi-layer perceptron of the initial energy storage battery weak supervision learning model to obtain a latent space, and the latent space is used to characterize the characteristics of the battery operation data.

[0101] The cluster center determination module 33 is used to cluster the battery operation data according to the latent space to obtain the cluster center, and the cluster center includes a label.

[0102] The training data determination module 34 is used to label the K nearest neighbor points of the cluster center and determine the training data according to the obtained label results, where K is a positive integer.

[0103] The training module 35 is used to train the initial energy storage battery weak supervision learning model based on the training data until the objective function meets the preset conditions to obtain the energy storage battery weak supervision learning model.

[0104] In some optional implementations of this embodiment, in the latent space determination module 32, the latent space is obtained according to the following formula (1):

[0105] (1)

[0106] in, It means at time t-1 The potential space output by the k-th layer perceptron, k is the number of layers of the multi-layer perceptron, f is the activation function, is the battery operation data at time t-1, is the weight parameter of the k-th layer perceptron, It refers to the bias of the k-th layer perceptron.

[0107] In some optional implementations of this embodiment, the training module 35 includes:

[0108] The reconstruction unit is used to reconstruct data in the latent space through an inverse multi-layer perceptron to obtain reconstructed data.

[0109] The training unit is used to train the initial energy storage battery weak supervision learning model based on the reconstructed data and the training data until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model.

[0110] Furthermore, the objective function is calculated according to the following formula (2):

[0111] (2)

[0112] Where L is the objective function, refers to the weight coefficient of the i-th training data, refers to the i-th training data, refers to the i-th reconstructed data, i refers to the number of training data, the value range of i is (1, N), i is a positive integer, j refers to the number of layers of the multilayer perceptron, the value range of j is (1, M), It refers to the output data of the i-th training data after the j-th layer perceptron. It refers to the output data of the i-th reconstructed data after the j-th layer of reverse perceptron, β refers to the threshold parameter, refers to the low-dimensional data of the i-th cluster center in the latent subspace, refers to the kth cluster center, the value range of k is (1, C), C refers to the total number of cluster centers, and e refers to the exponential function.

[0113] In some optional implementations of this embodiment, after the training module 35, the energy storage battery weak supervision learning model training device further includes:

[0114] A new data acquisition module is added, which is used to use the battery operation data of the next preset time period as the new data.

[0115] A new data latent space determination module is added, which is used to determine the latent space of new data based on the weakly supervised learning model of energy storage batteries.

[0116] The update module is used to update the weakly supervised learning model of the energy storage battery based on the latent space of the newly added data.

[0117] Furthermore, the update module includes:

[0118] The neighbor cluster determination unit is used to calculate the distance between the new data and the cluster center according to the potential space of the new data, and determine the neighbor cluster of the new data.

[0119] The distance calculation unit is used to calculate the distance between clusters in the nearest neighbor cluster, obtain the nearest neighbor distance, and determine the cluster radius of each cluster.

[0120] The target cluster determination unit is used to determine the target cluster based on the nearest neighbor distance and the cluster radius.

[0121] The updating unit is used to update the cluster center of the weakly supervised learning model of the energy storage battery based on the target cluster.

[0122] In some optional implementations of this embodiment, the training module 35 includes:

[0123] The reconstruction unit is used to reconstruct data in the latent space through an inverse multi-layer perceptron to obtain reconstructed data.

[0124] The training unit is used to train the initial energy storage battery weak supervision learning model based on the reconstructed data and the training data until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model.

[0125] For the specific limitations of the weakly supervised learning model training device for energy storage batteries, please refer to the limitations of the weakly supervised learning model training method for energy storage batteries mentioned above, which will not be repeated here. Each module in the above-mentioned weakly supervised learning model training device for energy storage batteries can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0126] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0127] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components connected to the memory 41, the processor 42, and the network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0128] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0129] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or D interface display memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as the program code for controlling electronic files, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0130] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the program code stored in the memory 41 or process data, such as running the program code for controlling the electronic file.

[0131] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0132] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores an interface display program, and the interface display program can be executed by at least one processor so that the at least one processor performs the steps of the weakly supervised learning model training method for energy storage batteries as described above.

[0133] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0134] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A weakly supervised learning model training method for energy storage batteries, characterized in that: The energy storage battery weak supervision learning model training method comprises: Acquire battery operation data for a preset time period, wherein the battery operation data is weakly supervised data; Based on a multi-layer perceptron of a weakly supervised learning model of an initial energy storage battery, data compression is performed on the battery operation data to obtain a latent space, wherein the latent space is used to characterize the characteristics of the battery operation data; Clustering the battery operation data according to the latent space to obtain cluster centers, wherein the cluster centers include labels; Labeling the K nearest neighbor points of the cluster center, and determining the training data according to the obtained labeling results, where K is a positive integer; Based on the training data, the initial energy storage battery weak supervision learning model is trained until the objective function meets the preset conditions to obtain the energy storage battery weak supervision learning model; The step of training the initial energy storage battery weak supervision learning model based on the training data until the objective function meets the preset conditions to obtain the energy storage battery weak supervision learning model includes: Through the reverse multi-layer perceptron, the data in the latent space is reconstructed to obtain the reconstructed data; Based on the reconstructed data and the training data, the initial energy storage battery weak supervision learning model is trained until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model; The step of training the initial energy storage battery weak supervision learning model based on the reconstructed data and the training data until the objective function meets the preset conditions to obtain the energy storage battery weak supervision learning model includes: The objective function is calculated according to the following formula: in, is the objective function, w i refers to the weight coefficient of the i-th sample data, x i refers to the i-th sample data, refers to the reconstructed data of the i-th sample data, i refers to the number of sample data, and the value range of i is [1, N]. j refers to the number of layers of the multilayer perceptron, and the value range of j is [1, M]. It refers to the output data of the i-th sample data after the j-th layer perceptron. refers to the output data of the i-th reconstructed data after the j-th layer of reverse perceptron, β refers to the threshold parameter, and h i refers to the low-dimensional data of the i-th cluster center in the latent subspace, c k refers to the kth cluster center, the value range of k is [1, C], C refers to the total number of cluster centers, and e refers to the exponential function; The multi-layer perceptron based on the weakly supervised learning model of the initial energy storage battery compresses the battery operation data to obtain the latent space, comprising: The latent space is obtained according to the following formula: in, It means R at time t-1 t-1 The potential space output by the k-th layer perceptron, k is the number of layers of the multi-layer perceptron, f is the activation function, R t-1 is the battery operation data at time t-1, is the weight parameter of the k-th layer perceptron, It refers to the bias of the k-th layer perceptron.

2. The energy storage battery weak supervision learning model training method according to claim 1, characterized in that: After the initial energy storage battery weak supervision learning model is trained based on the training data until the objective function satisfies a preset condition and the energy storage battery weak supervision learning model is obtained, the method further includes: The battery operation data of the next preset time period is used as the newly added data; Determine the potential space of the newly added data based on the weakly supervised learning model of the energy storage battery; Based on the latent space of the newly added data, the weakly supervised learning model of the energy storage battery is updated.

3. The weakly supervised learning model training method for energy storage batteries according to claim 2, characterized in that: The step of updating the weakly supervised learning model of the energy storage battery based on the latent space of the newly added data comprises: According to the latent space of the newly added data, the distance between the newly added data and the cluster center is calculated to determine the neighboring cluster of the newly added data; Calculating the distances between clusters in the nearest neighbor cluster to obtain nearest neighbor distances, and determining the cluster radius of each cluster; Determine a target cluster based on the neighbor distance and the cluster radius; Based on the target cluster, the cluster center of the weakly supervised learning model of the energy storage battery is updated.

4. A weakly supervised learning model training device for energy storage batteries, characterized in that: The energy storage battery weak supervision learning model training device comprises: A battery operation data acquisition module is used to acquire battery operation data for a preset time period, wherein the battery operation data is weakly supervised data; A latent space determination module, used for compressing the battery operation data based on a multi-layer perceptron of a weakly supervised learning model of an initial energy storage battery to obtain a latent space, wherein the latent space is used to characterize the characteristics of the battery operation data; A cluster center determination module, used to cluster the battery operation data according to the latent space to obtain a cluster center, wherein the cluster center includes a label; A training data determination module, used to label the K nearest neighbor points of the cluster center and determine the training data according to the obtained label results, where K is a positive integer; A training module, used to train the initial energy storage battery weak supervision learning model based on the training data until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model; Wherein, the training module includes: A reconstruction unit, used to reconstruct data in the latent space through an inverse multi-layer perceptron to obtain reconstructed data; A training unit, used to train the initial energy storage battery weak supervision learning model based on the reconstructed data and the training data until the objective function meets the preset conditions, thereby obtaining the energy storage battery weak supervision learning model; In the training unit, the objective function is calculated according to the following formula: in, is the objective function, w i refers to the weight coefficient of the i-th sample data, x i refers to the i-th sample data, refers to the reconstructed data of the i-th sample data, i refers to the number of sample data, and the value range of i is [1, N]. j refers to the number of layers of the multilayer perceptron, and the value range of j is [1, M]. It refers to the output data of the i-th sample data after the j-th layer perceptron. refers to the output data of the i-th reconstructed data after the j-th layer of reverse perceptron, β refers to the threshold parameter, and h i refers to the low-dimensional data of the i-th cluster center in the latent subspace, c k refers to the kth cluster center, the value range of k is [1, C], C refers to the total number of cluster centers, and e refers to the exponential function; The multi-layer perceptron based on the weakly supervised learning model of the initial energy storage battery compresses the battery operation data to obtain the latent space, comprising: The latent space is obtained according to the following formula: in, It means R at time t-1 t-1 The potential space output by the k-th layer perceptron, k is the number of layers of the multi-layer perceptron, f is the activation function, R t-1 is the battery operation data at time t-1, is the weight parameter of the k-th layer perceptron, It refers to the bias of the k-th layer perceptron.

5. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the energy storage battery weakly supervised learning model training method as described in any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the energy storage battery weakly supervised learning model training method as described in any one of claims 1 to 3 is implemented.

Citation Information

Patent Citations

  • Power distribution network operation mode identification method based on measurement big data analysis

    CN116247668A

  • Battery state prediction model training method, battery state prediction method and device

    CN117808116A

  • Method and apparatus for generating target re-recognition model and re-recognizing target

    US20210312208A1