Data processing method and device, electronic equipment and computer readable storage medium

Abstract

The embodiment of the invention discloses a data processing method and device, electronic equipment and a computer readable storage medium, and the method comprises the steps: inputting each battery data group in battery flow data into a pre-trained classification model for classification, and determining the class label of the battery data group, generating an incremental data set according to the battery data set corresponding to the new label, inputting the battery data set in the incremental data set into a pre-trained clustering model for clustering, and determining an incremental clustering result; and performing incremental learning training on the classification model according to an incremental clustering result, so as to determine a category label based on the trained classification model. Therefore, the category labels of the battery data sets under the newly-added categories can be determined through the method, the battery classification information can be expanded based on the newly-added battery data sets, and the accuracy and generalization of label determination are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data processing method, device, electronic device and computer-readable storage medium. Background Art

[0002] When battery big data aggregation analysis and model building are in progress, there is a strong reliance on the unique identification information of "battery system type". In the process of labeling battery system types, static information that remains unchanged throughout the battery life cycle is often used for clustering, so that labels are performed according to the categories in the clustering results.

[0003] However, in the existing clustering methods, since the static information used for clustering needs to strongly rely on the quality of static information transmitted to the cloud platform by new energy vehicles, two-wheeled vehicles, charging piles or energy storage, when used in aggregation, there will be results such as wrong classification, over-classification, and under-classification due to problems with individual information. In addition, there are difficulties in classifying "new types of low-quality battery information" and "new types of battery information" into designated categories and new classifications. Therefore, how to build a set of highly generalized, universal and applicable methods for new classifications is of great value and significance to solving the above problems. Summary of the invention

[0004] In view of this, an object of embodiments of the present invention is to provide a data processing method, device, electronic device and computer-readable storage medium to improve the accuracy and generalization of label determination.

[0005] In a first aspect, an embodiment of the present invention is to provide a data processing method, the method comprising: Acquiring battery flow data, wherein the battery flow data includes a plurality of battery data groups; Input each of the battery data groups into a pre-trained classification model for classification, and determine a category label of the battery data group, wherein the category label is an existing label or a new label; Generate an incremental data set according to the battery data group corresponding to the new label; Inputting the incremental data set into a pre-trained clustering model for clustering, and determining an incremental clustering result, wherein the incremental clustering result includes at least one new cluster, and each of the new clusters has a corresponding cluster center point and a newly added category label; Incremental learning training is performed on the classification model according to the incremental clustering result to determine the category label based on the trained classification model.

[0006] In a second aspect, an embodiment of the present invention is intended to provide a data processing device, the device comprising: An acquisition unit, configured to acquire battery flow data, wherein the battery flow data includes a plurality of battery data groups; A classification unit, configured to input each of the battery data groups into a pre-trained classification model for classification, determine a category label of the battery data group, wherein the category label is an existing label or a new label; and generate an incremental data set according to the battery data group corresponding to the new label; A clustering unit, configured to input the incremental data set into a pre-trained clustering model for clustering, and determine an incremental clustering result, wherein the incremental clustering result includes at least one new cluster, each of the new clusters having a corresponding cluster center point and a second category label; A training unit is used to perform incremental learning training on the classification model according to the incremental clustering result, so as to determine the category label based on the trained classification model.

[0007] In a third aspect, an embodiment of the present invention aims to provide a computer program product, wherein the computer program product comprises a computer program / instructions, and when the computer program / instructions are executed by a processor, the method as described in any one of the above items is implemented.

[0008] In a fourth aspect, an embodiment of the present invention aims to provide an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement a method as described in any one of the above items.

[0009] In a fifth aspect, an embodiment of the present invention aims to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method as described in any one of the above items is implemented.

[0010] The technical solution of the embodiment of the present invention determines the category label of the battery data group by inputting each battery data group in the battery flow data into a pre-trained classification model for classification, generates an incremental data set according to the battery data group corresponding to the new label, and inputs the battery data group in the incremental data set into a pre-trained clustering model for clustering to determine the incremental clustering result; performs incremental learning training on the classification model according to the incremental clustering result to determine the category label based on the trained classification model, which can not only determine the category label of the battery data group under the newly added category and improve the accuracy of label determination, but also enable the battery classification information to be expanded based on the newly added battery data group, thereby improving the generalization of label determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which: Figure 1 is a flow chart of a data processing method according to an embodiment of the present invention; Figure 2is a flow chart of determining a classification model according to an embodiment of the present invention; Figure 3 is a structural diagram of a classification model according to an embodiment of the present invention; Figure 4 is a flow chart of determining a category label according to an embodiment of the present invention; Figure 5 is a flow chart of incremental learning according to an embodiment of the present invention; Figure 6 is a schematic diagram of adjusting the classification model structure according to an embodiment of the present invention; Figure 7 is a flow chart of determining a category label according to an embodiment of the present invention; Figure 8 is a flow chart of a training clustering model according to an embodiment of the present invention; Fig. 9 is a schematic diagram of a clustering result information table according to an embodiment of the present invention; Fig.10 is a schematic diagram of a cluster center point information table according to an embodiment of the present invention; Fig.11 is a schematic diagram of an incremental clustering information table according to an embodiment of the present invention; Fig.12 is a schematic diagram of determining an input data source for incremental learning according to an embodiment of the present invention; Fig.13 is a schematic diagram of a gridded parameter combination according to an embodiment of the present invention; Fig.14 is a schematic diagram of a data processing device according to an embodiment of the present invention; Fig.15 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The present application is described below based on embodiments, but the present application is not limited to these embodiments. In the detailed description of the present application below, some specific details are described in detail. It is possible for those skilled in the art to fully understand the present application without the description of these details. In order to avoid confusing the essence of the present application, known methods, processes, flows, components and circuits are not described in detail.

[0013] In addition, persons of ordinary skill in the art will appreciate that the drawings provided herein are for illustration purposes and are not necessarily drawn to scale.

[0014] Unless the context clearly requires otherwise, the words "include", "comprising" and similar words throughout the application should be interpreted as including rather than exclusive or exhaustive; that is, the meaning is "including but not limited to".

[0015] In the description of this application, it should be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0016] The solutions described in this specification and the embodiments, if they involve information acquisition, will collect data under the premise of legality and compliance, ensure the legitimacy of the data source, and take appropriate technical and management measures to ensure data security. If it involves the processing of personal information, it will be processed on the premise of having a legal basis (such as obtaining the consent of the subject of personal information, or it is necessary to perform a contract, etc.), and will only be processed within the scope of regulations or agreements. The user's refusal to process personal information other than the necessary information for basic functions will not affect the user's use of basic functions.

[0017] In the prior art, the category labels of battery data are often determined by clustering battery static information, and the corresponding battery data are labeled based on the determined category labels. However, when the quality of the battery static information is not high or the battery static information is a newly added type of battery information, accurate classification cannot be achieved, and the generalization of the determined category labels is poor. In view of this, an embodiment of the present invention aims to provide a data processing method to improve the accuracy and generalization of battery data label determination.

[0018] Figure 1 is a flow chart of a data processing method according to an embodiment of the present invention. Figure 1 As shown, the data processing method in this embodiment includes the following steps.

[0019] In step S110 , battery flow data is acquired, where the battery flow data includes a plurality of battery data groups.

[0020] In this embodiment, the battery data group includes multiple battery characteristic data, including battery material type (brm_battery_type), battery rated capacity (brm_battery_power), battery rated total voltage (brm_battery_voltage), battery maximum operating temperature (bcp_max_temp), battery cell upper limit voltage (bcp_max_voltage), system maximum total voltage (bcp_total_voltage), battery nominal total energy (bcp_max_power). However, it should be understood that the type and quantity of battery characteristic data in the battery data group in this embodiment may include but are not limited to the above-mentioned characteristic data, which can be set according to the specific usage scenario, and is not limited here.

[0021] Optionally, in order to improve the accuracy of determining the battery data label, the battery flow data in this embodiment uses preprocessed data. When acquiring the battery flow data, in this embodiment, the original flow data can be first acquired, and the original flow data includes multiple battery static source data; then each battery static source data is preprocessed to obtain a corresponding battery data group, so as to generate the battery flow data according to each preprocessed battery data group. That is, the battery flow data corresponding to the original flow data is acquired by preprocessing each battery static source data in the original flow data, and each battery data group in the battery flow data has a corresponding battery static source data.

[0022] Furthermore, to ensure the orderliness of data processing, the original flow data obtained in this embodiment includes multiple battery static source data within a predetermined time period. The predetermined time period can be one day, one week, etc. The type and quantity of battery characteristic data in the battery static source data are the same as those in the battery flow data.

[0023] At the same time, when preprocessing the battery static source data, optionally, the preprocessing operation in this embodiment may include at least one of other types of data preprocessing operations such as data elimination, data filtering, data normalization, etc., so that higher quality battery flow data can be used for classification and labeling later.

[0024] Further optionally, when data is eliminated, in this embodiment, the characteristic data that exceeds the upper and lower limits can be marked according to the upper and lower limits set by the "end-to-cloud" transmission protocol or industry standards, and the entire data information group containing the characteristic data can be eliminated, that is, the battery static source data to which the characteristic data exceeds the upper and lower limits is eliminated.

[0025] When performing data filtering, in this embodiment, the characteristic data information of a specific type in the battery static source data can be subjected to consistency check filtering and business rule verification. For example, for the characteristic data of the battery rated capacity, in this embodiment, the available power difference diff_power of the battery is first determined based on the following formula: Where abs() is a function for obtaining absolute values, brm_battery_power is the rated capacity of the battery, bcp_max_power is the nominal total energy of the battery, and brm_battery_voltage is the rated total voltage of the battery.

[0026] Then, the available power difference is Compared with the preset power threshold t, When the battery static source data to which the characteristic data of the battery rated capacity belongs is filtered out.

[0027] For another example, for the characteristic data of battery material type, in this embodiment, it can be determined whether the battery material type in the static source data is the target type. If the battery material type is the target type, the static source data corresponding to the battery material type is retained; otherwise, if the battery material type is not the target type, the static source data corresponding to the battery material type is filtered out. The target type can be a material type that makes the nominal total energy of the battery fall within a specific upper and lower limit range.

[0028] When performing data normalization, the present embodiment may use Min-Max Normalization for processing, which can eliminate the dimensional influence between different feature data with similar classification attributes but different numerical magnitudes in the static source data, thereby improving data stability and accelerating the convergence speed of data processing results.

[0029] In step S120, each battery data group is input into a pre-trained classification model for classification, and a category label of the battery data group is determined. The category label is an existing label or a new label.

[0030] In this embodiment, the classification model is determined based on a pre-trained clustering model, and the pre-trained clustering model has fixed model parameters. At the same time, since the clustering model with fixed model parameters can only cluster battery data groups under different existing categories, and these existing categories are all obtained by clustering the clustering model. Therefore, determining the category label of the battery data group based on the classification model determined by the clustering model can enable the classification model to maintain the classification and recognition ability of the battery data group of the existing category, so as to quickly determine the category label of the battery data group under the existing category as the existing label under the corresponding existing category. At the same time, for the battery data group under the non-existing category, the corresponding category label can be quickly determined as a new label, so as to characterize the battery data group as a battery data group under the newly added category through the new label.

[0031] Optionally, the clustering model in this embodiment is determined based on an unsupervised training method. When determining the clustering model, in this embodiment, a historical data set is first obtained, and the historical data set includes multiple battery data groups within a historical time period; then the battery data groups in the historical data set are clustered based on a preset clustering algorithm to determine the historical clustering results and the clustering model, and the historical clustering results include at least one clustering cluster, and each clustering cluster has a corresponding cluster center point and an existing label.

[0032] Further optionally, the clustering model in this embodiment can adopt the DBSCAN model, wherein the DBSCAN model (Density-Based Spatial Clustering of Applications with Noise) is a density-based spatial clustering algorithm that can describe the compactness of a sample set based on a set of neighborhoods, divide areas with sufficient density into clusters, and find clusters of arbitrary shapes in a spatial database with noise, defining a cluster as the largest set of density-connected points. The DBSCAN model uses parameters (ε, MinPts) to describe the compactness of sample distribution in the neighborhood. Among them, ε describes the neighborhood distance threshold of a certain sample, and MinPts describes the threshold of the number of samples in the neighborhood with a distance of ε of a certain sample.

[0033] Furthermore, after the clustering model is determined through training, the model parameters in the trained clustering model are recorded in this embodiment, so that clustering is performed based on the recorded fixed model parameters in the future. At the same time, the classification model is determined based on the clustering results (historical clustering results) corresponding to the trained clustering model in this embodiment, so as to classify the battery data groups in the battery flow data based on the classification model.

[0034] Furthermore, in this embodiment, when determining the classification model, in this embodiment, Figure 2 The method shown determines the classification model, which specifically includes the following steps.

[0035] In step S210, a classification model is built according to the cluster center points of each cluster in the historical clustering results.

[0036] Figure 3 Schematic diagram of the structure of the classification model of the embodiment of the present invention. Figure 3 As shown, the classification model in this embodiment includes an input layer, a hidden layer and an output layer. The input layer is responsible for receiving the input battery data set and processing the data in the battery data set into a format that can be processed by the model. The hidden layer is responsible for extracting data features in the battery data set. The output layer is responsible for converting the data features extracted by the hidden layer into the final classification result or predicted value.

[0037] Furthermore, the input layer of the classification model of this embodiment is set based on the number of battery feature data in the battery data group, and the hidden layer and the output layer are set based on the number of cluster clusters identified by the clustering model, that is, the number of existing categories. For example, assuming that the battery feature data in the battery data group includes the aforementioned 7 types of battery feature data, the number of nodes in the input layer is set to 7; assuming that the number of existing categories that the clustering model can identify is m, and each existing category has a corresponding cluster center, that is, the clustering model corresponds to m cluster clusters and m cluster centers, then the number of nodes in the hidden layer and the output layer is set according to the m clusters and m cluster centers. Therefore, in this embodiment, a classification model is built by using the clustering performance of the clustering model, so that when determining the category label of the battery data group based on the classification model, the characteristic attributes under the same category can be effectively utilized to improve the accuracy of the category label determination.

[0038] Alternatively, if Figure 3 As shown, in order to quickly and intuitively determine the category label of the battery data group, the classification model in this embodiment is also provided with a probability prediction layer (i.e., a softmax layer) after the output layer. The softmax layer is used to realize multi-classification probability prediction, and the prediction probability of the battery data group under different existing categories is determined. The prediction probability is used to characterize the confidence that the battery data group belongs to the corresponding existing category. At the same time, the category label of the battery data group is determined by comparing the confidence represented by the prediction probability with the preset confidence threshold.

[0039] Further optionally, the classification model in this embodiment can adopt a radial basis neural network RBFNN model. The RBFNN model (Radial Basis Function Neural Network) is a neural network model widely used in classification, regression and clustering problems. The connection weights between neurons are calculated by radial basis functions. The radial basis function is a distance-based function that can measure the similarity between input data and neurons. In RBFNN, the input layer passes the feature vector to the neurons in the hidden layer. The hidden layer neurons use radial basis functions to calculate the similarity between the input data and the neurons, and pass the results to the neurons in the output layer. The output layer uses a set of basis functions to calculate the output value, usually a Gaussian function or a polynomial function. The parameters of the basis function are determined by training.

[0040] Therefore, in this embodiment, by combining the DBSCAN model and the RBFNN model to generate category labels for the battery data group, the density clustering advantage of DBSCAN and the nonlinear modeling ability of RBFNN can be innovatively combined to construct an efficient and flexible hybrid model. In this way, when determining the category label, the clustering center of DBSCAN can be directly mapped to the hidden layer basis function center of RBFNN, which not only retains the recognition ability of DBSCAN for complex data structures, but also introduces the nonlinear modeling advantage of RBFNN, so that the model can more accurately capture the local characteristics and complex distribution of the data, thereby improving the accuracy and stability of classification; and, in the DBSCAN-RBFNN joint incremental learning, the hidden layer basis function of RBFNN can capture the deep correlation between the data, thereby realizing the effective integration of new categories and the continuous optimization of old categories. This adaptability and scalability enables the model to maintain efficient and stable classification performance for a long time; finally, in terms of incremental learning, the architecture can reduce unnecessary computational overhead.

[0041] In step S220, a training data set is determined according to the battery data sets corresponding to the existing labels.

[0042] In this embodiment, after determining the model structure of the classification model, the training data set is determined by using the battery data sets corresponding to the existing labels, so that the classification model trained based on the training data set can quickly and accurately identify the battery data sets under the existing categories corresponding to the existing labels.

[0043] In step S230, the classification model is iteratively trained according to the training data set until the classification model obtained by training meets the preset conditions.

[0044] In this embodiment, the process of iteratively training the classification model according to the training data set is the process of adjusting the model parameters inside the classification model. In order to improve the performance of the trained classification model, a preset condition is pre-set in this embodiment. The preset condition can be that the model output accuracy reaches a preset accuracy threshold, the number of iterative training reaches a preset number, etc., which can be set according to the actual usage scenario.

[0045] After each training, the classification model obtained in the current training round will be evaluated based on the preset conditions until the classification model obtained through training meets the preset conditions, and the classification model that meets the preset conditions will be used for subsequent classification to determine the category label of the battery data group.

[0046] Figure 4 FIG. 1 is a flow chart of determining a category label according to an embodiment of the present invention. Figure 4 As shown, in this embodiment, the category label of the battery data group is determined by the following method.

[0047] In step S410, the battery data set is input into a pre-trained classification model for classification, and the confidence of the battery data set under each existing category is determined.

[0048] Optionally, in order to improve data processing efficiency, this embodiment inputs each battery data group in the battery flow data into a pre-trained classification model for classification, and determines the confidence of each battery data under different existing categories. For example, assuming that there are three existing categories, namely type-1, type-2 and type-3, then after inputting the feature data corresponding to a battery data group into the classification model for classification, the confidence of the battery data group belonging to type-1, type-2 and type-3 can be obtained.

[0049] In step S420, the relationship between the confidence level and the confidence threshold value of each existing category is determined.

[0050] In this embodiment, after determining the confidence that the battery data group belongs to each existing category, the confidence under each existing category is compared with a preset confidence threshold to determine the relationship between the confidence of the battery data group under each existing category and the corresponding confidence threshold.

[0051] Optionally, for different existing categories, the same confidence threshold may be used in this embodiment, or different confidence thresholds may be used, which may be set according to actual usage scenarios. Specifically, to further improve processing efficiency, the confidence thresholds corresponding to different existing categories in this embodiment use the same value.

[0052] In step S430 , in response to the confidence reaching a confidence threshold, an existing category corresponding to the confidence reaching the confidence threshold is determined as a category label of the battery data group.

[0053] In this embodiment, when the confidence of the battery data set under a certain existing category reaches the confidence threshold, that is, the confidence is greater than or equal to the confidence threshold, the existing category is determined as the category label of the battery data set. For example, assuming that the confidence of battery data set 1 under type-1 is greater than the confidence threshold, type-1 is used as the category label of battery data set 1, that is, the category label of battery data set 1 is the existing label type-1.

[0054] Optionally, when the number of existing categories whose confidence reaches the confidence threshold is multiple (i.e., greater than 1), this embodiment first determines the existing categories whose confidence reaches the confidence threshold as candidate categories, and then selects the existing category with the largest confidence from all candidate categories as the category label of the battery data group.

[0055] In step S440 , in response to each confidence level being less than a confidence level threshold, it is determined that the category label of the battery data group is a new label.

[0056] In this embodiment, when the confidence of the battery data group in each existing category is less than the confidence threshold, it indicates that the battery data group does not belong to all the current existing categories, that is, the battery data group belongs to the newly added category data. At this time, the category label of the battery data group is determined to be a new label.

[0057] In step S130, an incremental data set is generated according to the battery data set corresponding to the new tag.

[0058] In this embodiment, after determining the category label of each battery data group in the battery flow data, the battery data groups in the battery flow data can be divided into two categories according to the category label of each battery data group. One category is the battery data group that can be classified into an existing category, in which each battery data group has an existing label corresponding to the existing category, and the existing labels of different battery data groups are different or the same. By aggregating the battery data groups under each existing category, a classified data set and the classified data information corresponding to each existing category are obtained.

[0059] The other category is the battery data set that cannot be classified into the existing categories. The battery data set in this category has not yet determined the specific category it belongs to. It is new data and is identified by a unified category label - new label. By aggregating all battery data sets with the category label of new label, the corresponding incremental data set can be obtained.

[0060] For example, assuming that there are three existing categories, namely type-1, type-2 and type-3, and the battery flow data includes battery data groups 1-15, then after inputting the feature data corresponding to a battery data group into the classification model for classification, the confidence that the battery data group belongs to type-1, type-2 and type-3 can be obtained. After classifying battery data groups 1-15 using the pre-trained classification model, it is determined that the confidence of battery data groups 1-3 under the existing category type-1 is greater than the confidence threshold, the confidence of battery data groups 4-6 under the existing category type-2 is greater than the confidence threshold, and the confidence of battery data groups 7-9 under the existing category type-3 is greater than the confidence threshold, then battery data groups 1-9 are all battery data groups that can be classified into existing categories, and the category labels of battery data groups 1-3 are all existing labels type-1, the category labels of battery data groups 4-6 are all type-2, and the category labels of battery data groups 7-9 are all existing labels type-3.

[0061] At the same time, since the confidence levels of battery data groups 10-15 under the existing categories type-1, type-2 and type-3 are all less than the confidence threshold, it can be determined that battery data groups 10-15 are all battery data under the newly added categories, and the category labels of battery data groups 10-15 are all new labels. By aggregating battery data groups 10-15, an incremental data set is obtained.

[0062] In step S140, the incremental data set is input into the pre-trained clustering model for clustering, and an incremental clustering result is determined. The incremental clustering result includes at least one new cluster, and each new cluster has a corresponding cluster center point and a newly added category label.

[0063] In this embodiment, the incremental clustering results corresponding to the incremental data set are determined by inputting each battery data group in the incremental data set into a pre-trained clustering model for clustering, thereby determining a more specific category label for the battery data group whose category label is a new label, that is, a new category label.

[0064] For example, when the incremental data set includes the aforementioned battery data groups 10-15, the battery feature data in the battery data groups 10-15 will be input into the pre-trained clustering model for clustering, and the corresponding clustering results, that is, the incremental clustering results, can be determined. Assuming that two clusters are formed after clustering, one cluster includes the battery data groups 10-12, and the other cluster includes the battery data groups 13-15, then the two clusters are defined as new clusters, one new cluster includes the battery data groups 10-12, and the other new cluster includes the battery data groups 13-15. At the same time, in order to facilitate the overall label classification, the two new clusters will be numbered using the existing category numbers, and the newly added category corresponding to the new cluster where the battery data groups 10-12 are located will be defined as type-4, and the newly added category corresponding to the new cluster where the battery data groups 13-15 are located will be defined as type-5. Then, it can be determined that the category labels of the battery data groups 10-12 are the newly added category labels type-4, and the category labels of the battery data groups 13-15 are the newly added category labels type-5.

[0065] It should be noted that the clustering model used when clustering the incremental data set in this embodiment is the same clustering model as the clustering model used when clustering the battery flow data mentioned above, and the parameters in the clustering model are fixed parameters after freezing. This enables the clustering model to cluster the battery data groups under the existing categories while also clustering the battery data groups under the newly added categories, thereby determining the categories and category labels of the battery data groups under the newly added categories.

[0066] In step S150, incremental learning training is performed on the classification model according to the incremental clustering result to determine the category label based on the trained classification model.

[0067] In this embodiment, after determining the incremental clustering results corresponding to the incremental data set, by performing incremental learning training on the classification model according to the incremental clustering results, the trained classification model can identify and classify the battery data group under the newly added category, thereby determining the category label of the battery data group under the newly added category and improving the generalization of the category label determination.

[0068] Optionally, the category label in this embodiment is used to characterize the battery type, so that refined battery management and battery performance optimization can be achieved based on battery data of different battery types.

[0069] Furthermore, in this embodiment, when the classification model is incrementally trained, the structure and model parameters of the classification model are adjusted at the same time, so that the trained classification model can accurately identify the category of the battery data group under the newly added category, and improve the generalization and accuracy of the battery data group label determination. At the same time, such an adjustment method can enable the classification model to have the ability to incrementally fine-tune the model parameters and update the structure at the same time, and on the basis of ensuring the stability, continuous reliability of the historical classification label to the greatest extent, the battery classification information can be continuously expanded based on the newly added information, and has the ability to be applied throughout the business cycle.

[0070] Figure 5 is a flowchart of incremental learning in an embodiment of the present invention. Figure 5 As shown, in this embodiment, incremental learning and training of the classification model is achieved through the following method.

[0071] In step S510, the structure of the classification model is adjusted according to the cluster center points of each new cluster.

[0072] In this embodiment, since the hidden layer and output layer of the classification model are determined based on the cluster clusters identified by the clustering model, after the clustering model identifies new cluster clusters based on the incremental data set, that is, clusters the new clusters, the number of nodes in the hidden layer and output layer of the classification model will be adjusted according to the cluster center of each new cluster, so that the category labels that can be identified by the adjusted classification model and the categories that can be clustered by the clustering model are kept updated synchronously, so that the classification model can classify and determine the category labels for the battery data groups under the newly added categories during subsequent use, thereby improving the generalization and accuracy of the category label determination.

[0073] Figure 6 Schematic diagram of adjusting the classification model structure according to an embodiment of the present invention. Figure 6As shown, after the clustering model identifies the new clusters, in this embodiment, based on the original classification model structure, nodes corresponding to each new cluster are added in the hidden layer and output layer of the classification model according to the number of new clusters, thereby adjusting the structure of the original classification model. Compared with the classification model before adjustment, the adjusted classification model can extract and analyze battery data features under more categories, and can not only determine the category labels of battery data groups under existing categories, but also determine the category labels of battery data groups under newly added categories, and has higher generalization for the determination of category labels.

[0074] In step S520, incremental learning training is performed on the adjusted classification model according to the new category data information to determine the category label based on the trained classification model, and the new category data information includes the battery data group corresponding to each newly added category label.

[0075] Optionally, in this embodiment, the classification model with the adjusted structure can be iteratively trained only according to the new category information. For example, continuing with the above example, it can be determined that the new category information includes newly added categories type-4 and type-5, as well as battery data groups 10-12 under the newly added category type-4 and battery data groups 13-15 under the newly added category type-5. At this time, the classification model with the adjusted structure is iteratively trained using the new category information, so that the trained classification model can quickly identify the battery data group under the newly added category, and then determine the category label of the battery data group under the newly added category, thereby improving the accuracy and generalization of the overall label determination.

[0076] Alternatively, in order to take into account the classification model's recognition of battery data groups under newly added categories and battery data groups under existing categories at the same time, the adjusted classification model can also be iteratively trained according to the classification data information and the new category data information in this embodiment, and the classification data information is determined according to the battery data groups corresponding to each existing label. For example, continuing with the above example, it can be determined that the classification data information includes battery data groups 1-3 under existing labels type-1, type-2, and type-3, as well as battery data groups 1-3 under existing labels type-1, battery data groups 4-6 under existing labels type-2, and battery data groups 7-9 under existing labels type-3. New category information includes newly added categories type-4 and type-5, as well as battery data groups 10-12 under newly added categories type-4 and battery data groups 13-15 under newly added categories type-5. Afterwards, the adjusted classification model is iteratively trained through the classification data information and the new category data information, so that the trained classification model can simultaneously achieve effective integration of newly added categories and continuous optimization of existing categories, thereby improving the accuracy and generalization of category label determination.

[0077] Furthermore, in this embodiment, after completing the incremental learning and training of the classification model, the classification model obtained after the incremental learning and training will be used to process the subsequent input battery flow data to classify the newly added battery data group and determine its category label. In addition, when processing the newly added battery flow data, the above processing process will be performed again in this embodiment. While determining the accurate category label for the newly added battery data group, the classification model can be continuously optimized, the battery classification information can be continuously and dynamically expanded, and the classification model can be applied to the newly added classification, with higher generalization and versatility.

[0078] The technical solution of this embodiment uses a clustering model and a classification model in combination, and uses the clustering structure of the clustering model to build a classification model. After acquiring the newly added battery flow data, each battery data group in the battery flow data is input into a pre-trained classification model for classification to determine the category label of the battery data group, and an incremental data set is generated according to the battery data group corresponding to the new label, and the battery data group in the incremental data set is input into the pre-trained clustering model for clustering to determine the incremental clustering result; the classification model is incrementally learned and trained according to the incremental clustering result to determine the category label based on the trained classification model, which can not only determine the category label of the battery data group under the newly added category and improve the accuracy of label determination, but also enable the battery classification information to be expanded based on the newly added battery data group to improve the generalization of label determination.

[0079] For ease of understanding, the present embodiment takes the DBSCAN model as a clustering model and the RBFNN model as a classification model as an example to illustrate the specific process of determining the category label of the battery data group.

[0080] Figure 7 FIG. 1 is a flow chart of determining a category label according to an embodiment of the present invention. Figure 7 As shown, the process of determining category labels in this implementation includes three stages. The first stage is to train the clustering model. The second stage is to build a classification model based on the clustering model and deploy the classification model. The third stage is to classify based on the deployed classification model and the pre-trained clustering model to determine the category labels.

[0081] In the first stage, in this embodiment, the source data set T_dt_his is first obtained, and then the source data set T_dt_his is preprocessed to determine the corresponding historical data set TN_dt_his, and the DBSCAN model is unsupervisedly trained based on the historical data set TN_dt_his. Among them, the source data set includes multiple battery data groups within the historical time period. The battery data group in the source data set can be all battery static information data uploaded by all charging vehicles within one year before the model training. Each battery static information data corresponds to a vehicle, and each battery static information data includes battery material type (brm_battery_type), battery rated capacity (brm_battery_power), battery rated total voltage (brm_battery_voltage), battery maximum operating temperature (bcp_max_temp), battery cell upper limit voltage (bcp_max_voltage), system maximum total voltage (bcp_total_voltage), battery nominal total energy (bcp_max_power) and other battery characteristic data, and each battery static information can be recorded as {bbt, bbp, bbv, bmt, bmv, btv, bmp}.

[0082] When preprocessing the source data set, the same data elimination, data filtering, and data normalization methods in the aforementioned battery flow preprocessing can be used to process the battery characteristic data in the source data set and obtain a normalized battery data group. The normalized battery data group can be recorded as {n_bbt, n_bbp, n_bbv, n_bmt, n_bmv, n_btv, n_bmp}.

[0083] Figure 8 FIG. 1 is a flow chart of a training clustering model according to an embodiment of the present invention. Figure 8 As shown in , the DBSCAN model of this embodiment is trained using an unsupervised training method, which specifically includes the following steps: In step S810, model parameters are determined.

[0084] In this embodiment, at the beginning of the training process, the initial values ​​of the model parameters eps and MinPts are determined. Optionally, during the first training process, the initial values ​​of the parameters eps and MinPts can be selected based on experience, actual business processing requirements, etc. For example, the initial value of the parameter Eps can be set to eps_0 based on experience, and the initial value of the parameter MinPts can be selected based on business product requirements. For example, when the number of points in the classified neighborhood area is less than 100, that is, the number of mapped vehicles is small, on the one hand, it is likely to be an unreliable error clustering, and on the other hand, the new model needs to be further accumulated in quantity, so as to facilitate on-demand attention and in-depth analysis in the operation and use of business products, then the initial value of MinPts can be set to 100. At the same time, the Euclidean distance is selected to measure the distance between different neighborhoods to directly express the spatial distance characteristics.

[0085] At the same time, since the model training process is usually iterative training, new model parameters will be generated after each training is completed. Therefore, in this embodiment, the model parameters will be frozen after each training is completed, and the initial values ​​of the model parameters of the next training round will be determined based on the model parameters of the current training round.

[0086] In step S820, clustering is performed.

[0087] In this embodiment, DBACAN clustering is performed based on the above initial or parameters to generate an initial clustering result. In the initial clustering result, each sample point (i.e., battery data group) is assigned to a cluster, each cluster has a corresponding cluster number, i.e., the cluster number to which the sample belongs, and noise points are marked. At the same time, the clustering result information db_produce_x corresponding to the initial clustering result is as follows: Fig. 9 As shown in , the clustering result information also contains a specific list of sample points in each cluster. Among them, info_nums is the number of vehicles corresponding to each type of battery source data group after clustering (that is, the number of vehicles corresponding to different categories); type_counts is the number of vehicles corresponding to the cluster number (that is, the number of vehicles corresponding to the historical data set); map_typee is the cluster number, and x in db_produce_x is the serial number in the parameter adjustment process (that is, the iterative training process).

[0088] In step S830, the evaluation is verified.

[0089] Optionally, in this embodiment, the contour period is used to measure the closeness and separation of data points in the cluster to judge the quality of the clustering effect; at the same time, the data quality problems that affect the clustering effect are located and analyzed by checking the entire data interaction, upload and other links to ensure that clustering is based on high-quality data; the battery type is also confirmed by preset rules or manual verification to verify whether the battery data groups under the same category are clustered into a cluster in the clustering result. In addition, the final clustering parameters are continuously adjusted to determine.

[0090] Further, after the verification method determines that the current clustering effect is good and meets the verification target, the present embodiment continues to execute step S840. If the clustering effect does not meet the verification target, the present embodiment continues to adjust the clustering model parameters and returns to execute step S810.

[0091] In step S840, the model parameters and clustering information table are frozen.

[0092] In this embodiment, after the clustering result corresponding to the DBACAN model of the current training round meets the verification target, the model parameters eps and MinPts at the end of the training round will be frozen to determine the final parameters eps_final and MinPts_final. At the same time, the current clustering result (i.e., the clustering base result) is tabled to generate the final clustering information table db_produce_## (## represents the sequence number of the iterative training process).

[0093] In step S850, a cluster center point information table is generated.

[0094] In this embodiment, since the battery characteristic data in the clustering information table are all in a normalized data format, after determining the final clustering information table db_produce_##, the data is restored by performing a denormalization operation on the clustering information table db_produce_##, and the restored clustering information table is determined as the clustering basic information table db_produce_base, so that subsequent classification can be performed based on the existing categories in the clustering basic information table db_produce_base and the category labels of each existing category, wherein each existing category corresponds to a cluster cluster in the clustering result, and each cluster cluster has a corresponding cluster center.

[0095] Furthermore, after training and determining the DBACAN model, we can enter the second stage to build and deploy a classification model based on the clustering model.

[0096] When building a classification model based on DBACAN, in this embodiment, the cluster center information of each cluster in the clustering result corresponding to DBACAN will be extracted first. Optionally, the center point of the cluster in this embodiment can be represented by the mean of all data points in the corresponding cluster. When determining the cluster center point, the feature data of each battery data group in the cluster basic information table db_produce_base under different feature dimensions is used, and the calculation is performed by the following calculation formula: Among them, f ave represents the mean of the feature data, n is the number of the feature data, Cn is the weight of the feature data, and f is the value of the feature data.

[0097] After determining the mean of the feature data under each feature dimension, the cluster center point of the corresponding cluster can be determined. Fig.10 The information table db_center for the cluster center of each cluster is shown in . In the cluster center information table of the cluster cluster, each row corresponds to the cluster center of a cluster cluster, and each cluster center is represented by the mean of feature data under multiple feature dimensions, and each feature data mean is normalized data.

[0098] Then, in this embodiment, the cluster center point information of each cluster is combined, and the RBFNN model is built based on the determination method of the aforementioned classification model. The input layer of the RBFNN network matches the number of battery feature data in the battery data group with the category label to be determined, and the hidden layer and the output layer match the center point information of each cluster.

[0099] After the RBFNN model structure is built, the RBFNN model training begins. During the model training process, the RBFNN model will be initialized first in this embodiment. The initialization process includes applying the cluster center point vector of each cluster corresponding to DBACAN to the hidden layer basis function of the RBFNN model to determine the center point μ of the basis function; using the average distance from the data sample (i.e., the battery data group) in each cluster to the cluster center as the initial value of the width parameter σ of the basis function; randomly initializing the weights and biases between the hidden layer and the output layer; based on heuristic experience, initializing and setting parameters such as the learning rate and number of iterations.

[0100] Afterwards, based on the constructed network structure and initialization parameter values, the RBFNN model network is trained, and the K-fold cross validation is used to evaluate the training of the trained RBFNN model. During the K-fold cross training, the basis function, bandwidth parameter, learning rate and number of iterations remain unchanged throughout the training, and the weights and bias information between the hidden layer and the output layer are initialized and randomly adjusted during each K-fold training; the effect is evaluated based on the classification accuracy of the model on the validation set after K-fold training. When the K-fold test accuracy is stably maintained above the preset accuracy threshold c% (c is the tolerance coefficient value set based on business and product requirements), it is determined that the RBFNN model has achieved the expected effect, and the RBFNN model is ready for deployment and application.

[0101] Optionally, in order to improve the timeliness of classification processing, in this embodiment, the RBFNN model that meets the deployment application conditions will be deployed in the real-time computing engine to determine the category label of the battery data group based on the deployed RBFNN model.

[0102] Furthermore, after the RBFNN model is deployed, the third stage of the process of determining the category label is entered in this embodiment.

[0103] In the third stage, the present embodiment firstly uses the aforementioned preprocessing method to process each battery static source data in the original stream data, and obtains the preprocessed battery stream data X_dt_stream, which includes multiple battery data groups. Afterwards, each battery data group in the battery stream data input is processed by the RBFNN model to the input model, and the confidence of the battery data group under each existing category is determined. Each existing category has a corresponding cluster cluster in the clustering result of the DBSCAN model.

[0104] When the confidence is greater than the confidence threshold, it indicates that the corresponding battery data group belongs to an existing category. At this time, it can be determined that the category label of the battery data group is an existing label, and the existing label is an existing category corresponding to the confidence greater than the confidence threshold. When the confidence is greater than the confidence threshold, it indicates that the corresponding battery data group does not belong to an existing category but to a newly added category. At this time, it can be determined that the category label of the battery data group is a new label, and it is necessary to continue to determine which newly added category the battery data group specifically belongs to.

[0105] Furthermore, after determining the category label of each battery data group in the battery flow data, in this embodiment, the battery data groups in the battery flow data are classified into two categories according to the preliminary determination results of the category labels. One category is the battery data groups belonging to different existing categories with corresponding existing labels. These battery data groups with existing labels are collectively referred to as classified data information, and are stored in the hive storage table after denormalization, and finally form the classified data information table hive_dt_cla table for use by other algorithms and analyses; the other category is the new battery data groups that do not belong to the existing categories. These battery data groups together constitute the incremental data set and are stored in the incremental data information table hive_dt_discla table.

[0106] Afterwards, for the new category information in the hive_dt_discla table, the DBSCAN model with the same eps and MinPts parameters as the historical data is used for clustering to generate incremental clustering results. The same method of extracting cluster center point information is used to extract the cluster center point information corresponding to each new cluster in the incremental clustering results, and the following is obtained: Fig.11 The incremental clustering information table hive_add_dt_cla is shown, where map_type_add is defined as an incremental information category to distinguish historical classification information, that is, map_type_add is [1, 2, 3...].

[0107] After determining the incremental clustering information table hive_add_dt_cla, in this embodiment, based on the following Fig.12 The method shown in is used to determine the input data source for incremental learning of the classification model, providing data support for the incremental learning training of the classification model.

[0108] When determining the input data source for incremental learning of the classification model, in the first step, the feature data of the classification data information table hive_dt_cla and the clustering basic information table db_produce_base are matched, and information with the same category but data groups that cannot be completely matched is selected to form a new abnormal classification information table hive_dt_cla_diff.

[0109] In the second step, extract the maximum value of the classification number map_type_max in the clustering basic information table db_produce_base, and based on this value as the base, add the value to the newly added classification category map_type_add in the incremental clustering information table hive_add_dt_cla to form an incremental cluster number to update the incremental clustering information table hive_add_dt_cla. If the historical classification number is 100, the newly added categories are 101, 102, etc.

[0110] In the third step, the result information of the newly added anomaly classification information table hive_dt_cla_diff and the incremental clustering information table hive_add_dt_cla are collected to form the input data source hive_dt_update for RBFNN incremental update (i.e. incremental learning).

[0111] After determining the input data source hive_dt_update for incremental learning of the classification model, in this embodiment, the model structure of the RBFNN model will be adjusted based on the classification results of the battery flow data input this time. Specifically, first, based on the incremental clustering information table hive_add_dt_cla, the aforementioned method of determining the cluster center point by the mean is used to calculate the cluster center point information set db_center_add corresponding to the incremental clustering information table hive_add_dt_cla, and determine the corresponding number of newly added clusters to be n; then expand the hidden layer basis function nodes and output layer nodes in the RBFNN model to be updated, and increase the number of nodes by n; at the same time, initialize the weights and biases of the newly added hidden layer nodes to the output layer, assign new node center point data to the new hidden layer nodes, and other parameters remain unchanged. The changes in the network structure before and after the update can refer to the aforementioned Figure 6 As shown in , I will not repeat it here.

[0112] After that, the incremental information in the input data source hive_dt_update is used as input, and the latest updated network structure is used for training. In addition, optionally, for parameters that need to be adjusted based on training verification, such as learning rate η and number of iterations N, this embodiment can use a heuristic grid query optimization method to automatically traverse and determine the model parameter combination for each iterative training. The training order is as follows: Fig.13 The table shown in . Therefore, in each stage of incremental learning, the key parameters are optimized by adopting a gridded trial training mechanism in this embodiment. By setting a reasonable parameter range and change amplitude, combined with heuristic experience, multiple candidate parameter combinations are trained and tested, and the parameter combination with the best performance is selected for deployment. If the current parameter combination does not meet the performance requirements, continue training until a satisfactory solution is found. This mechanism ensures the stability and efficiency of the model in the incremental learning process, while avoiding the blindness and subjectivity of parameter selection, and finally has the ability to automatically deploy applications.

[0113] After each iterative training, in this embodiment, K-fold cross validation is used to perform training evaluation on the trained RBFNN model. After the accuracy of the output results of the K-fold test RBFNN model is stably maintained above c%, it is determined that the corresponding RBFNN model has achieved the expected effect and is ready for deployment and application, and can be used in the classification processing of subsequent input battery flow data.

[0114] Moreover, when new battery flow data arrives, the above processing process will be executed again in this embodiment. While determining accurate category labels for the newly added battery data groups, the classification model can be continuously optimized, the battery classification information can be continuously and dynamically expanded, and the classification model can be applied to the newly added classifications, with higher generalization and versatility.

[0115] Therefore, in this embodiment, the confidence threshold judgment mechanism can be introduced to innovatively realize the differentiated processing of new abnormal regression data and new category information, that is, the new category information is clustered and the new feature extraction and collection of new abnormal regression data is performed using DBSCAN with fixed parameters, supporting RBFNN network update and parameter adjustment. When the new category data increases significantly, the number of hidden layer neurons can be automatically increased or the basis function distribution can be adjusted to adapt to more complex classification tasks; at the same time, the fixed parameter DBSCAN incremental clustering and RBFNN model parameters are optimized based on the adaptive update mechanism to ensure that it maintains a high sensitivity and accuracy to new data while maintaining the stability of historical classification, and further improves the accuracy and generalization of determining the category label.

[0116] Fig.14 Schematic diagram of a data processing device according to an embodiment of the present invention. Fig.14 As shown, the data processing device in this embodiment includes: an acquisition unit 1, a classification unit 2, a clustering unit 3 and a training unit 4. Among them, the acquisition unit 1 is used to acquire battery flow data, and the battery flow data includes multiple battery data groups. The classification unit 2 is used to input each battery data group into a pre-trained classification model for classification, determine the category label of the battery data group, and the category label is an existing label or a new label; and generate an incremental data set according to the battery data group corresponding to the new label. The clustering unit 3 is used to input the incremental data set into a pre-trained clustering model for clustering, and determine the incremental clustering result. The incremental clustering result includes at least one new cluster, and each new cluster has a corresponding cluster center point and a second category label. The training unit 4 is used to perform incremental learning training on the classification model according to the incremental clustering result, so as to determine the category label based on the trained classification model.

[0117] Fig.15 Schematic diagram of an electronic device according to an embodiment of the present invention. Fig.15 As shown, Fig.15The electronic device shown is a general data processing device, which includes a general computer hardware structure, which at least includes a processor 51 and a memory 52. ​​The processor 51 and the memory 52 are connected via a bus 53. The memory 52 is suitable for storing instructions or programs executable by the processor 51. The processor 51 can be an independent microprocessor or a collection of one or more microprocessors. Thus, the processor 51 executes the instructions stored in the memory 52, thereby executing the method flow of the embodiment of the present invention as described above to realize the processing of data and the control of other devices. The bus 53 connects the above-mentioned multiple components together, and at the same time connects the above-mentioned components to the display controller 54 and the display device and the input / output (I / O) device 55. The input / output (I / O) device 55 can be a mouse, a keyboard, a modem, a network interface, a touch input device, a somatosensory input device, a printer, and other devices known in the art. Typically, the input / output device 55 is connected to the system through an input / output (I / O) controller 56.

[0118] It should be understood by those skilled in the art that the embodiments of the present application may be provided as methods, devices (equipment) or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0119] The present application is described with reference to flowcharts of methods, apparatuses (devices) and computer program products according to embodiments of the present application. It should be understood that each process in the flowchart can be implemented by computer program instructions.

[0120] These computer program instructions may be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device that implements the process Figure 1 A function specified in a process or multiple processes.

[0121] These computer program instructions may also be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce the instructions for implementing the process Figure 1 A device that specifies functions in a process or multiple processes.

[0122] Another embodiment of the present invention relates to a non-volatile storage medium for storing a computer-readable program, wherein the computer-readable program is used for a computer to execute part or all of the above method embodiments.

[0123] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by specifying relevant hardware through a program, and the program is stored in a storage medium, including several instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0124] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A data processing method, characterized in that: The method comprises: Acquiring battery flow data, wherein the battery flow data includes a plurality of battery data groups; Input each of the battery data groups into a pre-trained classification model for classification, and determine a category label of the battery data group, wherein the category label is an existing label or a new label; Generate an incremental data set according to the battery data group corresponding to the new label; Inputting the incremental data set into a pre-trained clustering model for clustering, and determining an incremental clustering result, wherein the incremental clustering result includes at least one new cluster, and each of the new clusters has a corresponding cluster center point and a newly added category label; Incremental learning training is performed on the classification model according to the incremental clustering result to determine the category label based on the trained classification model.

2. The method according to claim 1, characterized in that The performing incremental learning training on the classification model according to the incremental clustering result includes: Adjusting the structure of the classification model according to the cluster center points of each of the new clusters; The adjusted classification model is iteratively trained according to the new category data information to determine the category label based on the trained classification model, wherein the new category data information includes the battery data group corresponding to each of the newly added category labels.

3. The method according to claim 2, characterized in that The classification model includes an input layer, a hidden layer and an output layer, and adjusting the structure of the classification model according to the cluster center points of each of the clustered new clusters includes: The number of nodes in the hidden layer and the output layer in the classification model is adjusted according to the cluster center of each of the new clusters.

4. The method according to claim 2, characterized in that: The iterative training of the adjusted classification model according to the new category data information includes: The adjusted classification model is iteratively trained according to the classified data information and the new category data information, wherein the classified data information is determined according to the battery data group corresponding to each of the existing labels.

5. The method according to claim 1, characterized in that The clustering model is determined based on the following method: Acquire a historical data set, wherein the historical data set includes a plurality of battery data groups within a historical time period; The battery data group in the historical data set is clustered based on a preset clustering algorithm to determine a historical clustering result and a clustering model, wherein the historical clustering result includes at least one cluster, and each cluster has a corresponding cluster center point and an existing label.

6. The method according to claim 5, characterized in that The classification model is determined based on the following method: Building a classification model according to the cluster center points of each cluster in the historical clustering results; Determine a training data set according to the battery data set corresponding to each of the existing labels; The classification model is iteratively trained according to the training data set until the classification model obtained by training meets the preset conditions.

7. The method according to claim 1, characterized in that The step of inputting each of the battery data groups into a pre-trained classification model for classification and determining the category label of the battery data group comprises: Inputting the battery data set into a pre-trained classification model for classification, and determining the confidence of the battery data set under each existing category; In response to the confidence reaching a confidence threshold, determining an existing category corresponding to the confidence reaching the confidence threshold as a category label of the battery data group; In response to each of the confidences being less than a confidence threshold, determining the category label of the battery data set to be a new label.

8. The method according to claim 1, characterized in that The obtaining of battery flow data comprises: Acquire original stream data, where the original stream data includes multiple battery static source data; Preprocessing is performed on each of the battery static source data to determine a corresponding battery data group, so as to generate battery flow data according to each of the preprocessed battery data groups.

9. The method according to claim 1, characterized in that: The battery data group includes battery material type, battery rated capacity, battery rated total voltage, battery maximum operating temperature, battery cell upper limit voltage, system maximum total voltage, and battery nominal total energy.

10. The method according to claim 1, characterized in that The clustering model adopts the DBSCAN model, and the classification model adopts the radial basis neural network RBFNN model.

11. A data processing device, characterized in that: The device comprises: An acquisition unit, configured to acquire battery flow data, wherein the battery flow data includes a plurality of battery data groups; A classification unit, configured to input each of the battery data groups into a pre-trained classification model for classification, determine a category label of the battery data group, wherein the category label is an existing label or a new label; and generate an incremental data set according to the battery data group corresponding to the new label; A clustering unit, configured to input the incremental data set into a pre-trained clustering model for clustering, and determine an incremental clustering result, wherein the incremental clustering result includes at least one new cluster, each of the new clusters having a corresponding cluster center point and a second category label; A training unit is used to perform incremental learning training on the classification model according to the incremental clustering result, so as to determine the category label based on the trained classification model.

12. A computer program product, characterized in that The computer program product comprises a computer program / instructions, and when the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 10 is implemented.

13. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method according to any one of claims 1 to 10.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images