Evaluation model training method, battery evaluation method, electronic device and program product

Abstract

The present application is applicable to the technical field of battery evaluation, and provides an evaluation model training method, a battery evaluation method, an electronic device and a program product, comprising: acquiring battery data corresponding to a source domain to obtain first battery data, and acquiring battery data corresponding to a target domain to obtain second battery data, the battery working conditions corresponding to the source domain and the target domain are different, and the battery data reflect the charging performance and / or discharging performance of the battery; constructing a training sample set, wherein the training sample set comprises cross samples constructed according to the first battery data and the second battery data; and training an evaluation model according to the training sample set to obtain the trained evaluation model. The accuracy of the evaluation model can be improved.

Description

Technical Field

[0001] This application belongs to the technical field of battery evaluation, and particularly relates to an evaluation model training method, a battery evaluation method, an electronic device, and a program product. Background Art

[0002] Under the background of the global green energy transformation, the energy storage battery industry is rising rapidly. Energy storage batteries such as lithium-ion batteries are widely used in fields such as electric vehicles and smart grids, providing strong support for the green energy transformation. Since the health state of the battery directly affects the reliability, safety, and service life of the battery, evaluating the health state of the energy storage battery plays a crucial role in the safe and efficient operation of the energy storage battery and the sustainable development of the energy storage battery industry.

[0003] Currently, the health state of energy storage batteries is usually evaluated through an evaluation model. However, the current evaluation model usually assumes that the energy storage battery works under the same conditions, which does not conform to the working conditions of the energy storage battery in actual applications, resulting in poor accuracy of the evaluation results of the evaluation model. Summary of the Invention

[0004] Embodiments of this application provide an evaluation model training method, a battery evaluation method, an electronic device, and a program product, which can improve the accuracy of the evaluation model.

[0005] In a first aspect, an embodiment of this application provides an evaluation model training method, including:

[0006] Obtain battery data corresponding to the source domain to obtain first battery data, and obtain battery data corresponding to the target domain to obtain second battery data. The battery working conditions corresponding to the source domain and the target domain are different, and the battery data reflects the charging performance and / or discharging performance of the battery;

[0007] Construct a training sample set, where the training sample set includes cross samples constructed based on the first battery data and the second battery data;

[0008] Train an evaluation model according to the training sample set to obtain the trained evaluation model.

[0009] In a second aspect, an embodiment of this application provides a battery evaluation method, including:

[0010] Obtain battery data of the battery to be evaluated to obtain battery data to be evaluated;

[0011] Use the battery data to be evaluated as the input of the evaluation model to obtain an evaluation result output by the evaluation model. The evaluation result is used to indicate the health state of the battery to be evaluated, and the evaluation model is trained according to the evaluation model training method described in the first aspect above.

[0012] In a third aspect, an embodiment of the present application provides an evaluation model training device, including:

[0013] A battery data acquisition module, configured to acquire battery data corresponding to a source domain to obtain first battery data, and acquire battery data corresponding to a target domain to obtain second battery data, where the battery operating conditions corresponding to the source domain and the target domain are different, and the battery data reflects the charging performance and / or discharging performance of the battery;

[0014] A sample construction module, configured to construct a training sample set, where the training sample set includes cross samples constructed based on the first battery data and the second battery data;

[0015] A training module, configured to train an evaluation model according to the training sample set to obtain the trained evaluation model.

[0016] In a fourth aspect, an embodiment of the present application provides a battery evaluation device, including:

[0017] A battery data to be evaluated acquisition module, configured to acquire battery data of a battery to be evaluated to obtain battery data to be evaluated;

[0018] An evaluation module, configured to use the battery data to be evaluated as an input of the evaluation model to obtain an evaluation result output by the evaluation model, where the evaluation result is used to indicate the health status of the battery to be evaluated, and the evaluation model is trained according to the evaluation model training method described in the first aspect above.

[0019] In a fifth aspect, an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where when the processor executes the computer program, the steps of the evaluation model training method described in the first aspect above or the battery evaluation method described in the second aspect above are implemented.

[0020] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, where the computer storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the evaluation model training method described in the first aspect above or the battery evaluation method described in the second aspect above are implemented.

[0021] In a seventh aspect, an embodiment of the present application provides a computer program product, which when running on an electronic device causes the electronic device to execute the evaluation model training method described in the first aspect above or the battery evaluation method described in the second aspect above.

[0022] The beneficial effects of the embodiments of the present application compared with the prior art are:

[0023] In the embodiments of the present application, after obtaining the first battery data corresponding to the source domain and the second battery data corresponding to the target domain, a cross sample is constructed according to the first battery data and the second battery data. Since the battery data can reflect the charging performance and / or discharging performance of the battery, that is, the battery data can reflect the health state of the battery, and since the source domain and the target domain correspond to different battery operating conditions, the cross sample constructed from the first battery data and the second battery data can reflect the health state of the battery under different battery operating conditions. Furthermore, training the evaluation model according to the training sample set composed of the cross samples can enable the evaluation model to more comprehensively learn the state and performance changes of the battery during the training process, thereby being able to better enhance the adaptability of the evaluation model to other battery operating conditions, improve the generalization ability and evaluation accuracy of the evaluation model, and at the same time help reduce the data acquisition cost and improve the applicability of the evaluation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below.

[0025] Figure 1 is a schematic flowchart of a method for training an evaluation model provided by an embodiment of the present application;

[0026] Figure 2 is a schematic structural diagram of a spatio-temporal hypergraph provided by an embodiment of the present application;

[0027] Figure 3 is a schematic flowchart of a battery evaluation method provided by an embodiment of the present application;

[0028] Figure 4 is a schematic block diagram of a device for training an evaluation model provided by an embodiment of the present application;

[0029] Figure 5 is a schematic block diagram of a battery evaluation device provided by an embodiment of the present application;

[0030] Figure 6 is a schematic structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] In the following description, specific details such as specific system structures and technologies are presented for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0032] It should be understood that, as used in the specification of this application and the appended claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0033] It should also be understood that the term "and / or" used in the specification of this application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0034] In addition, in the description of the specification of this application and the appended claims, the terms "first", "second", "third", etc. are only used for differential description and should not be construed as indicating or implying relative importance.

[0035] Reference to "one embodiment" or "some embodiments" etc. described in the specification of this application means that a specific feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways.

[0036] Embodiment 1:

[0037] Figure 1 The flowchart of a method for training an evaluation model provided by an embodiment of this application is shown and described in detail as follows:

[0038] S101, obtain battery data corresponding to the source domain to obtain first battery data, and obtain battery data corresponding to the target domain to obtain second battery data. The battery working conditions corresponding to the above-mentioned source domain and the above-mentioned target domain are different, and the above-mentioned battery data reflects the charging performance and / or discharging performance of the battery.

[0039] The above-mentioned source domain and target domain correspond to different battery working conditions. The battery working conditions may include external conditions (such as temperature, humidity, pressure, or scenario type, etc.) and / or internal conditions (such as electrolyte concentration or working current, etc.) when the battery is working.

[0040] It should be understood that the source domain and the target domain can be determined according to actual application requirements.

[0041] For example, the source domain can be the actual application scenario of the evaluation model, and the target domain can be an application scenario different from the actual application scenario and corresponding to different battery operating conditions in the actual application scenario, so that the evaluation model can learn the state and performance changes of the battery under different battery operating conditions in other application scenarios, thereby being able to better improve the adaptability of the evaluation model in various application scenarios and various battery operating conditions.

[0042] In some embodiments, the batteries corresponding to the source domain and the target domain can be the same or different batteries, where the differences in the batteries include, but are not limited to, differences in aspects such as brand, type, model, and capacity.

[0043] The above battery data can include charging data and / or discharging data. The charging data includes, but is not limited to, data such as charging efficiency, charging rate, charging capacity, and charging voltage; the discharging data includes, but is not limited to, data such as discharging efficiency, discharging rate, discharging capacity, and discharging voltage.

[0044] Optionally, the above battery data can be time series data (i.e., data based on time order), reflecting the charging performance and / or discharging performance of the battery at different time nodes. For example, the battery data can include the charging data of multiple charging cycles of the battery, and / or, include the discharging data of multiple discharging cycles of the battery.

[0045] Optionally, to ensure that the evaluation model can better learn the state and performance changes of the battery under different operating conditions during training, the battery data corresponding to the source domain and the battery data corresponding to the target domain (i.e., the first battery data and the second battery data) can include the same type of data, such as both including charging data, or both including charging data and discharging data.

[0046] Optionally, before obtaining the battery data corresponding to the source domain and the battery data corresponding to the target domain, common features (such as key variables related to charging) that have a greater impact in both the source domain and the target domain can be analyzed according to the actual application scenario, and the data included in the battery data (such as the saturation duration of the charging voltage) can be determined according to this common feature.

[0047] It should be noted that one or more battery data corresponding to the source domain can be obtained according to the application requirements of the evaluation model to obtain the first battery data, and, one or more battery data corresponding to the target domain can be obtained according to the application requirements to obtain the second battery data. For example, obtain the battery data corresponding to the batteries under the first battery operating conditions in multiple application scenarios to obtain multiple types of first battery data (each application scenario corresponds to one type of first battery data). And / or, obtain the battery data corresponding to the batteries under the second battery operating conditions in multiple application scenarios to obtain multiple types of second battery data (each application scenario corresponds to one type of second battery data).

[0048] The application requirements of the evaluation model, also known as evaluation requirements, are used to reflect the requirements for evaluating the health status of the battery. Optionally, the application requirements include, but are not limited to, one or more of the requirements such as model requirements (e.g., the accuracy, generalization ability, or robustness of the model) and scenario requirements (e.g., applied to the source domain or applied to the target domain, etc.). For example, the application requirements may include: applied to the source domain, with strong generalization ability for unknown battery operating conditions.

[0049] S102, construct a training sample set, where the training sample set includes cross samples constructed based on the first battery data and the second battery data.

[0050] It should be understood that in the embodiments of the present application, each cross sample includes data in the first battery data and data in the second battery data. It should be noted that the cross sample can be constructed based on partial data in the first battery data and partial data in the second battery data.

[0051] In some embodiments, the training sample set may further include source domain samples constructed based on the first battery data, and / or target domain samples constructed based on the second battery data.

[0052] In some embodiments, the quantities corresponding to the cross samples, source domain samples, and target domain samples can be determined first, and then the corresponding quantities of samples can be constructed according to the quantities of various samples to obtain the training sample set. Optionally, the quantities corresponding to the cross samples, source domain samples, and target domain samples can be determined according to the application requirements of the evaluation model by means of intelligent algorithms, convolutional neural networks, or large models, etc.; or the quantities corresponding to the cross samples, source domain samples, and target domain samples can be determined according to the quantities sent by the user's corresponding terminal device (such as a mobile phone).

[0053] In some embodiments, considering that there may be differences such as dimensions between the first battery data and the second battery data, before constructing the training sample set, the first battery data and the second battery data can be preprocessed such as standardized to ensure the consistency of the first battery data and the second battery data through preprocessing, and avoid situations such as bias or inability to generalize during the training of the evaluation model caused by data differences, thereby improving the training accuracy of the evaluation model training.

[0054] In the embodiments of the present application, since the battery data can reflect the charging performance and / or discharging performance of the battery, that is, reflect the health status of the battery, and since the source domain and the target domain correspond to different battery operating conditions, therefore, cross samples are constructed based on the first battery data and the second battery data, and this cross can simultaneously reflect the health status of the battery under different battery operating conditions, and at the same time can better solve problems such as poor generalization ability and applicability of the model caused by single battery data or scarcity of battery data, effectively ensuring the training effect of the evaluation model.

[0055] S103. Train the evaluation model according to the above training sample set to obtain the trained evaluation model.

[0056] Optionally, the evaluation model can be constructed based on a convolutional neural network, a recurrent neural network, a time series network, a graph neural network, or a large model. The embodiments of the present application do not make specific limitations thereto.

[0057] In some embodiments, the evaluation model can be a model based on a spatial attention mechanism.

[0058] In some embodiments, during the process of training the evaluation model according to the training sample set, the evaluation model is updated based on the evaluation result of the evaluation model. After obtaining the updated evaluation model, if the updated evaluation model does not meet the training requirements (such as the accuracy reaching the accuracy threshold and the number of iterations reaching the number threshold, etc.), the updated evaluation model can be iterated based on the training sample set until the latest updated evaluation model meets the training requirements, and it is used as the trained evaluation model to evaluate the health state of the battery.

[0059] In the embodiments of the present application, since the battery data can reflect the charging performance and / or discharging performance of the battery, that is, the battery data can reflect the health state of the battery, and since the source domain and the target domain correspond to different battery operating conditions, therefore, by constructing cross samples based on the first battery data and the second battery data corresponding to the source domain, the obtained cross samples can reflect the health state of the battery under different battery operating conditions. Furthermore, training the evaluation model according to the training sample set composed of the cross samples can enable the evaluation model to more comprehensively learn the state and performance changes of the battery during the training process, thereby being able to better enhance the adaptability of the evaluation model to other battery operating conditions, improve the generalization ability and evaluation accuracy of the evaluation model, and at the same time be beneficial to reducing the data acquisition cost and improving the applicability of the evaluation model, which is beneficial to practical applications.

[0060] In some embodiments, the above step S102 includes:

[0061] A1. Construct a graph structure according to the above first battery data and the above second battery data to obtain the above cross samples.

[0062] A2. Determine the above training sample set according to the above cross samples.

[0063] The graph structure refers to a data structure composed of nodes (representing entities) and edges (representing the relationships between nodes), including but not limited to graph structures such as battery state graphs, battery attribute relationship graphs, or spatio-temporal hypergraphs.

[0064] It should be understood that when constructing source domain samples and / or target domain samples, a graph structure can be constructed according to the first battery data to obtain source domain samples, and a graph structure can be constructed according to the second battery data to obtain target domain samples.

[0065] Since the graph structure can more intuitively reflect the complex relationships during the battery charging and discharging process and the correlation between the changes in battery data and the health state of the battery during the charging and discharging process, therefore, constructing samples in the form of a graph structure according to the first battery data and the second battery data to train the evaluation model can enable the evaluation model to effectively capture these complex relationships and various correlations during the training process, thereby improving the evaluation accuracy of the battery health state.

[0066] In some embodiments, since a spatio-temporal hypergraph can connect each node in a fully connected manner and can combine the temporal and spatial correlation relationships between nodes, without pre-defining a graph structure according to a set rule or assumption to specify the spatial correlation between nodes, therefore, a spatio-temporal hypergraph can be constructed according to the first battery data and the second battery data to obtain cross samples. Correspondingly, when constructing source domain samples according to the first battery data and target domain samples according to the second battery data, they are also constructed into the data structure of a spatio-temporal hypergraph.

[0067] Constructing samples in the above manner enables the evaluation model to better learn the temporal and spatial dependence relationships of battery data based on the spatio-temporal hypergraph during the training process, effectively improving the accuracy of the evaluation model. At the same time, it can effectively avoid the problem that the use of a pre-defined fixed graph structure may cause the inability to dynamically adjust the connection relationship between nodes according to the actual battery data, and avoid problems such as the possible inability to cover comprehensively when defining the graph structure based on artificial experience or prior knowledge, further improving the accuracy of the evaluation model.

[0068] In some embodiments, the above step A1 includes:

[0069] A11. Determine the first proportion corresponding to the first battery data and the second proportion corresponding to the second battery data according to the application requirements of the above evaluation model.

[0070] A12. Construct the above graph structure according to the above first battery data, the above first proportion, the above second battery data, and the above second proportion to obtain the above cross samples.

[0071] In the embodiments of the present application, the application requirements at least include scenario requirements, so as to construct appropriate cross samples for training according to the scenario requirements of the evaluation model, ensuring that the evaluation model can have good evaluation performance while meeting the scenario requirements.

[0072] It should be understood that the first battery data and the second battery data are generally battery data obtained according to the same data acquisition rule, that is, in the embodiments of the present application, the first battery data and the second battery data are generally time series data with the same length. To maintain data consistency, when constructing a cross sample based on the first battery data and the second battery data, a cross sample can be constructed according to partial data in the first battery data and partial data in the second battery data, so that the length of the cross sample is consistent with the length of the obtained battery data.

[0073] Wherein, in the process of constructing the sample, first determine the proportion of the first battery data corresponding to the source domain in the cross sample according to the application requirements of the evaluation model to obtain the first proportion, and determine the proportion of the second battery data corresponding to the target domain in the cross sample according to the application requirements of the evaluation model to obtain the second proportion.

[0074] After determining the corresponding proportions of each battery data (i.e., the first battery data and the second battery data) in the cross sample, a graph structure can be constructed by combining each battery data and its corresponding proportion to obtain the required cross sample.

[0075] It should be understood that when the battery data includes multiple types of first battery data and / or multiple types of second battery data, the corresponding proportions of each type of battery data can be calculated according to the application requirements, and the sum of the corresponding proportions of each type of battery data is usually 1.

[0076] In some embodiments, the above first proportion and the above second proportion can be dynamically adjusted according to the data volume or acquisition cost of the battery data corresponding to the target domain.

[0077] In some embodiments, the application requirements may further include model requirements. Determine the first proportion and the second proportion according to the scenario requirements in the application requirements, and then perform weighted processing on the first proportion and the second proportion according to the model requirements in the application requirements to obtain the weighted first proportion and the weighted second proportion, and perform subsequent processing based on the weighted first proportion and the weighted second proportion.

[0078] In the embodiments of the present application, by controlling the proportions of the battery data in different domains in the cross sample according to the requirements of the application scenario of the evaluation model, the evaluation model can better learn the commonalities and differences between different domains based on the cross sample, so as to better solve problems such as generalization or data scarcity, and effectively control the performance of the trained evaluation model.

[0079] In some embodiments, the above step A12 includes:

[0080] Determine the first data to be embedded according to the above first battery data and the first proportion, and determine the second data to be embedded according to the above second battery data and the second proportion.

[0081] Construct the above graph structure based on the above first data to be embedded, the first embedding method, the above second data to be embedded, and the second embedding method to obtain the above cross-samples. The above first embedding method and the above second embedding method are determined according to the above application requirements.

[0082] It should be understood that the first embedding method and the second embedding method can be used to reflect the way of embedding the corresponding battery data into the graph structure, such as the position or level of embedding into the graph structure, etc. In the embodiments of the present application, the embedding method can be used to reflect the position of the corresponding battery data embedded in the graph structure.

[0083] Since the position of the battery data in the graph structure is usually related to the relevance and importance of the data, etc., the battery data with a higher position can be regarded as more important features, that is, the position of the battery data in the graph structure can affect the understanding and utilization direction of these battery data by the evaluation model, etc. Therefore, before constructing the graph structure based on the first battery data and the second battery data, the ways of embedding the data corresponding to the first battery data and the data corresponding to the second battery data into the graph structure can be determined according to the application requirements to obtain the first embedding method and the second embedding method. Subsequently, based on the first embedding method and the second embedding method, the first battery data and the second battery data are embedded into the graph structure, so that the obtained graph structure preferentially reflects the relevance and laws between important battery data, thereby being able to affect the learning effect of the evaluation model in a controllable manner, and thus controlling the performance of the evaluation model.

[0084] Optionally, the first data to be embedded can be determined by combining the application requirements, the first battery data, and the first ratio, and the second data to be embedded can be determined by combining the application requirements, the second battery data, and the second ratio.

[0085] For example, assume that the application requirements include: mainly applied to the battery evaluation of the target domain. At this time, the first battery data in the source domain is mainly used to achieve transfer learning, provide a benchmark or reference, etc. The corresponding data with a later time in the first battery data can be determined as the first data to be embedded according to the application requirements, the first battery data, and the first ratio, and the corresponding data with an earlier time in the second battery data can be determined as the second data to be embedded according to the application requirements, the second battery data, and the second ratio.

[0086] Assume that the first battery data includes the charging data of 4 charging cycles of the battery corresponding to the source domain (assumed to be denoted as N 11 , N 12、 N 13 and N 14 ), and the first ratio is 25%. Then, according to the above application requirements, 25% of the data with a later time in the first battery data (i.e., the charging data N 1-4 ) of the 4th charging cycle can be determined as the second data to be embedded.

[0087] Suppose the second battery data includes the charging data of the battery corresponding to the target domain for 4 charging cycles (denoted as N 21 , N 22、 N 23 and N 24 ), and the second proportion is 75%. Then, according to the above application requirements, the second data to be embedded can be 75% of the data in the second battery data with earlier time (i.e., the charging data of the 1st to 3rd charging cycles: N 21 -N 23 ) determined as the second data to be embedded.

[0088] At this time, according to N 21 , N 22 , N 23 and N 14 a graph structure is constructed. Suppose according to the above application requirements (i.e., mainly applied to the battery evaluation of the target domain), it can be determined that the first embedding method is to embed the nodes at the later positions in the graph structure, and the second embedding method is to embed at the earlier positions in the graph structure, that is, embed the first data to be embedded into the nodes at the earlier positions in the graph structure, and embed the second data to be embedded into the nodes at the later positions in the graph structure.

[0089] For another example, suppose the application requirements include: mainly used for the battery evaluation of the source domain. At this time, the second battery data corresponding to the target domain is mainly used to help the evaluation model learn the performance changes of batteries different from those in the source domain, and help the model adapt to the data differences from the source domain to the target domain. According to the above application requirements, it can be determined that the first embedding method is to embed at the earlier positions in the graph structure, and the second embedding method is random embedding, that is, determine the nodes corresponding to the second data to be embedded in the graph structure in a random manner, so that the nodes corresponding to the target domain are randomly distributed in the graph structure, thereby being able to better simulate the distribution differences between different domains and being beneficial to improving the generalization ability of the evaluation model.

[0090] In the embodiments of the present application, the positions of the battery data of different domains embedded in the graph structure are determined according to the application requirements of the evaluation model, so as to be able to control the learning effect of the evaluation model in combination with the application requirements, and significantly improve the training effect and application performance of the evaluation model.

[0091] In some embodiments, the above first battery data includes first charging data and first health values, the above second battery data includes second charging data and second health values, and the above step S101 includes:

[0092] Construct a spatio-temporal hypergraph according to the above first charging data and the above second charging data, and determine the label values corresponding to the spatio-temporal hypergraph according to the above first health values and the above second health values.

[0093] Determine the above cross samples according to the above spatio-temporal hypergraph and the above label values.

[0094] Determine the above-mentioned training sample set according to the above cross samples.

[0095] The above spatio-temporal hypergraph is a data structure that allows edges to connect any number of nodes, and these nodes can have spatio-temporal and spatial attributes simultaneously, so as to enable comprehensive connection and flexible representation of the dynamic interaction of entities (i.e., nodes) in time and space.

[0096] Optionally, the charging data may include constant-voltage charging data, charging voltage increase data, and charging voltage saturation data. It should be understood that the first battery data and the second battery data may include charging data corresponding to multiple charging cycles, so as to reflect the charging performance of the corresponding battery at different time nodes.

[0097] The constant-voltage charging data is used to reflect the charging performance of the battery under partial charging, such as the charging duration required for the battery to charge within a set voltage range (such as 3.9V - 4.2V) (referred to as the constant-voltage charging duration); the charging voltage increase data is used to reflect the voltage increase condition during the battery charging process (i.e., the change condition of the voltage difference between the real-time charging voltage and the initial charging voltage), such as the average charging voltage increase of the battery from the start of charging (such as 0% charge) to a specified state (such as 80% charge); the charging voltage saturation data is used to reflect the charging performance of the battery under full charging, such as the charging duration required for the battery to reach the charging cut-off voltage (such as 4.2V) (referred to as the saturation voltage charging duration).

[0098] It should be understood that the health value of the battery is used to reflect the health state of the battery, including but not limited to the SOH value (State of Health) of the battery or the actual capacity of the battery, etc.

[0099] Since the charging data of the battery can usually reflect information in multiple dimensions such as time, space, and electrochemical reactions, and there may be differences and commonalities in the charging data corresponding to different domains, therefore, in the embodiments of the present application, a spatio-temporal hypergraph can be constructed according to the first charging data and the second battery data, so that the finally obtained cross samples are data in the form of a spatio-temporal hypergraph. Furthermore, the evaluation model can be trained based on the spatio-temporal hypergraph that effectively reflects the complex changes during the battery charging process, improving the learning effect of the evaluation model.

[0100] Since the spatio-temporal hypergraph is constructed according to the first charging data and the second charging data, to ensure the accuracy of the label value of the cross sample, the label value of the cross sample can be comprehensively determined according to the first health value corresponding to the first charging data and the second health value corresponding to the second charging data, thereby improving the accuracy of the determined cross sample, that is, improving the accuracy of the finally constructed training sample set.

[0101] In some embodiments, the first charging data and the second charging data may include charging data corresponding to N charging cycles (N>1). The N charging cycles may be consecutive or non-consecutive. The above-mentioned charging data may include constant-voltage charging duration, average charging voltage increment, and saturation-voltage charging duration. When constructing a spatio-temporal hypergraph based on the first charging data and the second charging data, each data corresponding to each charging cycle may be used as a node, and every two nodes may be connected by an edge to obtain a spatio-temporal hypergraph.

[0102] It should be understood that when the first charging data includes charging data corresponding to N charging cycles, the corresponding first health value usually includes the health values corresponding to these N charging cycles. Since in a spatio-temporal hypergraph, label values may usually be set for each node, when determining the label values corresponding to the spatio-temporal hypergraph, the label values corresponding to each node of the spatio-temporal hypergraph may be determined according to the first health value and the second health value respectively. It should be understood that one of the charging data in each node corresponding to the spatio-temporal hypergraph, and the label value corresponding to the node is usually used to provide the health state of the battery related to the node, so that the model can more accurately understand the trend of the battery's health state changing over time.

[0103] Optionally, a cross-charging data (equivalent to the charging data composed of the first data to be embedded and the second data to be embedded) with the same time length (i.e., the number N of charging cycles in the first charging data or the second charging data) may be constructed according to partial charging data in the first charging data and partial charging data in the second charging data, and then a spatio-temporal hypergraph may be constructed according to the cross-charging data.

[0104] For example, assume that the first charging data sequentially includes charging data corresponding to 4 charging cycles N1, N2, N3, and N4: K1, K2, K3, and K4, and the second charging data sequentially includes charging data corresponding to 4 charging cycles N1, N2, N3, and N4: L1, L2, L3, and L4.

[0105] According to the application requirements, the first charging data, and the second charging data, cross-charging data with a length of 4 is constructed. For example, the cross-charging data is: K1, K2, L3, and K4. Another example is that the cross-charging data is: L1, K2, K3, and K4.

[0106] As Figure 2 shown, since each charging data includes 3 data (assumed to be constant-voltage charging duration, average charging voltage increment, and saturation-voltage charging duration), that is, the charging data corresponding to each time node corresponds to 3 nodes. Therefore, when constructing a spatio-temporal hypergraph according to the above cross-charging data, the constructed spatio-temporal hypergraph may include 12 nodes, corresponding to the constant-voltage charging duration, average charging voltage increment, and saturation-voltage charging duration of 4 time nodes respectively.

[0107] Optionally, in the process of constructing a spatiotemporal hypergraph based on the cross-charging data, the nodes corresponding to the target domain (assuming they are called target domain nodes) can be cross-domain connected with the nodes corresponding to the source domain (assuming they are called source domain nodes) in a fully connected manner, that is, each target domain node is connected to each source domain node to construct a cross-domain edge. In addition, nodes in the same domain can be connected in a fully connected manner.

[0108] Alternatively, each target domain node may be connected to a source domain target node based on similarity or prior knowledge, where the source domain target node is a source domain node that has an association relationship with the target domain node. For example, the source domain target node may be the M (e.g., 3) source domain nodes with the highest similarity to the target domain node, where M is usually greater than 0 and less than or equal to the number of source domain nodes.

[0109] In the embodiment of the present application, a spatiotemporal hypergraph is constructed as a cross sample based on the first charging data and the second charging data. Since the spatiotemporal hypergraph can effectively capture the complex patterns of battery charging data corresponding to different domains changing over time and space, it is helpful to mine potential cross-domain correlations. Therefore, by constructing cross samples in the above manner, a training sample set is obtained, so that the evaluation model can comprehensively learn the complex changes and dependencies in the battery charging process, help the evaluation model better generalize to unknown battery working conditions, and effectively improve the generalization ability and applicability of the evaluation model.

[0110] In some embodiments, before determining the label value corresponding to the spatiotemporal hypergraph according to the first health value and the second health value, the method further includes:

[0111] A first weight corresponding to the first health value is determined according to the application requirements of the evaluation model, and a second weight corresponding to the second health value is determined according to the application requirements.

[0112] Correspondingly, the determining of the label value corresponding to the spatiotemporal hypergraph according to the first health value and the second health value includes:

[0113] The label value is determined according to the first health value, the first weight, the second health value and the second weight.

[0114] Since the spatiotemporal hypergraph is constructed based on part of the first charging data and the second charging data, and combines charging data from different domains, before determining the label value corresponding to the spatiotemporal hypergraph, the weights corresponding to the first health value and the second health value can be determined first to obtain the first weight and the second weight, and then the final label value can be determined by weighted method.

[0115] Among them, since the application requirements of the evaluation model can reflect requirements such as the performance of the evaluation model, therefore, the first weight corresponding to the first health value can be determined according to the application requirements of the evaluation model, and the second weight corresponding to the second health value can be determined according to the application requirements, so that the finally determined label value can better guide the training of the evaluation model.

[0116] It should be understood that when determining the label value, it is usually necessary to determine the label value corresponding to each node of the spatio-temporal hypergraph. At this time, the determined first weight and second weight can be directly used as the weights corresponding to each node, that is, for each node, the same first weight is used to weight the corresponding first health value, and the same second weight is used to weight the corresponding second health value to calculate its label value.

[0117] Alternatively, for each node, the first weight and the second weight corresponding to the node can be determined respectively according to the domain to which the node belongs and the application requirements, and then the label value corresponding to the node is calculated by using the first weight, the first health value, the second weight and the second health value corresponding to the node.

[0118] As an example, the label value corresponding to the node of the spatio-temporal hypergraph can be expressed in the following form:

[0119] label = α1 * label1 + α2 * label2

[0120] Among them, label represents the label value corresponding to the node of the spatio-temporal hypergraph, α1 represents the first weight corresponding to the node of the spatio-temporal hypergraph, label1 represents the first health value corresponding to the node of the spatio-temporal hypergraph, α2 represents the second weight corresponding to the node of the spatio-temporal hypergraph, and label2 represents the second health value corresponding to the node of the spatio-temporal hypergraph. Optionally, the sum of α1 and α2 can be 1.

[0121] Optionally, during the process of iteratively training the evaluation model according to the constructed training sample set, the first weight and the second weight can be fine-tuned according to the evaluation result of the evaluation model and the application requirements in the previous iteration process, and the label values corresponding to each node of the spatio-temporal hypergraph in each cross-sample are updated based on the fine-tuned first weight and second weight to obtain the updated cross-sample, and then the current round of iterative training is performed based on the updated cross-sample.

[0122] In the embodiments of the present application, after determining the weights corresponding to different domains according to the evaluation requirements of the evaluation model to obtain the first weight and the second weight, the corresponding battery health values are weighted and calculated according to the first weight and the second weight to obtain the label values of each node of the spatio-temporal hypergraph, so as to improve the accuracy of the cross-sample that combines battery data from different domains, and thus can better guide the evaluation model to learn the complex relationship between battery charging data and battery health values, and improve the learning effect of the evaluation model.

[0123] In some embodiments, the above evaluation model is a model constructed based on a Fourier graph neural network, and the above step S103 includes:

[0124] Performing a Fourier graph operation on the above cross-samples through the above evaluation model, and evaluating according to the frequency-domain sample features obtained by the operation to obtain an evaluation value;

[0125] Updating the parameters of the above evaluation model according to the deviation between the above evaluation value corresponding to the above cross-samples and the true value to obtain the trained above evaluation model.

[0126] The above Fourier graph neural network, also known as the Fourier graph neural network, is a deep learning network that combines the characteristics of the Fourier transform and the graph neural network. It can utilize the characteristics of the Fourier transform to transform the data in the graph structure into the frequency-domain space for analysis, so as to better capture information such as periodicity and frequency in the data, and thus better understand the periodic patterns in the data.

[0127] The Fourier graph operation, also known as the Fourier graph operation, generally refers to an operation method that transforms the graph structure from the spatial domain to the frequency domain through transformation methods such as the Fourier transform, so as to perform operations such as point multiplication in the frequency domain.

[0128] Since the Fourier graph neural network can process the cross-samples through the Fourier graph operation to obtain frequency-domain sample features that can better reflect the features such as the frequency of the data in the cross-samples, therefore, an evaluation model is constructed based on the Fourier graph neural network. This evaluation model can perform a Fourier graph operation on the cross-samples and evaluate according to the frequency-domain sample features obtained by the operation, better capturing the features such as the periodicity or frequency of the data in the cross-samples, thereby providing a strong basis for evaluating the health state of the battery. Moreover, the combination of the graph neural network and the Fourier transform can make the evaluation model have higher accuracy and robustness when evaluating the battery health state, which is beneficial to practical applications. And, the evaluation model evaluates according to the frequency-domain sample features, which can simplify the convolution operation in the spatial domain into a point multiplication operation in the frequency domain, effectively simplifying the data processing complexity and improving the data processing efficiency.

[0129] After obtaining the evaluation value corresponding to the cross-samples, the parameters of the evaluation model can be updated through an update algorithm such as the backpropagation algorithm or the maximum marginal likelihood estimation according to the difference between the evaluation value and the corresponding true value, to obtain a trained evaluation model that can be used to evaluate the battery health state.

[0130] Optionally, when updating the parameters of the evaluation model according to the difference between the evaluation value and the corresponding true value, the difference between the evaluation value and the corresponding true value can be measured based on a preset loss function, and the parameters of the evaluation model can be optimized by minimizing this difference, so that the evaluation model can adapt to the characteristics of battery data in different domains and improve its generalization ability and applicability.

[0131] Optionally, the preset loss function can be the Maximum Mean Discrepancy (MMD) loss function. By minimizing the maximum mean discrepancy, the distribution difference between battery data in different domains can be realized, so that the trained evaluation model can better adapt to the characteristics of battery data in different domains.

[0132] It should be understood that the preset loss function can also be a loss function such as the mean square error function or covariance, etc., which can be specifically set according to the actual application scenario.

[0133] In some embodiments, a Gaussian process can be introduced into the evaluation model, and the cross-domain covariance function can be combined to capture the influence of battery data in a different domain (such as the target domain) on the battery health state evaluation. When the evaluation model evaluates the battery health state according to the frequency domain sample characteristics, the evaluation result can be adjusted by combining the influence of data in a different domain. Among them, the cross-domain covariance function is used to measure the similarity between battery data in different domains. At this time, the evaluation result is determined by the evaluation model based on the similarity between the source domain and the target domain in the frequency domain sample characteristics and the prior knowledge of the Gaussian process.

[0134] Corresponding to the evaluation model training method described in the above embodiments, Figure 3 The following shows a schematic flowchart of a battery evaluation method provided by an embodiment of the present application, which is described in detail as follows:

[0135] S301, obtain the battery data of the battery to be evaluated to obtain the battery data to be evaluated.

[0136] It should be understood that the battery working conditions corresponding to the battery to be evaluated can be the battery working conditions corresponding to the source domain or the target domain, or other battery working conditions.

[0137] It should be understood that when obtaining the battery data of the battery to be evaluated, the charging data and / or discharging data of the battery to be evaluated can be obtained to obtain the battery data to be evaluated that can reflect the charging performance and / or discharging performance to be evaluated, so that the evaluation model can better evaluate the health state of the battery to be evaluated according to the battery data to be evaluated.

[0138] S302. Use the battery data to be evaluated as the input of the evaluation model to obtain the evaluation result output by the evaluation model. The evaluation result is used to indicate the health status of the battery to be evaluated. The evaluation model is trained according to the evaluation model training method described in the above embodiments.

[0139] It should be understood that the evaluation result may include the SOH value of the battery to be evaluated or the actual capacity of the battery, etc., and is used to indicate the health status of the battery to be evaluated.

[0140] In the embodiments of the present application, since the evaluation model is trained based on the cross-samples including the first battery data and the second battery data, and the battery data reflects the charging performance and / or discharging performance of the battery, that is, reflects the health status of the battery, and since the source domain and the target domain correspond to different battery working conditions, the trained evaluation model can comprehensively understand the state and performance changes of the battery under different battery working conditions, has strong generalization ability and accuracy, and has a certain applicability to the batteries corresponding to other battery working conditions, so as to accurately analyze the health status of the battery to be evaluated according to the battery data to be evaluated and obtain a relatively accurate evaluation result.

[0141] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0142] Embodiment 2:

[0143] Corresponding to the evaluation model training method described in the above embodiments, Figure 4 The structural block diagram of the evaluation model training device provided by the embodiments of the present application is shown. For the sake of convenience of description, only the parts related to the embodiments of the present application are shown.

[0144] Refer to Figure 4 The device includes: a battery data acquisition module 41, a sample construction module 42, and a training module 43. Among them,

[0145] The battery data acquisition module is used to acquire the battery data corresponding to the source domain to obtain the first battery data, and acquire the battery data corresponding to the target domain to obtain the second battery data. The battery working conditions corresponding to the source domain and the target domain are different, and the battery data reflects the charging performance and / or discharging performance of the battery;

[0146] The sample construction module is used to construct a training sample set, and the training sample set includes cross-samples constructed according to the first battery data and the second battery data;

[0147] A training module, configured to train an evaluation model according to the above training sample set to obtain the trained evaluation model.

[0148] Corresponding to the battery evaluation method described in the above embodiments, Figure 5 The block diagram of the battery evaluation device provided by the embodiment of the present application is shown. For the sake of convenience of description, only the parts related to the embodiment of the present application are shown.

[0149] Referring to Figure 5 , the device includes: a battery data acquisition module 51 to be evaluated and an evaluation module 52. Among them,

[0150] The battery data acquisition module 51 to be evaluated is configured to acquire the battery data of the battery to be evaluated to obtain the battery data to be evaluated;

[0151] The evaluation module 52 is configured to use the above battery data to be evaluated as the input of the evaluation model to obtain the evaluation result output by the evaluation model. The above evaluation result is used to indicate the health status of the above battery to be evaluated. The above evaluation model is obtained through the above evaluation model training device.

[0152] It should be noted that the information interaction, execution process, etc. between the above devices / units, because they are based on the same concept as the method embodiment of the present application, for their specific functions and the technical effects brought, please refer to the method embodiment part for details, and will not be elaborated here.

[0153] Embodiment 3:

[0154] Figure 6 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 6 shown, the electronic device 6 of this embodiment includes: at least one processor 60 ( Figure 6 only one processor is shown in

[0155] ), a memory 61, and a computer program 62 stored in the memory 61 and executable on the at least one processor 60. When the processor 60 executes the computer program 62, the steps in any of the above method embodiments are implemented. Figure 6 The electronic device 6 may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The electronic device may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art can understand that

[0156] The so-called processor 60 may be a Central Processing Unit (CPU), and the processor 60 may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0157] In some embodiments, the memory 61 may be an internal storage unit of the electronic device 6, such as the hard disk or memory of the electronic device 6. In other embodiments, the memory 61 may also be an external storage device of the electronic device 6, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc., equipped on the electronic device 6. Further, the memory 61 may also include both the internal storage unit and the external storage device of the electronic device 6. The memory 61 is used to store an operating system, application programs, a BootLoader, data, and other programs, such as the program code of the computer program, etc. The memory 61 may also be used to temporarily store data that has been output or is to be output.

[0158] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiments can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of this application. The specific working processes of the units and modules in the above system can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.

[0159] An embodiment of the present application further provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor. When the processor executes the computer program, the steps in any of the above method embodiments are implemented.

[0160] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, which when executed by a processor can implement the steps in each of the above method embodiments.

[0161] An embodiment of the present application provides a computer program product, which when running on an electronic device enables the electronic device to implement the steps in each of the above method embodiments when executed.

[0162] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above method embodiments of the present application, a computer program can be used to instruct relevant hardware to complete. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps in each of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can at least include: any entity or device capable of carrying the computer program code to the photographing device / electronic device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk, or an optical disc, etc. In some jurisdictions, according to legislation and patent practice, the computer-readable medium cannot be an electrical carrier signal and a telecommunication signal.

[0163] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0164] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0165] In the embodiments provided in the present application, it should be understood that the disclosed device / network device and method can be implemented in other ways. For example, the device / network device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in electrical, mechanical or other forms.

[0166] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0167] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for training an evaluation model, characterized in that: include: Acquire battery data corresponding to a source domain to obtain first battery data, and acquire battery data corresponding to a target domain to obtain second battery data, wherein the battery operating conditions corresponding to the source domain and the target domain are different, and the battery data reflects charging performance and / or discharging performance of the battery; Constructing a training sample set, wherein the training sample set includes cross samples constructed according to the first battery data and the second battery data; The evaluation model is trained according to the training sample set to obtain the trained evaluation model.

2. The evaluation model training method according to claim 1, characterized in that: The step of constructing a training sample set includes: Constructing a graph structure according to the first battery data and the second battery data to obtain the cross sample; The training sample set is determined according to the cross samples.

3. The evaluation model training method according to claim 2, characterized in that: The step of constructing a graph structure according to the first battery data and the second battery data to obtain the cross sample includes: Determine a first proportion corresponding to the first battery data and a second proportion corresponding to the second battery data according to application requirements of the evaluation model; The graph structure is constructed according to the first battery data, the first proportion, the second battery data, and the second proportion to obtain the cross sample.

4. The evaluation model training method according to claim 3, characterized in that: The constructing the graph structure according to the first battery data, the first proportion, the second battery data, and the second proportion to obtain the cross sample includes: Determine first data to be embedded according to the first battery data and the first ratio, and determine second data to be embedded according to the second battery data and the second ratio; The graph structure is constructed according to the first data to be embedded, the first embedding method, the second data to be embedded, and the second embedding method to obtain the cross sample, and the first embedding method and the second embedding method are determined according to the application requirements.

5. The evaluation model training method according to claim 1, characterized in that: The first battery data includes first charging data and a first health value, the second battery data includes second charging data and a second health value, and the constructing of the training sample set includes: constructing a spatiotemporal hypergraph according to the first charging data and the second charging data, and determining a label value corresponding to the spatiotemporal hypergraph according to the first health value and the second health value; Determine the cross sample according to the spatiotemporal hypergraph and the label value; The training sample set is determined according to the cross samples.

6. The evaluation model training method according to claim 5, characterized in that: Before determining the label value corresponding to the spatiotemporal hypergraph according to the first health value and the second health value, the method further includes: Determine a first weight corresponding to the first health value according to application requirements of the evaluation model, and determine a second weight corresponding to the second health value according to the application requirements; Correspondingly, determining the label value corresponding to the spatiotemporal hypergraph according to the first health value and the second health value includes: The label value is determined according to the first health value, the first weight, the second health value, and the second weight.

7. The evaluation model training method according to any one of claims 1 to 6, characterized in that: The evaluation model is a model constructed based on a Fourier graph neural network, and the evaluation model is trained according to the training sample set to obtain the trained evaluation model, including: Performing Fourier graph operation on the cross samples through the evaluation model, evaluating according to the frequency domain sample characteristics obtained by the operation, and obtaining an evaluation value; The parameters of the evaluation model are updated according to the deviation between the evaluation value corresponding to the cross sample and the true value to obtain the trained evaluation model.

8. A battery evaluation method, characterized in that: include: Acquire battery data of the battery to be evaluated, and obtain the battery data to be evaluated; The battery data to be evaluated is used as an input of an evaluation model to obtain an evaluation result output by the evaluation model, wherein the evaluation result is used to indicate a health status of the battery to be evaluated, and the evaluation model is trained according to the evaluation model training method according to any one of claims 1 to 7.

9. An electronic 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 method according to any one of claims 1 to 7 or claim 8 is implemented.

10. A computer program product, characterized in that When the computer program product runs on an electronic device, the electronic device is enabled to execute the method according to any one of claims 1 to 7 or claim 8.

Images