A capacity prediction method, a training method and device of a capacity prediction model

Abstract

The application provides a capacity prediction method and a training method and device of a capacity prediction model, relates to the technical field of computers, and solves the technical problem that related technologies need to involve battery disassembly in the process of performing a discharge experiment, the steps are frequent and the workload is large, and the capacity of the battery cannot be quickly determined. The method comprises the following steps: acquiring the voltage of a to-be-identified battery at each moment in multiple moments, the voltage of M batteries corresponding to the to-be-identified battery at each moment, and the voltage of N batteries corresponding to the to-be-identified battery at each moment; generating target features of the to-be-identified battery at each moment based on the aggregated features of the to-be-identified battery at each moment, the internal resistance features of the to-be-identified battery at each moment, and initial features of the to-be-identified battery; and inputting the target features of the to-be-identified battery at each moment into a target capacity prediction model to obtain the capacity of the to-be-identified battery at each moment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, and particularly relates to a capacity prediction method, a training method and device of a capacity prediction model. BACKGROUND

[0002] At present, a user can determine the capacity of a battery through a discharge experiment.

[0003] However, the disassembly of the battery is required in the process of the discharge experiment, the steps are frequent and the workload is large, and the capacity of the battery cannot be quickly determined. SUMMARY

[0004] The present application provides a capacity prediction method, a training method and device of a capacity prediction model, and solves the technical problem that the related art requires the disassembly of the battery in the process of the discharge experiment, the steps are frequent and the workload is large, and the capacity of the battery cannot be quickly determined.

[0005] In a first aspect, the present application provides a capacity prediction method, comprising: obtaining the voltage of a to-be-identified battery at each time in a plurality of times, the voltage of M batteries corresponding to the to-be-identified battery at each time, and the voltage of N batteries corresponding to the to-be-identified battery at each time, the distance between the M batteries and the to-be-identified battery is less than or equal to a distance threshold, the voltage similarity between the N batteries and the to-be-identified battery is greater than or equal to a similarity threshold, M is an integer greater than or equal to 1, and N is an integer greater than or equal to 1; determining the aggregated feature of the to-be-identified battery at each time according to the voltage of the to-be-identified battery at each time, the voltage of the M batteries at each time, and the voltage of the N batteries at each time; generating the target feature of the to-be-identified battery at each time based on the aggregated feature of the to-be-identified battery at each time, the internal resistance feature of the to-be-identified battery at each time, and the initial feature of the to-be-identified battery; and inputting the target feature of the to-be-identified battery at each time into a target capacity prediction model to obtain the capacity of the to-be-identified battery at each time, the target capacity prediction model being trained based on the training method of the capacity prediction model provided by the present application.

[0006] Optionally, the determining the aggregated feature of the to-be-identified battery at each time according to the voltage of the to-be-identified battery at each time, the voltage of the M batteries at each time, and the voltage of the N batteries at each time comprises: determining a first feature and a second feature, the first feature being a sum of the voltage feature of the to-be-identified battery at a target time and the voltage feature of the M batteries at the target time, the second feature being a sum of the voltage feature of the to-be-identified battery at the target time and the voltage feature of the N batteries at the target time, the target time being one of the plurality of times, the voltage feature of a battery at a time being obtained by inputting the voltage of the battery at the time into a feature extraction network; splicing the first feature and the voltage feature of the to-be-identified battery at the target time to obtain a first spliced feature, and splicing the second feature and the voltage feature of the to-be-identified battery at the target time to obtain a second spliced feature; and splicing the first spliced feature and the second spliced feature to obtain the aggregated feature of the to-be-identified battery at the target time.

[0007] In the present application, the first spliced feature can represent the voltage of the to-be-identified battery at the target time from a spatial relationship, and the second spliced feature can represent the voltage of the to-be-identified battery at the target time from a semantic relationship. Therefore, the aggregated feature of the to-be-identified battery at the target time obtained by splicing the first spliced feature and the second spliced feature can represent the voltage of the to-be-identified battery at the target time in combination with the spatial relationship and the semantic relationship, which can accurately and effectively represent the voltage of the to-be-identified battery at the target time, thereby improving the accuracy of capacity prediction.

[0008] Optionally, the capacity prediction method further comprises: determining that a voltage similarity between a target battery and the to-be-identified battery satisfies the following formula:

[0009]

[0010] wherein S represents the voltage similarity between the target battery and the to-be-identified battery, A i represents the voltage of the target battery at the i th time, A i-1 represents the voltage of the target battery at the (i-1) th time, B i represents the voltage of the to-be-identified battery at the i th time, B i-1 represents the voltage of the to-be-identified battery at the (i-1) th time, the target battery being one of the N batteries, T representing the number of the plurality of times, and 2≤i≤T.

[0011] In the present application, the electronic device can determine the voltage similarity between the target battery and the to-be-identified battery based on the voltage of the target battery at each time point (i.e., the i th time point), the voltage of the target battery at the previous time point (i.e., the i-1 th time point) of each time point, the voltage of the to-be-identified battery at each time point, and the voltage of the to-be-identified battery at the previous time point, thereby facilitating and expediting the determination of the voltage similarity between the target battery and the to-be-identified battery, and further improving the efficiency of the electronic device in determining the aggregated feature of the to-be-identified battery at each time point.

[0012] Optionally, the capacity prediction method further includes: obtaining attribute information of the to-be-identified battery, the attribute information of the to-be-identified battery including at least one of an identifier of the to-be-identified battery, a manufacturer of the to-be-identified battery, a production batch of the to-be-identified battery, a use duration of the to-be-identified battery, and a discharge frequency of the to-be-identified battery; performing encoding processing on the attribute information of the to-be-identified battery to obtain encoding information corresponding to the to-be-identified battery; and obtaining an initial feature of the to-be-identified battery based on the encoding information of the to-be-identified battery and the embedded feature.

[0013] In the present application, for a certain battery (e.g., a to-be-identified battery), the attribute information of the to-be-identified battery can represent the original state (including the production state and the use state of the to-be-identified battery, etc.) of the to-be-identified battery. Therefore, the encoding information obtained based on the attribute information and the initial feature obtained based on the encoding information can accurately and effectively represent the feature of the to-be-identified battery in the production process and the use process, thereby improving the effectiveness of capacity prediction.

[0014] In a second aspect, the present application provides a training method of a capacity prediction model, including: obtaining a voltage of an identified battery at each time point in a plurality of time points, a voltage of X batteries corresponding to the identified battery at each time point, and a voltage of Y batteries corresponding to the identified battery at each time point, a distance between the X batteries and the identified battery being less than or equal to a distance threshold, a voltage similarity between the Y batteries and the identified battery being greater than or equal to a similarity threshold, X being an integer greater than or equal to 1, and Y being an integer greater than or equal to 1; determining an aggregated feature of the identified battery at each time point according to the voltage of the identified battery at each time point, the voltage of the X batteries at each time point, and the voltage of the Y batteries at each time point; generating a target feature of the identified battery at each time point based on the aggregated feature of the identified battery at each time point, an internal resistance feature of the identified battery at each time point, and an initial feature of the identified battery; and training an initial capacity prediction model based on the target feature of the identified battery at each time point to generate a target capacity prediction model.

[0015] In the present application, since the identified battery aggregation feature at each time point can represent the voltage condition of the identified battery at each time point in combination with spatial relationship and semantic relationship, and the target feature of the identified battery at each time point generated by the electronic device based on the aggregation feature of the battery to be identified at each time point, the internal resistance feature of the identified battery at each time point, and the initial feature of the identified battery can accurately and effectively represent the actual charging and discharging condition of the identified battery at each time point. Moreover, the electronic device trains the initial capacity prediction model based on the target feature of the identified battery at each time point, and can generate a capacity prediction model (i.e., a target capacity prediction model) with high prediction accuracy. Furthermore, the electronic device predicts the capacity of the battery based on the target capacity prediction model, which can improve the prediction efficiency of the battery capacity.

[0016] Optionally, the initial capacity prediction model is trained based on the target feature of the identified battery at each time point to generate a target capacity prediction model, specifically including: inputting the target feature of the identified battery at each time point into the initial capacity prediction model to obtain a predicted value of the identified battery at each time point; obtaining a true value of the identified battery at each time point; determining a target loss based on the predicted value of the identified battery at each time point and the true value of the identified battery at each time point; updating parameters in the initial capacity prediction model based on the target loss to generate the target capacity prediction model.

[0017] In the present application, the predicted value of the identified battery at each time point is the capacity of the identified battery at each time point predicted by the initial capacity prediction model, and the true value of the identified battery at each time point is the true capacity of the identified battery at each time point. In this way, the target loss determined by the electronic device based on the predicted value and the true value can accurately and effectively represent the difference between the result predicted by the initial capacity prediction model and the true result. Then, the electronic device updates the parameters in the initial capacity prediction model based on the target loss, which can conveniently and quickly generate a target capacity prediction model, thereby improving the efficiency of model training.

[0018] Optionally, the determining, according to the voltage of the identified battery at each time, the voltage of the X batteries at each time, and the voltage of the Y batteries at each time, the aggregated feature of the identified battery at each time, specifically includes: determining a third feature and a fourth feature, the third feature being a sum of the voltage feature of the identified battery at a target time and the voltage feature of the X batteries at the target time, the fourth feature being a sum of the voltage feature of the identified battery at the target time and the voltage feature of the Y batteries at the target time, the target time being one of the plurality of times, the voltage feature of a battery at a time being obtained by inputting the voltage of the battery at the time into the feature extraction network; splicing the third feature and the voltage feature of the identified battery at the target time to obtain a third spliced feature, and splicing the fourth feature and the voltage feature of the identified battery at the target time to obtain a fourth spliced feature; splicing the third spliced feature and the fourth spliced feature to obtain the aggregated feature of the identified battery at the target time.

[0019] In the present application, the third spliced feature can represent the voltage of the identified battery at the target time from a spatial relationship, and the fourth spliced feature can represent the voltage of the identified battery at the target time from a semantic relationship. Therefore, the aggregated feature of the identified battery at the target time obtained by splicing the third spliced feature and the fourth spliced feature can represent the voltage of the identified battery at the target time in combination with the spatial relationship and the semantic relationship, can accurately and effectively represent the voltage of the identified battery at the target time, and can further improve the accuracy of model training.

[0020] Optionally, the training method of the capacity prediction model further includes: determining that a voltage similarity between a preset battery and the identified battery satisfies the following formula:

[0021]

[0022] wherein S' represents the voltage similarity between the preset battery and the identified battery, C i represents the voltage of the preset battery at the i th time, C i-1 represents the voltage of the preset battery at the (i-1) th time, D i represents the voltage of the identified battery at the i th time, D i-1 represents the voltage of the identified battery at the (i-1) th time, the preset battery being one of the Y batteries, T representing the number of the plurality of times, and 2≤i≤T.

[0023] In the present application, the electronic device can determine the voltage similarity between the preset battery and the identified battery based on the voltage of the preset battery at each time (i.e., the i th time), the voltage of the preset battery at the previous time (i.e., the i-1 th time) of each time, the voltage of the identified battery at each time, and the voltage of the identified battery at the previous time, thereby facilitating and expediting the determination of the voltage similarity between the preset battery and the identified battery, and further improving the efficiency of the electronic device in determining the aggregated features of the identified battery at each time.

[0024] Optionally, the training method of the capacity prediction model further includes: obtaining attribute information of the identified battery, the attribute information of the identified battery including at least one of an identifier of the identified battery, a manufacturer of the identified battery, a production batch of the identified battery, a use duration of the identified battery, and a discharge frequency of the identified battery; performing encoding processing on the attribute information of the identified battery to obtain encoding information corresponding to the identified battery; and obtaining an initial feature of the identified battery based on the encoding information of the identified battery and the embedded feature.

[0025] In the present application, for a certain battery (e.g., an identified battery), the attribute information of the identified battery can represent the original state (including the production state and the use state of the identified battery, etc.) of the identified battery. Therefore, the encoding information obtained based on the attribute information and the initial feature obtained based on the encoding information can accurately and effectively represent the feature of the identified battery in the production process and the use process, thereby improving the effectiveness of model training.

[0026] In a third aspect, the present application provides a capacity prediction device, comprising: an acquisition module, a determination module and a processing module; the acquisition module is configured to acquire the voltage of a to-be-identified battery at each time point in a plurality of time points, the voltage of M batteries corresponding to the to-be-identified battery at each time point, and the voltage of N batteries corresponding to the to-be-identified battery at each time point, the distance between the M batteries and the to-be-identified battery is less than or equal to a distance threshold, the voltage similarity between the N batteries and the to-be-identified battery is greater than or equal to a similarity threshold, M is an integer greater than or equal to 1, and N is an integer greater than or equal to 1; the determination module is configured to determine the aggregated feature of the to-be-identified battery at each time point according to the voltage of the to-be-identified battery at each time point, the voltage of the M batteries at each time point, and the voltage of the N batteries at each time point; the processing module is configured to generate the target feature of the to-be-identified battery at each time point based on the aggregated feature of the to-be-identified battery at each time point, the internal resistance feature of the to-be-identified battery at each time point, and the initial feature of the to-be-identified battery; and the processing module is further configured to input the target feature of the to-be-identified battery at each time point into a target capacity prediction model to obtain the capacity of the to-be-identified battery at each time point, wherein the target capacity prediction model is obtained by training the capacity prediction model according to the training method provided by the present application.

[0027] Optionally, the determination module is specifically configured to determine a first feature and a second feature, the first feature being a sum of the voltage feature of the to-be-identified battery at a target time point and the voltage feature of the M batteries at the target time point, the second feature being a sum of the voltage feature of the to-be-identified battery at the target time point and the voltage feature of the N batteries at the target time point, the target time point being one of the plurality of time points, and the voltage feature of a battery at a time point being obtained by inputting the voltage of the battery at the time point into a feature extraction network; the processing module is specifically configured to splice the first feature and the voltage feature of the to-be-identified battery at the target time point to obtain a first spliced feature, and splice the second feature and the voltage feature of the to-be-identified battery at the target time point to obtain a second spliced feature; and the processing module is further configured to splice the first spliced feature and the second spliced feature to obtain the aggregated feature of the to-be-identified battery at the target time point.

[0028] Optionally, the determination module is further configured to determine that the voltage similarity between the target battery and the to-be-identified battery satisfies the following formula:

[0029]

[0030] wherein S represents the voltage similarity between the target battery and the to-be-identified battery, A i represents the voltage of the target battery at the i-th time point, A i-1 represents the voltage of the target battery at the (i-1)-th time point, B idenotes the voltage of the battery to be identified at the i-th time point, B i-1 denotes the voltage of the battery to be identified at the i-1-th time point, the target battery is one of the N batteries, T denotes the number of time points, and 2≤i≤T.

[0031] Optionally, the acquisition module is further configured to acquire attribute information of the battery to be identified, the attribute information of the battery to be identified including at least one of an identifier of the battery to be identified, a manufacturer of the battery to be identified, a production batch of the battery to be identified, a use duration of the battery to be identified, and a discharge frequency of the battery to be identified; the processing module is further configured to perform encoding processing on the attribute information of the battery to be identified to obtain encoding information corresponding to the battery to be identified; and the processing module is further configured to obtain an initial feature of the battery to be identified based on the encoding information of the battery to be identified and the embedded feature.

[0032] In a fourth aspect, the present application provides a training device for a capacity prediction model, including an acquisition module, a determination module, and a processing module. The acquisition module is configured to acquire a voltage of an identified battery at each time point in a plurality of time points, a voltage of X batteries corresponding to the identified battery at the each time point, and a voltage of Y batteries corresponding to the identified battery at the each time point, the distance between the X batteries and the identified battery being less than or equal to a distance threshold, the voltage similarity between the Y batteries and the identified battery being greater than or equal to a similarity threshold, X being an integer greater than or equal to 1, and Y being an integer greater than or equal to 1. The determination module is configured to determine an aggregated feature of the identified battery at the each time point according to the voltage of the identified battery at the each time point, the voltage of the X batteries at the each time point, and the voltage of the Y batteries at the each time point. The processing module is configured to generate a target feature of the identified battery at the each time point based on the aggregated feature of the identified battery at the each time point, an internal resistance feature of the identified battery at the each time point, and an initial feature of the identified battery. The processing module is further configured to train an initial capacity prediction model based on the target feature of the identified battery at the each time point, to generate a target capacity prediction model.

[0033] Optionally, the processing module is specifically configured to input the target feature of the identified battery at the each time point into the initial capacity prediction model to obtain a predicted value of the identified battery at the each time point. The acquisition module is further configured to acquire a true value of the identified battery at the each time point. The determination module is further configured to determine a target loss based on the predicted value of the identified battery at the each time point and the true value of the identified battery at the each time point. The processing module is further specifically configured to update parameters in the initial capacity prediction model based on the target loss to generate the target capacity prediction model.

[0034] Optionally, the determining module is specifically configured to determine a third feature and a fourth feature, the third feature being a sum of the voltage feature of the identified battery at a target time and voltage features of the X batteries at the target time, the fourth feature being a sum of the voltage feature of the identified battery at the target time and voltage features of the Y batteries at the target time, the target time being one of the plurality of times, the voltage feature of a battery at a time being obtained by inputting a voltage feature of the battery at the time into the feature extraction network; the processing module is further configured to splice the third feature and the voltage feature of the identified battery at the target time to obtain a third spliced feature, and splice the fourth feature and the voltage feature of the identified battery at the target time to obtain a fourth spliced feature; and the processing module is further configured to splice the third spliced feature and the fourth spliced feature to obtain an aggregated feature of the identified battery at the target time.

[0035] Optionally, the determining module is further configured to determine that a voltage similarity between a preset battery and the identified battery satisfies the following formula:

[0036]

[0037] wherein S' represents the voltage similarity between the preset battery and the identified battery, C i represents the voltage of the preset battery at the i th time, C i-1 represents the voltage of the preset battery at the (i-1) th time, D i represents the voltage of the identified battery at the i th time, D i-1 represents the voltage of the identified battery at the (i-1) th time, the preset battery being one of the Y batteries, T representing the number of the plurality of times, and 2≤i≤T.

[0038] Optionally, the obtaining module is further configured to obtain attribute information of the identified battery, the attribute information of the identified battery including at least one of an identifier of the identified battery, a manufacturer of the identified battery, a production batch of the identified battery, a use duration of the identified battery, and a discharge frequency of the identified battery; the processing module is further configured to perform encoding processing on the attribute information of the identified battery to obtain encoding information corresponding to the identified battery; and the processing module is further configured to obtain an initial feature of the identified battery based on the encoding information of the identified battery and the embedded feature.

[0039] In a fifth aspect, the present application provides an electronic device, comprising a processor and a memory configured to store processor-executable instructions; wherein the processor is configured to execute the instructions to implement any of the optional capacity prediction methods in the first aspect, or to implement any of the training methods of the optional capacity prediction model in the second aspect.

[0040] In a sixth aspect, the present application provides a computer-readable storage medium, and the computer-readable storage medium stores instructions. When the instructions in the computer-readable storage medium are executed by an electronic device, the electronic device can execute any of the optional capacity prediction methods in the first aspect, or execute any of the training methods of the optional capacity prediction model in the second aspect.

[0041] The capacity prediction method, the training method and the device of the capacity prediction model provided by the present application can be used to obtain the voltage of a to-be-identified battery at each time point, the voltage of M batteries corresponding to the to-be-identified battery at each time point, and the voltage of N batteries corresponding to the to-be-identified battery at each time point. Then, the electronic device can determine the aggregated feature of the to-be-identified battery at each time point according to the voltage of the to-be-identified battery at each time point, the voltage of the M batteries at each time point, and the voltage of the N batteries at each time point. The electronic device can generate the target feature of the to-be-identified battery at each time point based on the aggregated feature of the to-be-identified battery at each time point, the internal resistance feature of the to-be-identified battery at each time point, and the initial feature of the to-be-identified battery. Finally, the electronic device inputs the target feature of the to-be-identified battery at each time point into a target capacity prediction model to obtain the capacity of the to-be-identified battery at each time point. In the present application, the aggregated feature of the to-be-identified battery at each time point can represent the voltage condition of the to-be-identified battery at each time point in combination with the spatial relationship and the semantic relationship. In addition, the target feature of the to-be-identified battery at each time point generated based on the aggregated feature of the to-be-identified battery at each time point, the internal resistance feature of the to-be-identified battery at each time point, and the initial feature of the to-be-identified battery can accurately and effectively represent the actual charging and discharging condition of the to-be-identified battery at each time point. Moreover, the electronic device can accurately and effectively determine the capacity of the to-be-identified battery at each time point based on the target feature of the to-be-identified battery at each time point, thereby improving the prediction efficiency of the battery capacity. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced.

[0043] Figure 1 A flowchart of a capacity prediction method provided by an embodiment of the present application;

[0044] Figure 2 A flowchart of another capacity prediction method provided by an embodiment of the present application;

[0045] Figure 3 A schematic diagram of a spatial relationship graph and a semantic relationship graph provided by an embodiment of the present application;

[0046] Figure 4 A flowchart of another capacity prediction method provided by the embodiment of the application is shown in FIG. 6.

[0047] Figure 5 A flowchart of another capacity prediction method provided by the embodiment of the application is shown in FIG. 6.

[0048] Figure 6 A schematic diagram of encoding the attribute information of the battery provided by the embodiment of the application is shown in FIG. 7.

[0049] Figure 7 A flowchart of the training method of the capacity prediction model provided by the embodiment of the application is shown in FIG. 8.

[0050] Figure 8 A flowchart of another training method of the capacity prediction model provided by the embodiment of the application is shown in FIG. 9.

[0051] Figure 9 A flowchart of another training method of the capacity prediction model provided by the embodiment of the application is shown in FIG. 9.

[0052] Figure 10 A schematic diagram of the electronic device obtaining the voltage and capacity of the identified battery at a certain time provided by the embodiment of the application is shown in FIG. 10.

[0053] Figure 11 A flowchart of another training method of the capacity prediction model provided by the embodiment of the application is shown in FIG. 9.

[0054] Figure 12 A flowchart of another training method of the capacity prediction model provided by the embodiment of the application is shown in FIG. 9.

[0055] Figure 13 A structural schematic diagram of the capacity prediction device provided by the embodiment of the application is shown in FIG. 11.

[0056] Figure 14 A structural schematic diagram of another capacity prediction device provided by the embodiment of the application is shown in FIG. 12.

[0057] Figure 15 A structural schematic diagram of the training device of the capacity prediction model provided by the embodiment of the application is shown in FIG. 13.

[0058] Figure 16 A structural schematic diagram of another training device of the capacity prediction model provided by the embodiment of the application is shown in FIG. 14. DETAILED DESCRIPTION

[0059] The capacity prediction method, the training method and device of the capacity prediction model provided by the embodiment of the application will be described in detail below with reference to the accompanying drawings.

[0060] The terms "first" and "second" and the like in the description of the present application and in the claims of the present application are used for the purpose of differentiating different objects, and are not used for the purpose of describing a specific order of the objects, for example, the first feature and the second feature are used for the purpose of differentiating different features, and are not used for the purpose of describing a specific order of the features.

[0061] In addition, the terms "comprise" and "have" and any variations thereof mentioned in the description of the present application are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally further include other steps or units not listed, or can optionally further include other steps or units inherent to the process, method, product or device.

[0062] It should be noted that in the embodiments of the present application, the words "exemplary" or "for example" are used to mean serving as an example, instance, or illustration. Any embodiment or design presented as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or advantageous than other embodiments or design solutions. Rather, the use of "exemplary" or "for example" is intended to present concepts in a concrete manner.

[0063] In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0064] Based on the description in the background, since in the related art, the battery needs to be disassembled during the discharge experiment, the steps are frequent and the workload is large, and the capacity of the battery may not be quickly determined. Based on this, the embodiments of the present application provide a capacity prediction method, a training method and device of a capacity prediction model. Since the aggregation features of the to-be-identified battery at each time point can represent the voltage condition of the to-be-identified battery at each time point in combination with the spatial relationship and the semantic relationship, in addition, the electronic device can accurately and effectively represent the actual charging and discharging condition of the to-be-identified battery at each time point based on the target features of the to-be-identified battery at each time point generated based on the aggregation features of the to-be-identified battery at each time point, the internal resistance features of the to-be-identified battery at each time point, and the initial features of the to-be-identified battery. And the electronic device can accurately and effectively determine the capacity of the to-be-identified battery at each time point based on the target features of the to-be-identified battery at each time point, thereby improving the prediction efficiency of the battery capacity.

[0065] Exemplarily, the electronic device performing the capacity prediction method provided in the embodiments of the present application can be a mobile phone, a tablet computer, a desktop computer, a laptop computer, a handheld computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a cellular phone, a personal digital assistant (PDA), an augmented reality (AR) \ virtual reality (VR) device, and the like. The specific form of the electronic device is not specially limited in the present application. The electronic device can perform human-computer interaction with a user through one or more of a keyboard, a touchpad, a touch screen, a remote controller, voice interaction, a handwriting device, and the like.

[0066] Optionally, the electronic device can be a stand-alone physical server, a server cluster or a distributed system composed of multiple physical servers, a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

[0067] As shown in Figure 1 The capacity prediction method provided in the embodiments of the present application can include S101-S104.

[0068] S101, the electronic device obtains the voltage of the to-be-identified battery at each time point in a plurality of time points, the voltage of the M batteries corresponding to the to-be-identified battery at each time point, and the voltage of the N batteries corresponding to the to-be-identified battery at each time point.

[0069] The distance between the M batteries and the to-be-identified battery is less than or equal to a distance threshold, the voltage similarity between the N batteries and the to-be-identified battery is greater than or equal to a similarity threshold, M is an integer greater than or equal to 1, and N is an integer greater than or equal to 1.

[0070] The battery (for example, the to-be-identified battery) in the embodiments of the present application can be a storage battery capable of charging and discharging operations.

[0071] Exemplarily, the battery can be a lead-acid storage battery.

[0072] In an implementation form of the embodiment of the application, the electronic device can also acquire the position information of the to-be-identified battery and the position information of each battery in the plurality of batteries, and determine the distance between the to-be-identified battery and each battery based on the position information of the to-be-identified battery and the position information of each battery. Then the electronic device can determine the M batteries from the plurality of batteries based on the distance between the to-be-identified battery and each battery.

[0073] In another implementation form of the embodiment of the application, the electronic device can also determine the M batteries with the smallest distance between the to-be-identified battery from the plurality of batteries as the M batteries corresponding to the to-be-identified battery.

[0074] In an optional implementation form, the electronic device can also acquire the voltage of each battery in the plurality of batteries at each time point, and determine the voltage similarity between the to-be-identified battery and each battery based on the voltage of the to-be-identified battery at each time point and the voltage of each battery at each time point. Then the electronic device can determine the N batteries from the plurality of batteries based on the voltage similarity between the to-be-identified battery and each battery.

[0075] In another optional implementation form, the electronic device can also determine the N batteries with the largest voltage similarity between the to-be-identified battery from the plurality of batteries as the N batteries corresponding to the to-be-identified battery.

[0076] In an optional implementation form, the electronic device can input the voltage of the to-be-identified battery at each time point, the voltage of the M batteries at each time point, and the voltage of the N batteries at each time point into a feature extraction network to obtain the voltage feature of the to-be-identified battery at each time point, the voltage feature of the M batteries at each time point, and the voltage feature of the N batteries at each time point. Then the electronic device can determine the average value of the voltage feature of the to-be-identified battery at each time point, the voltage feature of the M batteries at each time point, and the voltage feature of the N batteries at each time point as the aggregated feature of the to-be-identified battery at each time point.

[0077] In an optional implementation form, the electronic device can input the voltage of the to-be-identified battery at each time point, the voltage of the M batteries at each time point, and the voltage of the N batteries at each time point into a feature extraction network to obtain the voltage feature of the to-be-identified battery at each time point, the voltage feature of the M batteries at each time point, and the voltage feature of the N batteries at each time point. Then the electronic device can determine the average value of the voltage feature of the to-be-identified battery at each time point, the voltage feature of the M batteries at each time point, and the voltage feature of the N batteries at each time point as the aggregated feature of the to-be-identified battery at each time point.

[0078] In another optional implementation form, the electronic device can also splice the voltage feature of the to-be-identified battery at each time point, the voltage feature of the M batteries at each time point, and the voltage feature of the N batteries at each time point to obtain the aggregated feature of the to-be-identified battery at each time point.

[0079] In combination withFigure 1 As shown in Figure 2 In an implementation manner of the embodiment of the present application, the electronic device determines the aggregated feature of the to-be-identified battery at each time according to the voltage of the to-be-identified battery at each time, the voltage of the M batteries at each time, and the voltage of the N batteries at each time, and specifically can include S1021-S1023.

[0080] S1021, the electronic device determines the first feature and the second feature.

[0081] The first feature is the sum of the voltage feature of the to-be-identified battery at the target time and the voltage feature of the M batteries at the target time, and the second feature is the sum of the voltage feature of the to-be-identified battery at the target time and the voltage feature of the N batteries at the target time. The target time is one of the plurality of times, and the voltage feature of a battery at a time is obtained by inputting the voltage feature of the battery at the time into the feature extraction network.

[0082] S1022, the electronic device splices the first feature and the voltage feature of the to-be-identified battery at the target time to obtain a first spliced feature, and splices the second feature and the voltage feature of the to-be-identified battery at the target time to obtain a second spliced feature.

[0083] In combination with the description of the above embodiment, it should be understood that, since the distance between the M batteries and the to-be-identified battery is less than or equal to the distance threshold, that is, the distance between the M batteries and the to-be-identified battery is small, the first spliced feature (or the first feature) can represent the voltage condition of the to-be-identified battery at a certain time (i.e., the target time) from the spatial relationship. Since the voltage similarity between the N batteries and the to-be-identified battery is greater than or equal to the similarity threshold, that is, the voltage similarity between the N batteries and the to-be-identified battery is large, the second spliced feature (or the second feature) can represent the voltage condition of the to-be-identified battery at the target time from the semantic relationship.

[0084] S1023, the electronic device splices the first spliced feature and the second spliced feature to obtain the aggregated feature of the to-be-identified battery at the target time.

[0085] It can be understood that, since the first spliced feature can represent the voltage condition of the to-be-identified battery at the target time from the spatial relationship, and the second spliced feature can represent the voltage condition of the to-be-identified battery at the target time from the semantic relationship. Therefore, the aggregated feature of the to-be-identified battery at the target time obtained by splicing the first spliced feature and the second spliced feature can represent the voltage condition of the to-be-identified battery at the target time in combination with the spatial relationship and the semantic relationship, which can accurately and effectively represent the voltage condition of the to-be-identified battery at the target time, and further can improve the accuracy of capacity prediction.

[0086] In one alternative implementation, the electronic device can determine the M batteries and N batteries mentioned above based on the spatial relationship graph and the semantic relationship graph, respectively, and then determine the aggregate features of the battery to be identified at a certain time (e.g., the target time).

[0087] Specifically, a spatial relationship graph is used to characterize the correlation between a battery and other batteries from a spatial (or distance) perspective. That is, an electronic device can construct this spatial relationship graph based on the distance between the battery to be identified and each of the multiple batteries mentioned above. A semantic relationship graph is used to characterize the correlation between a battery and other batteries from a semantic perspective. That is, an electronic device can construct this semantic relationship graph based on the voltage similarity between the battery to be identified and each of the batteries mentioned above.

[0088] It should be understood that when a battery experiences thermal runaway, its own temperature rises, and the heat is transferred to adjacent batteries through convection, conduction, and radiation, causing safety issues in the circuit containing that battery. The spatial relationship (or distance relationship) between a battery and other batteries can effectively characterize the temperature changes between them. Therefore, spatial relationship diagrams can effectively improve the efficiency and accuracy of battery capacity prediction.

[0089] Understandably, batteries may remain in a float charging state for extended periods. Due to changes in internal chemical substances, some batteries may experience a sudden voltage surge during discharge, causing them to reach their discharge cutoff voltage prematurely and reducing their discharge capacity. If a battery is spatially close to other batteries, its electrolyte concentration may differ significantly from the others. Therefore, constructing a semantic relationship graph for electronic devices can accurately characterize the voltage similarity between batteries, specifically representing the degradation patterns of similar batteries over time and accurately depicting the differential voltage similarity between them. Thus, semantic relationship graphs can also improve the efficiency and accuracy of battery capacity prediction.

[0090] The following example illustrates the spatial relationship graph and semantic relationship graph constructed by an electronic device.

[0091] For example, such as Figure 3 As shown, the spatial relationship diagram includes 16 batteries (i.e., 16 small boxes), and the semantic relationship diagram also includes 16 batteries.

[0092] Suppose that M is 4 and N is 3. The electronic device determines that the middle one of the batteries circled by the dashed line in the virtual space relationship diagram is the to-be-identified battery, and the remaining 4 batteries circled by the dashed line are the M batteries corresponding to the to-be-identified battery. The electronic device also determines that the lowermost battery circled by the dashed line in the semantic relationship diagram is the to-be-identified battery, and the remaining 3 batteries circled by the dashed line are the N batteries corresponding to the to-be-identified battery.

[0093] The electronic device can then obtain a combination relationship diagram of the to-be-identified battery, the remaining 4 batteries, and the remaining 3 batteries based on the target time. Figure 3 The electronic device can further determine the aggregated feature of the to-be-identified battery at the target time based on the voltage features of the 8 batteries included in the combination relationship diagram at the target time.

[0094] In S103, the electronic device generates a target feature of the to-be-identified battery at each time based on the aggregated feature of the to-be-identified battery at each time, the internal resistance feature of the to-be-identified battery at each time, and the initial feature of the to-be-identified battery.

[0095] In an optional implementation, the electronic device can splice the aggregated feature of the to-be-identified battery at each time, the internal resistance feature of the to-be-identified battery at each time, and the initial feature of the to-be-identified battery to obtain the target feature of the to-be-identified battery at each time.

[0096] In another optional implementation, the electronic device can also determine the average of the aggregated feature of the to-be-identified battery at each time, the internal resistance feature of the to-be-identified battery at each time, and the initial feature of the to-be-identified battery as the target feature of the to-be-identified battery at each time.

[0097] In the embodiments of the present application, the electronic device can obtain the internal resistance of the to-be-identified battery at each time in the plurality of times, and input the internal resistance of the to-be-identified battery at each time into the feature extraction network, so as to obtain the internal resistance feature of the to-be-identified battery at each time.

[0098] In S104, the electronic device inputs the target feature of the to-be-identified battery at each time into a target capacity prediction model to obtain the capacity of the to-be-identified battery at each time.

[0099] The target capacity prediction model is obtained based on the training method of the capacity prediction model provided in the embodiments of the present application.

[0100] In the embodiments of the present application, the target capacity prediction model is an accurate higher capacity prediction model that has been trained, and the target capacity prediction model is used to predict the capacity of a battery at a time.

[0101] The technical solutions provided by the above embodiments can bring at least the following beneficial effects: as can be seen from S101-S104, the electronic device can obtain the voltage of the to-be-identified battery at each of the plurality of time points, the voltage of the M batteries corresponding to the to-be-identified battery at each of the plurality of time points, and the voltage of the N batteries corresponding to the to-be-identified battery at each of the plurality of time points; then the electronic device can determine the aggregated feature of the to-be-identified battery at each of the plurality of time points according to the voltage of the to-be-identified battery at each of the plurality of time points, the voltage of the M batteries at each of the plurality of time points, and the voltage of the N batteries at each of the plurality of time points; and generate the target feature of the to-be-identified battery at each of the plurality of time points based on the aggregated feature of the to-be-identified battery at each of the plurality of time points, the internal resistance feature of the to-be-identified battery at each of the plurality of time points, and the initial feature of the to-be-identified battery; finally, the electronic device inputs the target feature of the to-be-identified battery at each of the plurality of time points into a target capacity prediction model to obtain the capacity of the to-be-identified battery at each of the plurality of time points. In the embodiments of the present application, since the aggregated feature of the to-be-identified battery at each of the plurality of time points can represent the voltage condition of the to-be-identified battery at each of the plurality of time points in combination with the spatial relationship and the semantic relationship, in addition, the target feature of the to-be-identified battery at each of the plurality of time points generated based on the aggregated feature of the to-be-identified battery at each of the plurality of time points, the internal resistance feature of the to-be-identified battery at each of the plurality of time points, and the initial feature of the to-be-identified battery can accurately and effectively represent the actual charging and discharging condition of the to-be-identified battery at each of the plurality of time points. Moreover, the electronic device can accurately and effectively determine the capacity of the to-be-identified battery at each of the plurality of time points based on the target feature of the to-be-identified battery at each of the plurality of time points, thereby improving the prediction efficiency of the battery capacity.

[0102] In combination Figure 1 As Figure 4 indicated, the capacity prediction method provided by the embodiments of the present application can further include S105.

[0103] 105. The electronic device determines that the voltage similarity between the target battery and the to-be-identified battery satisfies the following formula:

[0104]

[0105] wherein S represents the voltage similarity between the target battery and the to-be-identified battery, A i represents the voltage of the target battery at the i th time point, A i-1 represents the voltage of the target battery at the (i-1) th time point, B i represents the voltage of the to-be-identified battery at the i th time point, B i-1 represents the voltage of the to-be-identified battery at the (i-1) th time point, the target battery is one of the N batteries, and T represents the number of the plurality of time points, 2≤i≤T.

[0106] In this embodiment of the application, the electronic device can conveniently and quickly determine the voltage similarity between the target battery and the battery to be identified based on the voltage of the target battery at each time (i.e., the i-th time), the voltage of the target battery at the previous time (i.e., the (i-1)-th time), the voltage of the battery to be identified at each time, and the voltage of the battery to be identified at the previous time. This can improve the efficiency of the electronic device in determining the aggregate features of the battery to be identified at each time.

[0107] In one alternative implementation, for any battery (including the battery to be identified and each of the aforementioned batteries), the electronic device can generate a differential sequence for the battery based on its voltage at each of the aforementioned time points. This differential sequence can effectively characterize the fluctuations of the battery. Specifically, the differential sequence includes the difference between the voltage of the battery at each time point and the voltage of the battery at the previous time point. The electronic device can then determine the voltage similarity between the battery to be identified and each of the aforementioned batteries based on the differential sequence of the battery to be identified and the differential sequences of each of the aforementioned batteries, thus accurately characterizing the voltage similarity between the battery to be identified and each of the aforementioned batteries.

[0108] Combination Figure 1 ,like Figure 5 As shown, the capacity prediction method in this application embodiment further includes S106-108.

[0109] S106. The electronic device acquires the attribute information of the battery to be identified.

[0110] The attribute information of the battery to be identified includes at least one of the following: the battery's identifier, the battery's manufacturer, the battery's production batch, the battery's usage time, and the number of times the battery has been discharged.

[0111] Optionally, the attribute information of the battery to be identified may also include the type of battery and the model of the battery.

[0112] S107. The electronic device encodes the attribute information of the battery to be identified to obtain the coded information corresponding to the battery to be identified.

[0113] In this embodiment of the application, when the attribute information of the battery to be identified includes more than one item, the electronic device can encode at least two items of information separately (for example, each item of information can be assigned a different identifier, i.e., 0 or 1, etc.), and then arrange and combine the codes of the at least two items of information to generate the coded information corresponding to the battery to be identified.

[0114] Optionally, the above encoding process can be one-hot encoding.

[0115] S108. The electronic device obtains the initial characteristics of the battery to be identified based on the coded information and embedded features of the battery to be identified.

[0116] In this embodiment of the application, the electronic device can determine the initial characteristics of the battery to be identified, satisfying the following formula:

[0117] G = E × F

[0118] Wherein, G represents the initial features of the battery to be identified, E represents the coded information corresponding to the battery to be identified, and F represents the embedded features of the battery to be identified.

[0119] In one alternative implementation, the electronic device can input the encoded information corresponding to the battery to be identified into an embedding layer (which includes the aforementioned embedded features), and based on the embedding layer, the encoded information corresponding to the battery to be identified can be reduced in dimensionality to a denser feature (or vector).

[0120] For example, such as Figure 6 As shown, the electronic device can obtain attribute information for 16 batteries (i.e., 16 small boxes). The attribute information for each battery includes its identifier, manufacturer, production batch, usage time, and discharge cycles. Specifically, Figure 6 The document only shows the attribute information of three batteries (identified as 1, 2, and 3).

[0121] The electronic device then encodes the attribute information of the three batteries (battery 1, battery 2, and battery 3) to obtain the corresponding encoded information for each of the three batteries. Afterward, based on the encoded information and embedded features of each of the three batteries, the electronic device can obtain the initial characteristics of each of the three batteries.

[0122] In this embodiment, for a particular battery (e.g., a battery to be identified), the attribute information of the battery to be identified can characterize the original state of the battery (including its production state and usage state). Thus, the electronic device, based on the encoded information obtained from this attribute information and the initial features obtained from this encoded information, can accurately and effectively characterize the characteristics of the battery to be identified during the production and usage processes, thereby improving the effectiveness of capacity prediction.

[0123] like Figure 7 As shown, the training method for the capacity prediction model provided in this application embodiment may include S201-S204.

[0124] S201, the electronic device obtains the voltage of the identified battery at each of the plurality of time points, the voltage of the X batteries corresponding to the identified battery at each of the plurality of time points, and the voltage of the Y batteries corresponding to the identified battery at each of the plurality of time points.

[0125] wherein the distance between the X batteries and the identified battery is less than or equal to a distance threshold, the voltage similarity between the Y batteries and the identified battery is greater than or equal to a similarity threshold, X is an integer greater than or equal to 1, and Y is an integer greater than or equal to 1.

[0126] It should be noted that the number of the identified battery can be one or multiple. The number of the identified battery is not limited in the embodiments of the present application.

[0127] In addition, the specific number of the X batteries can be the same as or different from the specific number of the M batteries, and the specific number of the Y batteries can be the same as or different from the specific number of the N batteries. The specific number of the X batteries, the specific number of the Y batteries, the specific number of the M batteries, and the specific number of the N batteries are not limited in the embodiments of the present application.

[0128] It should be noted that the specific process of determining the X batteries corresponding to the identified battery and the Y batteries corresponding to the identified battery from the plurality of batteries by the electronic device is the same as or similar to the explanation of determining the M batteries corresponding to the to-be-identified battery and the N batteries corresponding to the to-be-identified battery from the plurality of batteries by the electronic device, which will not be described herein again.

[0129] S202, the electronic device determines the aggregated feature of the identified battery at each of the plurality of time points according to the voltage of the identified battery at each of the plurality of time points, the voltage of the X batteries at each of the plurality of time points, and the voltage of the Y batteries at each of the plurality of time points.

[0130] In an optional implementation, the electronic device can input the voltage of the identified battery at each of the plurality of time points, the voltage of the X batteries at each of the plurality of time points, and the voltage of the Y batteries at each of the plurality of time points into a feature extraction network to obtain the voltage feature of the identified battery at each of the plurality of time points, the voltage feature of the X batteries at each of the plurality of time points, and the voltage feature of the Y batteries at each of the plurality of time points. Then the electronic device can determine the average of the voltage feature of the identified battery at each of the plurality of time points, the voltage feature of the X batteries at each of the plurality of time points, and the voltage feature of the Y batteries at each of the plurality of time points as the aggregated feature of the identified battery at each of the plurality of time points.

[0131] In another optional implementation, the electronic device can also splice the voltage feature of the identified battery at each of the plurality of time points, the voltage feature of the X batteries at each of the plurality of time points, and the voltage feature of the Y batteries at each of the plurality of time points to obtain the aggregated feature of the identified battery at each of the plurality of time points.

[0132] In combination Figure 7 As Figure 8 shown, in an implementation of the embodiment of the present application, the electronic device determines the aggregated feature of the identified battery at each time according to the voltage of the identified battery at each time, the voltage of the X batteries at each time, and the voltage of the Y batteries at each time, which can specifically include S2021-S2023.

[0133] S2021, the electronic device determines a third feature and a fourth feature.

[0134] The third feature is the sum of the voltage feature of the identified battery at the target time and the voltage feature of the X batteries at the target time, and the fourth feature is the sum of the voltage feature of the identified battery at the target time and the voltage feature of the Y batteries at the target time. The target time is one of the plurality of times, and the voltage feature of a battery at a time is obtained by inputting the voltage of the battery at the time into the voltage feature extraction network.

[0135] S2022, the electronic device splices the third feature and the voltage feature of the identified battery at the target time to obtain a third spliced feature, and splices the fourth feature and the voltage feature of the identified battery at the target time to obtain a fourth spliced feature.

[0136] It can be understood that, since the distance between the X batteries and the identified battery is less than or equal to the distance threshold, i.e., the distance between the X batteries and the identified battery is small, the third spliced feature (or the third feature described above) can represent the voltage condition of the identified battery at a certain time (i.e., the target time) from the spatial relationship. Since the voltage similarity between the Y batteries and the identified battery is greater than or equal to the similarity threshold, i.e., the voltage similarity between the Y batteries and the identified battery is large, the fourth spliced feature (or the fourth feature described above) can represent the voltage condition of the identified battery at the target time from the semantic relationship.

[0137] S2023, the electronic device splices the third spliced feature and the fourth spliced feature to obtain the aggregated feature of the identified battery at the target time.

[0138] In the embodiments of the present application, the third splicing feature can represent the voltage condition of the identified battery at the target moment from a spatial relationship, and the fourth splicing feature can represent the voltage condition of the identified battery at the target moment from a semantic relationship. Therefore, the electronic device can represent the voltage condition of the identified battery at the target moment by splicing the third splicing feature and the fourth splicing feature to obtain the aggregated feature of the identified battery at the target moment, which can accurately and effectively represent the voltage condition of the identified battery at the target moment, thereby improving the accuracy of model training.

[0139] In the embodiments of the present application, the third splicing feature can represent the voltage condition of the identified battery at the target moment from a spatial relationship, and the fourth splicing feature can represent the voltage condition of the identified battery at the target moment from a semantic relationship. Therefore, the electronic device can represent the voltage condition of the identified battery at the target moment by splicing the third splicing feature and the fourth splicing feature to obtain the aggregated feature of the identified battery at the target moment, which can accurately and effectively represent the voltage condition of the identified battery at the target moment, thereby improving the accuracy of model training.

[0140] In the embodiments of the present application, the third splicing feature can represent the voltage condition of the identified battery at the target moment from a spatial relationship, and the fourth splicing feature can represent the voltage condition of the identified battery at the target moment from a semantic relationship. Therefore, the electronic device can represent the voltage condition of the identified battery at the target moment by splicing the third splicing feature and the fourth splicing feature to obtain the aggregated feature of the identified battery at the target moment, which can accurately and effectively represent the voltage condition of the identified battery at the target moment, thereby improving the accuracy of model training.

[0141] In the embodiments of the present application, the third splicing feature can represent the voltage condition of the identified battery at the target moment from a spatial relationship, and the fourth splicing feature can represent the voltage condition of the identified battery at the target moment from a semantic relationship. Therefore, the electronic device can represent the voltage condition of the identified battery at the target moment by splicing the third splicing feature and the fourth splicing feature to obtain the aggregated feature of the identified battery at the target moment, which can accurately and effectively represent the voltage condition of the identified battery at the target moment, thereby improving the accuracy of model training.

[0142] In the embodiments of the present application, the electronic device can obtain the internal resistance of the identified battery at each moment in the above plurality of moments, and input the internal resistance of the identified battery at each moment into the feature extraction network, so as to obtain the internal resistance feature of the identified battery at each moment.

[0143] In the embodiments of the present application, the third splicing feature can represent the voltage condition of the identified battery at the target moment from a spatial relationship, and the fourth splicing feature can represent the voltage condition of the identified battery at the target moment from a semantic relationship. Therefore, the electronic device can represent the voltage condition of the identified battery at the target moment by splicing the third splicing feature and the fourth splicing feature to obtain the aggregated feature of the identified battery at the target moment, which can accurately and effectively represent the voltage condition of the identified battery at the target moment, thereby improving the accuracy of model training.

[0144] It should be understood that the initial capacity prediction model is a capacity prediction model that is not trained or in an initial state, and the prediction accuracy (or precision) of the initial capacity prediction model may not be very high. The electronic device trains the initial capacity prediction model to generate an accurate and higher target capacity prediction model.

[0145] The technical solutions provided by the above embodiments can bring at least the following beneficial effects: as can be seen from S201-S204, the electronic device can obtain the voltage of the identified battery at each of the plurality of time points, the voltage of the X batteries corresponding to the identified battery at each of the plurality of time points, and the voltage of the Y batteries corresponding to the identified battery at each of the plurality of time points; then the electronic device can determine the aggregated feature of the identified battery at each of the plurality of time points according to the voltage of the identified battery at each of the plurality of time points, the voltage of the X batteries at each of the plurality of time points, and the voltage of the Y batteries at each of the plurality of time points; and generate the target feature of the identified battery at each of the plurality of time points based on the aggregated feature of the identified battery at each of the plurality of time points, the internal resistance feature of the identified battery at each of the plurality of time points, and the initial feature of the identified battery; finally, the electronic device can train the initial capacity prediction model based on the target feature of the identified battery at each of the plurality of time points to generate a target capacity prediction model. In the embodiments of the present application, since the aggregated feature of the identified battery at each of the plurality of time points can represent the voltage condition of the identified battery at each of the plurality of time points in combination with the spatial relationship and the semantic relationship, and the target feature of the identified battery at each of the plurality of time points generated by the electronic device based on the aggregated feature of the identified battery at each of the plurality of time points, the internal resistance feature of the identified battery at each of the plurality of time points, and the initial feature of the identified battery can accurately and effectively represent the actual charging and discharging condition of the identified battery at each of the plurality of time points. Moreover, the electronic device can train the initial capacity prediction model based on the target feature of the identified battery at each of the plurality of time points to generate a capacity prediction model (i.e., the target capacity prediction model) with high prediction accuracy. Furthermore, the electronic device can predict the capacity of the battery based on the target capacity prediction model, which can improve the prediction efficiency of the battery capacity.

[0146] In combination Figure 7 As Figure 9 indicated, in an implementation manner of the embodiments of the present application, the electronic device trains the initial capacity prediction model based on the target feature of the identified battery at each of the plurality of time points to generate a target capacity prediction model, which can specifically include S2041-S2044.

[0147] S2041, the electronic device inputs the target feature of the identified battery at each of the plurality of time points into the initial capacity prediction model to obtain a predicted value of the identified battery at each of the plurality of time points.

[0148] It should be understood that the predicted value of the identified battery at each of the plurality of time points is the capacity of the identified battery at each of the plurality of time points predicted by the initial capacity prediction model.

[0149] S2042, the electronic device obtains a true value of the identified battery at each of the plurality of time points.

[0150] It can be understood that the real value of the identified battery at each time point is the capacity (or real capacity) of the identified battery at each time point.

[0151] In the embodiment of the application, the electronic device can obtain the voltage of the identified battery at each time point during the float charging process of the identified battery, and can obtain the capacity of the identified battery at each time point during the core capacity discharge process of the identified battery.

[0152] Optionally, the electronic device can also obtain the internal resistance of the identified battery at each time point and the temperature of the identified battery at each time point during the float charging process. The electronic device can also obtain the current of the identified battery at each time point during the core capacity discharge process.

[0153] For example, as shown in FIG. 1, the electronic device can obtain the voltage of the identified battery at t1 during the float charging process of the identified battery, and obtain the capacity of the identified battery at t1 during the core capacity discharge process of the identified battery. Figure 10

[0154] S2043, the electronic device determines a target loss based on the predicted value of the identified battery at each time point and the real value of the identified battery at each time point.

[0155] The target loss is used to represent the degree of inconsistency between the predicted value of the identified battery at each time point and the real value of the identified battery at each time point.

[0156] Optionally, the electronic device can determine the above-mentioned target loss in combination with the cross-entropy function.

[0157] S2044, the electronic device updates the parameters in the initial capacity prediction model based on the target loss to generate a target capacity prediction model.

[0158] In one implementation manner of the embodiment of the application, the electronic device can iteratively update the parameters of the initial capacity prediction model based on the target loss, until the prediction accuracy (or precision) of the current capacity prediction model is greater than or equal to an accuracy threshold, at which time the electronic device can determine the current capacity prediction model as the target capacity prediction model.

[0159] ​In this embodiment, the predicted value of the identified battery at each time step is the capacity predicted by the initial capacity prediction model at that time step, and the true value of the identified battery at each time step is the true capacity of the identified battery at that time step. Thus, the target loss determined by the electronic device based on the predicted value and the true value can accurately and effectively characterize the difference between the result predicted by the initial capacity prediction model and the true result. Then, the electronic device updates the parameters in the initial capacity prediction model based on the target loss, enabling convenient and quick generation of the target capacity prediction model and improving the efficiency of model training.

[0160] Combination Figure 7 ,like Figure 11 As shown, the training method of the capacity prediction model in this application embodiment may further include S205.

[0161] S205. The electronic device determines that the voltage similarity between the preset battery and the identified battery satisfies the following formula:

[0162]

[0163] Where S' represents the voltage similarity between the preset battery and the identified battery, and C i C represents the voltage of the preset battery at time i. i-1 D represents the voltage of the preset battery at time i-1. i D represents the voltage of the identified battery at the i-th time. i-1 This indicates the voltage of the identified battery at the (i-1)th time. The preset battery is one of the Y batteries. T represents the number of times, and 2≤i≤T.

[0164] In this embodiment of the application, the electronic device can conveniently and quickly determine the voltage similarity between the preset battery and the identified battery based on the voltage of the preset battery at each time (i.e., the i-th time), the voltage of the preset battery at the previous time (i.e., the (i-1)-th time), the voltage of the identified battery at each time, and the voltage of the identified battery at the previous time. This can improve the efficiency of the electronic device in determining the aggregate characteristics of the identified battery at each time.

[0165] Combination Figure 7 ,like Figure 12 As shown, the training method of the capacity prediction model in this application embodiment further includes S206-S208.

[0166] S206. The electronic device acquires the attribute information of the identified battery.

[0167] The attribute information of the identified battery includes at least one of an identifier of the identified battery, a manufacturer of the identified battery, a production batch of the identified battery, a use duration of the identified battery, and a discharge frequency of the identified battery.

[0168] In S207, the electronic device encodes the attribute information of the identified battery to obtain encoding information corresponding to the identified battery.

[0169] It should be noted that the process of encoding the attribute information of the identified battery by the electronic device is the same as or similar to the process of encoding the attribute information of the battery to be identified by the electronic device, which will not be described here.

[0170] In S208, the electronic device obtains initial features of the identified battery based on the encoding information of the identified battery and the embedded features.

[0171] Optionally, the electronic device can determine the product of the encoding information of the identified battery and the embedded features as the initial features of the identified battery.

[0172] In the embodiments of the present application, for a certain battery (for example, the identified battery), the attribute information of the identified battery can represent the original state (including the production state and the use state of the identified battery) of the identified battery. Therefore, the encoding information obtained based on the attribute information and the initial features obtained based on the encoding information can accurately and effectively represent the feature conditions of the identified battery in the production process and the use process, thereby improving the effectiveness of model training.

[0173] The embodiments of the present application can divide the electronic device and the like into functional modules according to the above method examples. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one processing module. The integrated module can be realized in the form of hardware or in the form of a software functional module. It should be noted that the division of modules in the embodiments of the present application is illustrative, and is only a logical functional division. Actual implementation can have another division manner.

[0174] In the case of dividing each functional module corresponding to each function, Figure 13 A possible structural schematic diagram of the capacity prediction device involved in the above embodiments is shown, as shown in Figure 13 The capacity prediction device 10 can include an acquisition module 101, a determination module 102, and a processing module 103.

[0175] The acquisition module 101 is configured to acquire a voltage of a to-be-identified battery at each time point in a plurality of time points, voltages of M batteries corresponding to the to-be-identified battery at the each time point, and voltages of N batteries corresponding to the to-be-identified battery at the each time point, the M batteries being less than or equal to a distance threshold in distance from the to-be-identified battery, the N batteries being greater than or equal to a similarity threshold in voltage similarity from the to-be-identified battery, M being an integer greater than or equal to 1, and N being an integer greater than or equal to 1.

[0176] The determination module 102 is configured to determine an aggregated feature of the to-be-identified battery at the each time point according to the voltage of the to-be-identified battery at the each time point, the voltages of the M batteries at the each time point, and the voltages of the N batteries at the each time point.

[0177] The processing module 103 is configured to generate a target feature of the to-be-identified battery at the each time point based on the aggregated feature of the to-be-identified battery at the each time point, an internal resistance feature of the to-be-identified battery at the each time point, and an initial feature of the to-be-identified battery.

[0178] The processing module 103 is further configured to input the target feature of the to-be-identified battery at the each time point into a target capacity prediction model to obtain a capacity of the to-be-identified battery at the each time point, the target capacity prediction model being obtained by training a capacity prediction model according to an embodiment of the present application.

[0179] Optionally, the determination module 102 is specifically configured to determine a first feature and a second feature, the first feature being a sum of a voltage feature of the to-be-identified battery at a target time point and voltage features of the M batteries at the target time point, the second feature being a sum of the voltage feature of the to-be-identified battery at the target time point and voltage features of the N batteries at the target time point, the target time point being one of the plurality of time points, and the voltage feature of a battery at a time point being obtained by inputting a voltage of the battery at the time point into a feature extraction network.

[0180] The processing module 103 is specifically configured to splice the first feature and the voltage feature of the to-be-identified battery at the target time point to obtain a first spliced feature, and splice the second feature and the voltage feature of the to-be-identified battery at the target time point to obtain a second spliced feature.

[0181] The processing module 103 is further specifically configured to splice the first spliced feature and the second spliced feature to obtain the aggregated feature of the to-be-identified battery at the target time point.

[0182] Optionally, the determination module 102 is further configured to determine that a voltage similarity between a target battery and the to-be-identified battery satisfies the following formula:

[0183]

[0184] wherein S represents a voltage similarity between the target battery and the battery to be identified, A i represents a voltage of the target battery at the i th moment, A i-1 represents a voltage of the target battery at the (i-1) th moment, B i represents a voltage of the battery to be identified at the i th moment, B i-1 represents a voltage of the battery to be identified at the (i-1) th moment, the target battery is one of the N batteries, T represents a number of the plurality of moments, and 2≤i≤T.

[0185] Optionally, the obtaining module 101 is further configured to obtain attribute information of the battery to be identified, the attribute information of the battery to be identified including at least one of an identifier of the battery to be identified, a manufacturer of the battery to be identified, a production batch of the battery to be identified, a use duration of the battery to be identified, and a discharge frequency of the battery to be identified.

[0186] The processing module 103 is further configured to perform encoding processing on the attribute information of the battery to be identified to obtain encoding information corresponding to the battery to be identified.

[0187] The processing module 103 is further configured to obtain an initial feature of the battery to be identified based on the encoding information of the battery to be identified and the embedded feature.

[0188] In the case of using an integrated unit, Figure 14 A possible structural schematic diagram of the capacity prediction apparatus involved in the above embodiments is shown. As shown in the figure, Figure 14 The capacity prediction apparatus 20 can include a processing module 201 and a communication module 202. The processing module 201 can be configured to control and manage the actions of the capacity prediction apparatus 20. The communication module 202 can be configured to support the communication of the capacity prediction apparatus 20 with other entities. Optionally, as shown in the figure, Figure 14 The capacity prediction apparatus 20 can further include a storage module 203 configured to store program codes and data of the capacity prediction apparatus 20.

[0189] The processing module 201 can be a processor or a controller. The communication module 202 can be a transceiver, a transceiver circuit, or a communication interface, etc. The storage module 203 can be a memory.

[0190] When the processing module 201 is a processor, the communication module 202 is a transceiver, and the storage module 203 is a memory, the processor, the transceiver, and the memory can be connected through a bus. The bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0191] In the case of dividing each functional module according to each function, Figure 15 A possible structural schematic diagram of the training device of the capacity prediction model involved in the above embodiments is shown as follows. Figure 15 As shown in the figure, the training device 30 of the capacity prediction model can include an acquisition module 301, a determination module 302, and a processing module 303.

[0192] The acquisition module 301 is configured to acquire the voltage of an identified battery at each of a plurality of time points, the voltage of X batteries corresponding to the identified battery at each of the time points, and the voltage of Y batteries corresponding to the identified battery at each of the time points, the distance between the X batteries and the identified battery being less than or equal to a distance threshold, the voltage similarity between the Y batteries and the identified battery being greater than or equal to a similarity threshold, X being an integer greater than or equal to 1, and Y being an integer greater than or equal to 1.

[0193] The determination module 302 is configured to determine the aggregated feature of the identified battery at each of the time points according to the voltage of the identified battery at each of the time points, the voltage of the X batteries at each of the time points, and the voltage of the Y batteries at each of the time points.

[0194] The processing module 303 is configured to generate the target feature of the identified battery at each of the time points based on the aggregated feature of the identified battery at each of the time points, the internal resistance feature of the identified battery at each of the time points, and the initial feature of the identified battery.

[0195] The processing module 303 is further configured to train the initial capacity prediction model based on the target feature of the identified battery at each of the time points to generate a target capacity prediction model.

[0196] Optionally, the processing module 303 is specifically configured to input the target feature of the identified battery at each of the time points into the initial capacity prediction model to obtain a predicted value of the identified battery at each of the time points.

[0197] The acquisition module 301 is further configured to acquire the true value of the identified battery at each of the time points.

[0198] The determining module 302 is further configured to determine a target loss based on the predicted value of the identified battery at each time point and the actual value of the identified battery at each time point.

[0199] The processing module 303 is further configured to update the parameters in the initial capacity prediction model based on the target loss to generate the target capacity prediction model.

[0200] Optionally, the determining module 302 is specifically configured to determine a third feature and a fourth feature, the third feature being a sum of the voltage feature of the identified battery at a target time point and the voltage features of the X batteries at the target time point, the fourth feature being a sum of the voltage feature of the identified battery at the target time point and the voltage features of the Y batteries at the target time point, the target time point being one of the plurality of time points, and the voltage feature of a battery at a time point being a voltage input feature extraction network result of the voltage of the battery at the time point.

[0201] The processing module 303 is further configured to concatenate the third feature and the voltage feature of the identified battery at the target time point to obtain a third concatenated feature, and concatenate the fourth feature and the voltage feature of the identified battery at the target time point to obtain a fourth concatenated feature.

[0202] The processing module 303 is further configured to concatenate the third concatenated feature and the fourth concatenated feature to obtain an aggregated feature of the identified battery at the target time point.

[0203] Optionally, the determining module 302 is further configured to determine that a voltage similarity between a preset battery and the identified battery satisfies the following formula:

[0204]

[0205] wherein S' represents the voltage similarity between the preset battery and the identified battery, C i represents the voltage of the preset battery at the i th time point, C i-1 represents the voltage of the preset battery at the (i-1) th time point, D i represents the voltage of the identified battery at the i th time point, D i-1 represents the voltage of the identified battery at the (i-1) th time point, the preset battery being one of the Y batteries, T representing the number of the plurality of time points, and 2≤i≤T.

[0206] Optionally, the obtaining module 301 is further configured to obtain attribute information of the identified battery, the attribute information of the identified battery including at least one of an identifier of the identified battery, a manufacturer of the identified battery, a production batch of the identified battery, a use duration of the identified battery, and a discharge frequency of the identified battery.

[0207] The processing module 303 is further configured to encode the attribute information of the identified battery to obtain the encoding information corresponding to the identified battery.

[0208] The processing module 303 is further configured to obtain the initial feature of the identified battery based on the encoding information of the identified battery and the embedded feature.

[0209] In the case of using an integrated unit, Figure 16 A possible structural schematic diagram of the training apparatus of the capacity prediction model involved in the above embodiments is shown. As shown in Figure 16 The training apparatus 40 of the capacity prediction model can include a processing module 401 and a communication module 402. The processing module 401 can be configured to control and manage the actions of the training apparatus 40 of the capacity prediction model. The communication module 402 can be configured to support the communication between the training apparatus 40 of the capacity prediction model and other entities. Optionally, as shown in Figure 16 The training apparatus 40 of the capacity prediction model can further include a storage module 403 configured to store the program codes and data of the training apparatus 40 of the capacity prediction model.

[0210] The processing module 401 can be a processor or a controller. The communication module 402 can be a transceiver, a transceiving circuit, or a communication interface, etc. The storage module 403 can be a memory.

[0211] When the processing module 401 is a processor, the communication module 402 is a transceiver, and the storage module 403 is a memory, the processor, the transceiver, and the memory can be connected through a bus. The bus can be a PCI bus or an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0212] It should be understood that, in various embodiments of the present application, the size of the sequence number of the above processes does not mean the order of execution, and the execution order of the processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0213] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0214] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the system, device and unit described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0215] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0216] In the above embodiments, all or part can be realized by software, hardware, firmware or any combination thereof. When realized by software, all or part can be realized in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or data storage device including one or more servers, data centers, etc. integrated with one or more media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0217] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be limited by the protection scope of the claims.

Claims

1. A capacity prediction method characterized by, The method comprises: obtaining the voltage of the to-be-identified battery at each time in a plurality of times, the voltage of the M batteries corresponding to the to-be-identified battery at each time, and the voltage of the N batteries corresponding to the to-be-identified battery at each time, the distance between the M batteries and the to-be-identified battery being less than or equal to a distance threshold, the voltage similarity between the N batteries and the to-be-identified battery being greater than or equal to a similarity threshold, M being an integer greater than or equal to 1, and N being an integer greater than or equal to 1; determining the aggregated feature of the to-be-identified battery at each time according to the voltage of the to-be-identified battery at each time, the voltage of the M batteries at each time, and the voltage of the N batteries at each time; generating the target feature of the to-be-identified battery at each time based on the aggregated feature of the to-be-identified battery at each time, the internal resistance feature of the to-be-identified battery at each time, and the initial feature of the to-be-identified battery; inputting the target feature of the to-be-identified battery at each time into a target capacity prediction model to obtain the capacity of the to-be-identified battery at each time, the target capacity prediction model being trained based on a training method of a capacity prediction model, and the training method of the capacity prediction model comprising: obtaining the voltage of the identified battery at each time in a plurality of times, the voltage of the X batteries corresponding to the identified battery at each time, and the voltage of the Y batteries corresponding to the identified battery at each time, the distance between the X batteries and the identified battery being less than or equal to a distance threshold, the voltage similarity between the Y batteries and the identified battery being greater than or equal to a similarity threshold, X being an integer greater than or equal to 1, and Y being an integer greater than or equal to 1; determining the aggregated feature of the identified battery at each time according to the voltage of the identified battery at each time, the voltage of the X batteries at each time, and the voltage of the Y batteries at each time; generating the target feature of the identified battery at each time based on the aggregated feature of the identified battery at each time, the internal resistance feature of the identified battery at each time, and the initial feature of the identified battery; training an initial capacity prediction model based on the target feature of the identified battery at each time to generate a target capacity prediction model.

2. The capacity prediction method according to claim 1, characterized by, The determination of the aggregated feature of the to-be-identified battery at each time according to the voltage of the to-be-identified battery at each time, the voltage of the M batteries at each time, and the voltage of the N batteries at each time comprises: determining a first feature and a second feature, the first feature being a sum of a voltage feature of the to-be-identified battery at a target time and voltage features of the M batteries at the target time, the second feature being a sum of the voltage feature of the to-be-identified battery at the target time and voltage features of the N batteries at the target time, the target time being one of the plurality of times, the voltage feature of a battery at a time being obtained by inputting a voltage of the battery at the time into a voltage feature extraction network; concatenating the first feature and the voltage feature of the to-be-identified battery at the target time to obtain a first concatenated feature, and concatenating the second feature and the voltage feature of the to-be-identified battery at the target time to obtain a second concatenated feature; concatenating the first concatenated feature and the second concatenated feature to obtain an aggregated feature of the to-be-identified battery at the target time.

3. The capacity prediction method according to claim 1, characterized by, The method further comprises: determining that a voltage similarity between a target battery and the to-be-identified battery satisfies the following formula: in, This indicates the voltage similarity between the target battery and the battery to be identified. Indicates that the target battery is in the first Voltage at a given moment Indicates that the target battery is in the first Voltage at a given moment This indicates that the battery to be identified is in the first... Voltage at a given moment This indicates that the battery to be identified is in the first... The voltage at a given moment, wherein the target battery is one of the N batteries. This indicates the number of the multiple moments. .

4. The capacity prediction method according to any one of claims 1 to 3, characterized by, The method further comprises: obtaining attribute information of the to-be-identified battery, the attribute information of the to-be-identified battery including at least one of an identifier of the to-be-identified battery, a manufacturer of the to-be-identified battery, a production batch of the to-be-identified battery, a use duration of the to-be-identified battery, and a discharge frequency of the to-be-identified battery; encoding the attribute information of the to-be-identified battery to obtain encoding information corresponding to the to-be-identified battery; obtaining an initial feature of the to-be-identified battery based on the encoding information of the to-be-identified battery and the embedded feature.

5. A training method for a capacity prediction model, characterized in that, The method comprises: obtaining a voltage of an identified battery at each time of a plurality of times, a voltage of X batteries corresponding to the identified battery at the each time, and a voltage of Y batteries corresponding to the identified battery at the each time, a distance between the X batteries and the identified battery being less than or equal to a distance threshold, a voltage similarity between the Y batteries and the identified battery being greater than or equal to a similarity threshold, X being an integer greater than or equal to 1, and Y being an integer greater than or equal to 1; determining an aggregated feature of the identified battery at the each time according to the voltage of the identified battery at the each time, the voltage of the X batteries at the each time, and the voltage of the Y batteries at the each time; generating a target feature of the identified battery at the each time based on the aggregated feature of the identified battery at the each time, an internal resistance feature of the identified battery at the each time, and an initial feature of the identified battery; training an initial capacity prediction model based on the target feature of the identified battery at the each time to generate a target capacity prediction model. 6.The method of Claim 5, wherein, Training an initial capacity prediction model based on the target feature of the identified battery at the each time to generate a target capacity prediction model comprises: inputting the target feature of the identified battery at the each time into the initial capacity prediction model to obtain a predicted value of the identified battery at the each time; obtaining a true value of the identified battery at the each time; determine a target loss based on the predicted value of the identified battery at each time point and the true value of the identified battery at each time point; update parameters in the initial capacity prediction model based on the target loss to generate the target capacity prediction model. 7.The method of Claim 5, wherein, The determining the aggregated feature of the identified battery at each time point according to the voltage of the identified battery at each time point, the voltage of the X batteries at each time point, and the voltage of the Y batteries at each time point comprises: determining a third feature and a fourth feature, the third feature being a sum of the voltage feature of the identified battery at a target time point and the voltage feature of the X batteries at the target time point, and the fourth feature being a sum of the voltage feature of the identified battery at the target time point and the voltage feature of the Y batteries at the target time point, the target time point being one of the plurality of time points, and the voltage feature of a battery at a time point being obtained by inputting the voltage of the battery at the time point into a feature extraction network; splicing the third feature and the voltage feature of the identified battery at the target time point to obtain a third spliced feature, and splicing the fourth feature and the voltage feature of the identified battery at the target time point to obtain a fourth spliced feature; splicing the third spliced feature and the fourth spliced feature to obtain the aggregated feature of the identified battery at the target time point. 8.The method of Claim 5, wherein, The method further comprises: determining that a voltage similarity between a preset battery and the identified battery satisfies the following formula: in, This indicates the voltage similarity between the preset battery and the identified battery. Indicates that the preset battery is in the first Voltage at a given moment Indicates that the preset battery is in the first Voltage at a given moment This indicates that the identified battery is in the first... Voltage at a given moment This indicates that the identified battery is in the first... The voltage at a given moment, wherein the preset battery is one of the Y batteries. This indicates the number of the multiple moments. . 9.The method of Claim 5-8, wherein, The method further comprises: obtaining attribute information of the identified battery, the attribute information of the identified battery comprising at least one of an identifier of the identified battery, a manufacturer of the identified battery, a production batch of the identified battery, a use duration of the identified battery, and a discharge frequency of the identified battery; encoding the attribute information of the identified battery to obtain encoding information corresponding to the identified battery; obtaining an initial feature of the identified battery based on the encoding information of the identified battery and the embedded feature.

10. A capacity prediction device characterized by comprising: comprise: an obtaining module, a determining module, and a processing module; The obtaining module is configured to obtain a voltage of a to-be-identified battery at each time point in a plurality of time points, a voltage of M batteries corresponding to the to-be-identified battery at each time point, and a voltage of N batteries corresponding to the to-be-identified battery at each time point, a distance between the M batteries and the to-be-identified battery being less than or equal to a distance threshold, a voltage similarity between the N batteries and the to-be-identified battery being greater than or equal to a similarity threshold, M being an integer greater than or equal to 1, and N being an integer greater than or equal to 1; The determining module is configured to determine an aggregated feature of the to-be-identified battery at each time point according to the voltage of the to-be-identified battery at each time point, the voltage of the M batteries at each time point, and the voltage of the N batteries at each time point. The processing module is configured to generate target features of the to-be-identified battery at each time point based on the aggregate features of the to-be-identified battery at the each time point, the internal resistance features of the to-be-identified battery at the each time point, and initial features of the to-be-identified battery. The processing module is further configured to input the target features of the to-be-identified battery at the each time point into a target capacity prediction model to obtain capacities of the to-be-identified battery at the each time point, the target capacity prediction model being obtained based on a training method of a capacity prediction model, and the training method of the capacity prediction model comprising: obtaining voltages of an identified battery at each time point in a plurality of time points, voltages of X batteries corresponding to the identified battery at the each time point, and voltages of Y batteries corresponding to the identified battery at the each time point, the X batteries being less than or equal to a distance threshold in distance from the identified battery, the Y batteries being greater than or equal to a similarity threshold in voltage similarity from the identified battery, X being an integer greater than or equal to 1, and Y being an integer greater than or equal to 1; determining aggregate features of the identified battery at the each time point according to the voltages of the identified battery at the each time point, the voltages of the X batteries at the each time point, and the voltages of the Y batteries at the each time point; generating target features of the identified battery at the each time point based on the aggregate features of the identified battery at the each time point, internal resistance features of the identified battery at the each time point, and initial features of the identified battery; training an initial capacity prediction model based on the target features of the identified battery at the each time point to generate a target capacity prediction model.

11. The capacity prediction apparatus according to claim 10, wherein the determining module is specifically configured to determine a first feature and a second feature, the first feature being a sum of a voltage feature of the to-be-identified battery at a target time point and voltage features of the M batteries at the target time point, and the second feature being a sum of the voltage feature of the to-be-identified battery at the target time point and voltage features of the N batteries at the target time point, the target time point being one of the plurality of time points, and a voltage feature of a battery at a time point being obtained by inputting a voltage of the battery at the time point into a feature extraction network; the processing module is specifically configured to splice the first feature and the voltage feature of the to-be-identified battery at the target time point to obtain a first spliced feature, and splice the second feature and the voltage feature of the to-be-identified battery at the target time point to obtain a second spliced feature; the processing module is further specifically configured to splice the first spliced feature and the second spliced feature to obtain the aggregate features of the to-be-identified battery at the target time point.

12. The capacity prediction apparatus according to claim 10, wherein the determining module is further configured to determine that a voltage similarity between a target battery and the to-be-identified battery satisfies the following formula: in, This indicates the voltage similarity between the target battery and the battery to be identified. Indicates that the target battery is in the first Voltage at a given moment Indicates that the target battery is in the first Voltage at a given moment This indicates that the battery to be identified is in the first... Voltage at a given moment This indicates that the battery to be identified is in the first... The voltage at a given moment, wherein the target battery is one of the N batteries. This indicates the number of the multiple moments. .

13. The capacity prediction apparatus according to any one of claims 10-12, wherein The acquisition module is further configured to acquire attribute information of the to-be-identified battery, the attribute information of the to-be-identified battery including at least one of an identifier of the to-be-identified battery, a manufacturer of the to-be-identified battery, a production batch of the to-be-identified battery, a use duration of the to-be-identified battery, and a discharge frequency of the to-be-identified battery. The processing module is further configured to perform encoding processing on the attribute information of the to-be-identified battery to obtain encoding information corresponding to the to-be-identified battery. The processing module is further configured to obtain an initial feature of the to-be-identified battery based on the encoding information of the to-be-identified battery and the embedded feature. 14.A device for training a capacity prediction model, comprising: Comprise: An acquisition module, a determination module, and a processing module; The acquisition module is configured to acquire, at each of a plurality of time points, a voltage of an identified battery, voltages of X batteries corresponding to the identified battery at the time point, and voltages of Y batteries corresponding to the identified battery at the time point, the X batteries being within a distance threshold from the identified battery, the Y batteries having a voltage similarity greater than or equal to a similarity threshold with the identified battery, X being an integer greater than or equal to 1, and Y being an integer greater than or equal to 1. The determination module is configured to determine, at each of the time points, an aggregated feature of the identified battery based on the voltage of the identified battery at the time point, the voltages of the X batteries at the time point, and the voltages of the Y batteries at the time point. The processing module is configured to generate, at each of the time points, a target feature of the identified battery based on the aggregated feature of the identified battery at the time point, an internal resistance feature of the identified battery at the time point, and the initial feature of the identified battery. The processing module is further configured to train an initial capacity prediction model based on the target feature of the identified battery at each of the time points, to generate a target capacity prediction model.

15. The training apparatus of the capacity prediction model according to claim 14, wherein The processing module is specifically configured to input the target feature of the identified battery at each of the time points into the initial capacity prediction model to obtain a predicted value of the identified battery at each of the time points. The acquisition module is further configured to acquire a true value of the identified battery at each of the time points. The determination module is further configured to determine a target loss based on the predicted value of the identified battery at each of the time points and the true value of the identified battery at each of the time points. The processing module is further specifically configured to update parameters in the initial capacity prediction model based on the target loss to generate the target capacity prediction model.

16. The training apparatus of the capacity prediction model according to claim 14, wherein The determining module is specifically configured to determine a third feature and a fourth feature, the third feature being a sum of the voltage feature of the identified battery at a target moment and voltage features of the X batteries at the target moment, the fourth feature being a sum of the voltage feature of the identified battery at the target moment and voltage features of the Y batteries at the target moment, the target moment being one of the plurality of moments, and the voltage feature of a battery at a moment being obtained by inputting a voltage of the battery at the moment into a feature extraction network. The processing module is further configured to splice the third feature and the voltage feature of the identified battery at the target moment to obtain a third spliced feature, and splice the fourth feature and the voltage feature of the identified battery at the target moment to obtain a fourth spliced feature. The processing module is further configured to splice the third spliced feature and the fourth spliced feature to obtain the aggregated feature of the identified battery at the target moment.

17. The device of claim 14, wherein the determining module is further configured to determine that a voltage similarity between a preset battery and the identified battery satisfies the following formula:

18. The device of any one of claims 14-17, wherein the obtaining module is further configured to obtain attribute information of the identified battery, the attribute information of the identified battery including at least one of an identifier of the identified battery, a manufacturer of the identified battery, a production batch of the identified battery, a use duration of the identified battery, and a discharge frequency of the identified battery. in, This indicates the voltage similarity between the preset battery and the identified battery. Indicates that the preset battery is in the first Voltage at a given moment Indicates that the preset battery is in the first Voltage at a given moment This indicates that the identified battery is in the first... Voltage at a given moment This indicates that the identified battery is in the first... The voltage at a given moment, wherein the preset battery is one of the Y batteries. This indicates the number of the multiple moments. . The processing module is further configured to perform encoding processing on the attribute information of the identified battery to obtain encoding information corresponding to the identified battery. The processing module is further configured to obtain an initial feature of the identified battery based on the encoding information of the identified battery and the embedded feature. The electronic device includes: a processor; 19. An electronic device, comprising: a memory configured to store instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the capacity prediction method of any one of claims 1-4, or the training method of the capacity prediction model of any one of claims 5-9. When the instructions in the computer-readable storage medium are executed by the electronic device, the electronic device is enabled to perform the capacity prediction method of any one of claims 1-4, or the training method of the capacity prediction model of any one of claims 5-9. ​ 20. A computer-readable storage medium having stored thereon instructions, the instructions comprising, ​

Images