Battery capacity prediction model training method, battery capacity prediction method and related equipment

Abstract

The embodiment of the invention provides a battery capacity prediction model training method, a battery capacity prediction method and related equipment thereof, and the battery capacity prediction model training method comprises the steps: obtaining a battery capacity label, and obtaining working condition data of a battery working condition and a vehicle working condition corresponding to the battery capacity label, the working condition data further comprises the whole vehicle battery consistency evaluation data and / or the vehicle type, the battery capacity prediction model is trained based on the working condition data, and the battery capacity prediction capacity and prediction precision of the model under various conditions are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of batteries, and in particular to a battery capacity prediction model training, a battery capacity prediction method and related equipment. Background Art

[0002] With the continuous development of new energy, electric vehicles have shown great market prospects with their advantages of high environmental protection, low noise and low cost, and can effectively promote energy conservation and emission reduction, and promote the sustainable development and progress of society. In the development of electric vehicles, battery capacity has an important impact. Based on the importance of battery capacity, in order to improve the reliability of electric vehicles, some technical solutions for battery capacity prediction have emerged.

[0003] However, these technical solutions cannot predict the battery capacity fluctuation of some electric vehicles very well, and there is a problem of large prediction deviation. Therefore, how to improve the accuracy of battery capacity prediction is a technical problem to be solved. Summary of the invention

[0004] The embodiments of the present application provide a battery capacity prediction model training method, a battery capacity prediction method and related equipment, which can improve the accuracy of battery capacity prediction.

[0005] In a first aspect, a battery capacity prediction model training method is provided, comprising: Get the battery capacity label; Acquire the operating condition data of the battery operating condition and the vehicle operating condition corresponding to the battery capacity label, wherein the operating condition data includes: operating condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle cumulative mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model; Based on the operating condition data, a battery capacity prediction model is trained.

[0006] The technical solution of the embodiment of the present application, after obtaining the battery capacity label, obtains operating condition data of multiple dimensions including battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes vehicle battery consistency evaluation data and / or vehicle model, and the vehicle battery consistency or vehicle model has a great influence on the battery capacity. For example, if the vehicle battery consistency is relatively good, the charging and discharging of each battery of the vehicle is relatively balanced, and the overall battery capacity of the vehicle is high. If the vehicle battery consistency is relatively poor, the charging and discharging of each battery of the vehicle is relatively unbalanced, resulting in a low overall battery capacity of the vehicle. Different models have different algorithms and strategies for battery management, and the differences in algorithms and strategies will also have a great impact on the battery capacity. Therefore, based on the operating condition data obtained in the embodiment of the present application, the battery capacity prediction model is trained, so that the data used for training can cover various operating conditions of the vehicle, and the trained battery capacity prediction model is obtained, which effectively improves the model's capacity prediction ability and prediction accuracy for the vehicle under various conditions.

[0007] In a possible implementation, the training of the battery capacity prediction model based on the operating condition data includes: Extracting a target operating condition index sample corresponding to the battery capacity tag from the operating condition data; By performing correlation analysis on the battery capacity labels and the corresponding target operating condition indicator samples, target operating condition indicator samples that meet the correlation conditions are screened out to obtain a training data set; Using the training data set, an initial battery capacity prediction model is trained, and an error influencing factor of the initial battery capacity prediction model is analyzed, wherein the battery capacity prediction model is used to predict the battery capacity; Based on the error influencing factor, performing sample denoising on the training data set to obtain an updated training data set; Based on the updated training data set, the initial battery capacity prediction model is trained to obtain a trained battery capacity prediction model.

[0008] The technical solution of this implementation method extracts target operating condition indicator samples corresponding to battery capacity labels from operating condition data, so that the samples can cover various operating conditions of the vehicle, and performs correlation analysis on the target operating condition indicator samples to screen out target operating condition indicator samples that meet the correlation conditions, thereby achieving sample dimensionality reduction. The initial battery capacity prediction model is then trained through the training data set to analyze the error influence shadow, so that the training data set can be subjected to sample denoising based on the error influence factor to obtain an updated training data set, and the initial battery capacity prediction model is trained to obtain a trained battery capacity prediction model, thereby effectively improving the model's capacity prediction capability and prediction accuracy for vehicles under various conditions.

[0009] In a possible implementation, the obtaining of the battery capacity tag includes: obtaining a battery capacity tag that satisfies a tag selection condition, where the tag selection condition includes at least one of the following conditions: The battery corresponding to the battery capacity label should be discharged to the preset low-end capacity threshold before charging; The battery corresponding to the battery capacity label should reach the preset full-charge threshold before charging ends; After the battery corresponding to the battery capacity label is charged, the battery rest time should reach the preset rest time threshold; The battery corresponding to the battery capacity label does not experience any preset impact event during the charging process; The distribution ratio of battery capacity labels meets the distribution ratio condition.

[0010] Through the technical solution of this implementation method, it is possible to effectively select battery capacity labels whose corresponding data are pure, accurate, certain, with little interference and reasonable distribution, thereby improving the prediction accuracy of the model.

[0011] In a possible implementation, the step of obtaining a battery capacity tag that meets a tag selection condition includes: Obtain the battery capacity label of the corresponding battery before charging, when the discharge reaches the preset capacity low-end threshold, and before the end of charging, the battery reaches the preset capacity full-charge threshold, and after the end of charging, the battery rest time reaches the preset rest time threshold, and no preset influencing event occurs during the charging process of the battery, to obtain an initial battery capacity label set; Counting the distribution ratio of the initial battery capacity tag set in multiple capacity intervals; According to the distribution ratio, the battery capacity labels that meet the distribution ratio condition are screened out from the initial battery capacity labels.

[0012] Through the technical solution of this implementation method, reliable battery capacity labels are first screened out, and then screened through distribution percentage statistics, which can improve the credibility and distribution rationality of samples in each capacity interval and avoid too many or too few interval samples that do not meet the prediction requirements.

[0013] In a possible implementation, the performing of correlation analysis on the battery capacity tag and the corresponding target operating condition indicator sample to screen out the target operating condition indicator sample that meets the correlation condition includes: Performing a first correlation analysis between the battery capacity tag and the corresponding target operating condition index sample to obtain a first correlation analysis result between the battery capacity tag and the corresponding target operating condition index sample; Based on the first correlation analysis result, target operating condition indicator samples that meet the first correlation condition are screened out for the battery capacity label.

[0014] Through the technical solution of this implementation method, based on the correlation between the battery capacity label and the target operating condition indicator sample, it is possible to effectively exclude the operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity label, thereby improving the model training efficiency and model prediction accuracy.

[0015] In a possible implementation, the performing of correlation analysis on the battery capacity tag and the corresponding target operating condition indicator sample to screen out the target operating condition indicator sample that meets the correlation condition includes: Performing a second correlation analysis between a plurality of target operating condition indicator samples corresponding to any of the battery capacity tags to obtain a second correlation analysis result between the plurality of target operating condition indicator samples; Based on the second correlation analysis result, a target operating condition indicator sample that meets the second correlation condition is screened out for any of the battery capacity tags.

[0016] Through the technical solution of this implementation method, it is possible to effectively eliminate redundant samples based on the correlation between multiple target operating condition indicator samples, reduce data complexity, and improve model training efficiency and model prediction accuracy.

[0017] In a possible implementation, the correlation condition includes: The correlation between any of the battery capacity tags and the corresponding target operating condition indicator samples is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples corresponding to any of the battery capacity tags is less than or equal to a second preset correlation threshold.

[0018] The technical solution of this implementation method makes the correlation between the screened battery capacity label and the corresponding target operating condition indicator sample greater than or equal to the first preset correlation threshold, and the correlation between multiple target operating condition indicator samples is less than or equal to the second preset correlation threshold, thereby effectively excluding operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity label, and eliminating redundant samples from multiple strongly correlated operating condition indicator samples, reducing data complexity, and improving model training efficiency and model prediction accuracy.

[0019] In a possible implementation, after screening out target operating condition indicator samples that meet the correlation condition, the method further includes: Calculating the covariance matrix corresponding to the set of target operating condition index samples; Determining, based on the covariance matrix, a feature contribution corresponding to each target operating condition index sample in the set of target operating condition index samples; Based on the characteristic contribution corresponding to each target operating condition index sample, the target operating condition index samples whose characteristic contribution does not meet the contribution condition are removed from the set of target operating condition index samples to obtain an updated target operating condition index sample set; The training data set is obtained based on the updated target operating condition index sample set.

[0020] The technical solution of this implementation method calculates the covariance matrix of the set of target operating condition index samples, decomposes the feature contribution, screens the target operating condition index samples that meet the contribution conditions based on the feature contribution, and eliminates the target operating condition index samples that do not meet the conditions, thereby improving the sample quality of the target operating condition index sample set, achieving sample dimensionality reduction, and effectively improving the model training efficiency and training effect.

[0021] In a possible implementation, the training of the initial battery capacity prediction model using the training data set and the analysis of the error influencing factors of the initial battery capacity prediction model include: Inputting the target operating condition index samples in the training data set into an initial battery capacity prediction model to obtain a capacity prediction result output by the initial battery capacity prediction model; Based on the battery capacity label corresponding to the target operating condition index sample and the capacity prediction result, calculating the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation of the initial battery capacity prediction model; Based on the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation, the error influencing factor of the initial battery capacity prediction model is analyzed.

[0022] The technical solution of this implementation method analyzes the distribution of the mean square error, mean absolute error, determination coefficient and each prediction deviation to analyze the error influencing factor, and performs sample denoising on the training data set based on the error influencing factor, thereby reducing the error caused by abnormal samples, improving the model training effect, and obtaining a more optimized battery capacity prediction model.

[0023] In a second aspect, a battery capacity prediction method is provided, comprising: receiving a battery capacity prediction request, determining an operating condition index of a battery to be tested; inputting the operating condition index of the battery to be tested into a battery capacity prediction model, and obtaining a capacity prediction result output by the battery capacity prediction model; Among them, the battery capacity prediction model is pre-trained based on operating condition data, and the operating condition data includes: operating condition data corresponding to multiple dimensions in battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model.

[0024] Through the technical solution of this implementation method, the battery capacity prediction model can effectively avoid the interference of error factors, and the operating condition data of the pre-trained model also includes the vehicle battery consistency evaluation data and / or vehicle model, and the vehicle battery consistency or vehicle model has a great influence on the battery capacity. For example, if the vehicle battery consistency is relatively good, the charging and discharging of each battery in the vehicle is relatively balanced, and the overall battery capacity of the vehicle is high. If the vehicle battery consistency is relatively poor, the charging and discharging of each battery in the vehicle is relatively unbalanced, resulting in a low overall battery capacity of the vehicle. Different models have different algorithms and strategies for battery management, and the differences in algorithms and strategies will also have a great impact on the battery capacity. Therefore, the data used for training can cover various operating conditions of the vehicle, learn the mapping relationship between different battery capacities and different operating condition data, and then accurately predict the battery capacity based on the operating condition indicators in the application, effectively improving the capacity prediction capability and prediction accuracy of the vehicle under various conditions.

[0025] In a third aspect, a battery capacity prediction model training device is provided, comprising: A battery capacity label acquisition module is used to acquire the battery capacity label; A working condition data acquisition module, used to acquire working condition data of the battery working condition and the vehicle working condition corresponding to the battery capacity label, wherein the working condition data includes: working condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle cumulative mileage, and the working condition data also includes: vehicle battery consistency assessment data and / or vehicle model; The model training module is used to train the battery capacity prediction model based on the operating condition data.

[0026] The technical solution of the embodiment of the present application, after obtaining the battery capacity label, obtains operating condition data of multiple dimensions including battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes vehicle battery consistency evaluation data and / or vehicle model, and the vehicle battery consistency or vehicle model has a great influence on the battery capacity. For example, if the vehicle battery consistency is relatively good, the charging and discharging of each battery of the vehicle is relatively balanced, and the overall battery capacity of the vehicle is high. If the vehicle battery consistency is relatively poor, the charging and discharging of each battery of the vehicle is relatively unbalanced, resulting in a low overall battery capacity of the vehicle. Different models have different algorithms and strategies for battery management, and the differences in algorithms and strategies will also have a great impact on the battery capacity. Therefore, based on the operating condition data, the battery capacity prediction model is trained, so that the data used for training can cover various operating conditions of the vehicle, and the trained battery capacity prediction model is obtained, which effectively improves the model's capacity prediction ability and prediction accuracy for the vehicle under various conditions.

[0027] Optionally, the model training module includes: An index sample extraction submodule, used to extract a target operating condition index sample corresponding to the battery capacity tag from the operating condition data; A training data set acquisition submodule is used to perform correlation analysis on the battery capacity label and the corresponding target operating condition indicator sample, filter out the target operating condition indicator sample that meets the correlation condition, and obtain the training data set; An error analysis submodule, used to train an initial battery capacity prediction model using the training data set and analyze error influencing factors of the initial battery capacity prediction model, wherein the battery capacity prediction model is used to predict battery capacity; A sample denoising submodule, configured to perform sample denoising on the training data set based on the error influencing factor to obtain an updated training data set; The training submodule is used to train the initial battery capacity prediction model based on the updated training data set to obtain a trained battery capacity prediction model.

[0028] The technical solution of this implementation method extracts target operating condition indicator samples corresponding to battery capacity labels from operating condition data, so that the samples can cover various operating conditions of the vehicle, and performs correlation analysis on the target operating condition indicator samples to screen out target operating condition indicator samples that meet the correlation conditions, thereby achieving sample dimensionality reduction. The initial battery capacity prediction model is then trained through the training data set to analyze the error influence shadow, so that the training data set can be subjected to sample denoising based on the error influence factor to obtain an updated training data set, and the initial battery capacity prediction model is trained to obtain a trained battery capacity prediction model, thereby effectively improving the model's capacity prediction capability and prediction accuracy for vehicles under various conditions.

[0029] In a possible implementation, the battery capacity tag acquisition module is used to acquire a battery capacity tag that meets a tag selection condition, where the tag selection condition includes at least one of the following conditions: The battery corresponding to the battery capacity label should be discharged to the preset low-end capacity threshold before charging; The battery corresponding to the battery capacity label should reach the preset full-charge threshold before charging ends; After the battery corresponding to the battery capacity label is charged, the battery rest time should reach the preset rest time threshold; The battery corresponding to the battery capacity label does not experience any preset impact event during the charging process; The distribution ratio of battery capacity labels meets the distribution ratio condition.

[0030] Through the technical solution of this implementation method, it is possible to effectively select battery capacity labels whose corresponding data are pure, accurate, certain, with little interference and reasonable distribution, thereby improving the prediction accuracy of the model.

[0031] In a possible implementation, the battery capacity tag acquisition module includes: The initial label set acquisition submodule is used to acquire the battery capacity label of the corresponding battery before charging, when the discharge reaches the preset capacity low-end threshold, and before the end of charging, the battery reaches the preset capacity full-charge threshold, and after the end of charging, the battery rest time reaches the preset rest time threshold, and no preset influencing event occurs during the charging process, so as to obtain the initial battery capacity label set; A proportion statistics submodule, used to count the distribution proportions of the initial battery capacity tag set in multiple capacity intervals; The proportion screening submodule is used to screen out the battery capacity labels that meet the distribution proportion conditions from the initial battery capacity labels according to the distribution proportion.

[0032] Through the technical solution of this implementation method, reliable battery capacity labels are first screened out, and then screened through distribution percentage statistics, which can improve the credibility and distribution rationality of samples in each capacity interval and avoid too many or too few interval samples that do not meet the prediction requirements.

[0033] In some embodiments, the training data set acquisition submodule includes: A first correlation calculation submodule, used to perform a first correlation analysis between the battery capacity tag and the corresponding target operating condition index sample, to obtain a first correlation analysis result between the battery capacity tag and the corresponding target operating condition index sample; The first correlation screening submodule is used to screen target operating condition indicator samples that meet a first correlation condition for the battery capacity tag based on the first correlation analysis result.

[0034] Through the technical solution of this implementation method, based on the correlation between the battery capacity label and the target operating condition indicator sample, it is possible to effectively exclude the operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity label, thereby improving the model training efficiency and model prediction accuracy.

[0035] In some embodiments, the training data set acquisition submodule includes: A second correlation calculation submodule, configured to perform a second correlation analysis between a plurality of target operating condition index samples corresponding to any of the battery capacity tags, and obtain a second correlation analysis result between the plurality of target operating condition index samples; The second correlation screening submodule is used to screen out target operating condition indicator samples that meet the second correlation condition for any of the battery capacity tags based on the second correlation analysis result.

[0036] Through the technical solution of this implementation method, it is possible to effectively eliminate redundant samples based on the correlation between multiple target operating condition indicator samples, reduce data complexity, and improve model training efficiency and model prediction accuracy.

[0037] In some embodiments, the relevance condition includes: The correlation between any of the battery capacity tags and the corresponding target operating condition indicator samples is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples corresponding to any of the battery capacity tags is less than or equal to a second preset correlation threshold.

[0038] The technical solution of this implementation method can screen out battery capacity labels and corresponding target operating condition indicator samples whose correlation is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples is less than or equal to a second preset correlation threshold, thereby effectively excluding operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity labels, and eliminating redundant samples from multiple strongly correlated operating condition indicator samples, reducing data complexity, and improving model training efficiency and model prediction accuracy.

[0039] In some embodiments, the apparatus further comprises: A covariance matrix calculation module, used to calculate the covariance matrix corresponding to the set of target operating condition index samples; A feature contribution acquisition submodule, used to determine the feature contribution corresponding to each target operating condition index sample in the set of target operating condition index samples based on the covariance matrix; A removal submodule, configured to remove target operating condition index samples whose feature contributions do not meet the contribution condition from the set of target operating condition index samples based on the feature contributions corresponding to the respective target operating condition index samples, so as to obtain an updated target operating condition index sample set; The training set acquisition submodule is used to obtain the training data set based on the updated target operating condition index sample set.

[0040] The technical solution of this implementation method calculates the covariance matrix of the set of target operating condition index samples, decomposes the feature contribution, screens the target operating condition index samples that meet the contribution conditions based on the feature contribution, and eliminates the target operating condition index samples that do not meet the conditions, thereby improving the sample quality of the target operating condition index sample set, achieving sample dimensionality reduction, and effectively improving the model training efficiency and training effect.

[0041] In some embodiments, the error analysis submodule includes: An input submodule, used to input the target operating condition index samples in the training data set into the initial battery capacity prediction model to obtain the capacity prediction result output by the initial battery capacity prediction model; A calculation submodule, used to calculate the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation of the initial battery capacity prediction model based on the battery capacity label corresponding to the target operating condition index sample and the capacity prediction result; The analysis submodule is used to analyze the error influencing factors of the initial battery capacity prediction model based on the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation.

[0042] The technical solution of this implementation method analyzes the distribution of the mean square error, mean absolute error, determination coefficient and each prediction deviation to analyze the error influencing factor, and performs sample denoising on the training data set based on the error influencing factor, thereby reducing the error caused by abnormal samples, improving the model training effect, and obtaining a more optimized battery capacity prediction model.

[0043] In a fourth aspect, a battery capacity prediction device is provided, comprising: a prediction request receiving module, used to receive a battery capacity prediction request and determine the operating condition index of the battery to be tested; a prediction result output module, used to input the operating condition index of the battery to be tested into a battery capacity prediction model, and obtain a capacity prediction result output by the battery capacity prediction model; wherein the battery capacity prediction model is pre-trained based on operating condition data, and the operating condition data includes: operating condition data corresponding to multiple dimensions in battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model.

[0044] Through the technical solution of this implementation method, the battery capacity prediction model can effectively avoid the interference of error factors, and the operating condition data of the pre-trained model also includes the vehicle battery consistency evaluation data and / or vehicle model, and the vehicle battery consistency or vehicle model has a great influence on the battery capacity. For example, if the vehicle battery consistency is relatively good, the charging and discharging of each battery in the vehicle is relatively balanced, and the overall battery capacity of the vehicle is high. If the vehicle battery consistency is relatively poor, the charging and discharging of each battery in the vehicle is relatively unbalanced, resulting in a low overall battery capacity of the vehicle. Different models have different algorithms and strategies for battery management, and the differences in algorithms and strategies will also have a great impact on the battery capacity. Therefore, the data used for training can cover various operating conditions of the vehicle, learn the mapping relationship between different battery capacities and different operating condition data, and then accurately predict the battery capacity based on the operating condition indicators in the application, effectively improving the capacity prediction capability and prediction accuracy of the vehicle under various conditions.

[0045] In a fifth aspect, an electronic device is provided, comprising: a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein when the program or instruction is executed by the processor, the steps of the battery capacity prediction model training method described in the first aspect or the battery capacity prediction method described in the second aspect are implemented.

[0046] In a sixth aspect, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the battery capacity prediction model training method described in the first aspect or the battery capacity prediction method described in the second aspect are implemented.

[0047] In a seventh aspect, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of the battery capacity prediction model training method described in the first aspect or the battery capacity prediction method described in the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the drawings without paying creative work.

[0049] Figure 1 is an architecture diagram of a battery capacity prediction system applicable to an embodiment of the present application; Figure 2 This is one of the schematic flow charts of a battery capacity prediction model training method provided in an embodiment of the present application; Figure 3 This is the second schematic flow chart of a battery capacity prediction model training method provided in an embodiment of the present application; Figure 4 This is a schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Figure 5 This is a fourth schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Figure 6 This is a fifth schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Figure 7 This is a sixth schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Figure 8 This is the seventh schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Fig. 9 This is an eighth schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Fig.10 This is a ninth schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Fig.11 This is a schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Fig.12 This is a schematic flowchart of a battery capacity prediction model training method provided in an embodiment of the present application; Fig.13 is a schematic flow chart of a battery capacity prediction method provided in an embodiment of the present application; Fig.14 It is a schematic structural block diagram of a battery capacity prediction model training device provided in an embodiment of the present application; Fig.15 is a schematic structural block diagram of a battery capacity prediction device provided in an embodiment of the present application; Fig.16 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0050] The following detailed description and drawings of the embodiments of the present application are used to illustrate the principles of the present application, but cannot be used to limit the scope of the present application, that is, the present application is not limited to the described embodiments.

[0051] In the description of this application, it should be noted that, unless otherwise specified, "multiple" means more than two; the terms "upper", "lower", "left", "right", "inner", "outer", etc., indicating directions or positional relationships, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting this application. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0052] In the development of electric vehicles, battery capacity has an important impact. Based on the importance of battery capacity, in order to improve the reliability of electric vehicles, some technical solutions for battery capacity prediction have emerged. However, these technical solutions cannot predict the battery capacity fluctuations of some electric vehicles very well, and there is a problem of large prediction deviation.

[0053] In view of this, in order to improve the accuracy of battery capacity prediction, the present application proposes a new battery capacity prediction model training method, a battery capacity prediction method and related equipment.

[0054] Figure 1 The figure shows an architecture diagram of a battery capacity prediction system applicable to an embodiment of the present application.

[0055] like Figure 1 As shown, the battery capacity prediction system 100 may include: an on-board terminal 110 and a server 120 in the cloud. The on-board terminal 110 and the server 120 are connected via the Internet of Vehicles. The vehicle deployed with the on-board terminal 110 includes at least one battery pack, and the at least one battery pack as a whole may be collectively referred to as a battery 121. In terms of the type of battery, the battery 121 may be any type of battery, including but not limited to: lithium-ion batteries (such as ternary lithium-ion batteries), lithium metal batteries, lithium-sulfur batteries, lead-acid batteries, nickel-cathode batteries, nickel-hydrogen batteries, or lithium-air batteries, etc. In the embodiment of the present application, the specific type and scale of the battery 121 are not specifically limited.

[0056] Optionally, the vehicle may be an electric vehicle (eg, including a pure electric vehicle and a plug-in hybrid electric vehicle).

[0057] For example, the battery capacity prediction model training method and the battery capacity prediction method provided in the embodiment of the present application can be applied to the server 120 in the cloud, and the vehicle terminal 110 is connected to the vehicle network to obtain the battery operating data and vehicle operating data of the vehicle. The server 120 extracts the target operating condition index sample from the acquired data to train the battery capacity prediction model, obtains the trained battery capacity prediction model, and uses the trained battery capacity prediction model to predict the battery capacity based on the newly input operating condition index.

[0058] It should be noted that Figure 1 The battery capacity prediction system shown is an optional application scenario of the present application. According to actual application needs, the present application can also be applied to other optional application scenarios.

[0059] Figure 2 The schematic flowchart of a battery capacity prediction model training method proposed in the embodiment of the present application is shown. Optionally, the method of the embodiment of the present application can be applied to the above Figure 1 Server 120 is shown.

[0060] like Figure 2 As shown, the battery capacity prediction model training method may include the following steps.

[0061] Step 210: Obtain the battery capacity label.

[0062] The battery capacity label is the label data used to train the battery capacity prediction model, which can be understood as the actual capacity value corresponding to the working condition indicator sample. The battery capacity label can be obtained through real historical data, experimental measurement, database query, or Internet of Vehicles.

[0063] Optionally, in some embodiments, step 210 may obtain a battery capacity tag that meets a tag selection condition.

[0064] The label selection condition is a condition for filtering the battery capacity label. Optionally, the label selection condition can be customized or set by the system default and saved in the configuration information, so that the label selection condition can be read from the configuration information when needed; or, the label selection condition can be written into the prediction logic, and the label selection condition written therein is used for filtering when the prediction logic is executed.

[0065] In some embodiments, the label selection conditions may include at least one of the following conditions: before charging, the battery should reach a preset capacity low-end threshold when discharging; before charging is completed, the battery should reach a preset capacity full-charge threshold when charging is completed; after charging is completed, the battery should reach a preset rest time threshold when the battery is idle; during the charging process, no preset impact event occurs. Among them, the preset capacity low-end threshold, the preset capacity full threshold, the preset rest time threshold, and the preset impact event can be obtained based on experience or historical data statistics.

[0066] Exemplarily, the tag selection conditions may be as follows: the battery should be discharged to a low end before charging, such as the state of charge (SOC) dropping to 20%; after charging, the battery should be left to stand for at least 1 hour; and no preset influencing events such as backhaul and discharge occur during the charging process.

[0067] In other embodiments, the tag selection conditions may include: time range conditions, battery type conditions, battery environment temperature conditions, etc.

[0068] For example, the battery capacity label may be expressed in percentage to represent the state of health (SOH) of the battery.

[0069] Taking the battery capacity prediction of lithium iron phosphate (LFP) batteries as an example, considering the charging platform characteristics of LFP batteries and the polarization phenomenon of batteries during charging, we can screen battery capacity labels that meet the conditions of discharging to the low end, such as SOC dropping to 20%, charging to the full charge cut-off voltage, such as 3.7V, and sufficient standing time after charging, such as 1 hour, and no influence of feedback, discharge, etc. during the charging process.

[0070] Step 220: Obtain the operating condition data of the battery operating condition and the vehicle operating condition corresponding to the battery capacity label, wherein the operating condition data includes: operating condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model.

[0071] It is understandable that the operating data corresponding to the battery capacity tag can be the battery operating data corresponding to the charging process of the battery capacity tag and the vehicle operating data corresponding to multiple time nodes from the beginning of the vehicle life cycle to the charging time node, including at least: battery storage, battery cycle, battery temperature, battery state of charge and vehicle cumulative mileage corresponding to multiple dimensions of operating data. The operating data also includes: vehicle battery consistency assessment data and / or vehicle model.

[0072] Battery storage is the operating condition data used to describe the storage status of the battery. For example, battery storage may include any one or more operating condition data such as storage time and storage conditions, which may be expressed as: Battery cycle is a condition data used to describe the cycle use of the battery. For example, the battery cycle may include condition data such as the number of cycles.

[0073] Battery temperature is a working condition data used to describe the temperature of the battery. For example, the battery temperature may include working condition data such as the ambient temperature during the battery charging process and the battery's own temperature.

[0074] The battery state of charge (SOC) is the percentage of the battery's current remaining power to its total capacity.

[0075] The vehicle's cumulative mileage is used to describe the vehicle's cumulative mileage from the beginning of its life cycle to the charging time node.

[0076] The vehicle battery consistency assessment is a consistency assessment of the battery composition of the vehicle. For example, an electric vehicle includes battery A, battery B, and battery C. By evaluating the matching degree of multiple dimensions of the AC battery (such as but not limited to performance, function, and process parameters), consistency assessment data can be obtained. For another example, the difference between the SOC and SOH of multiple batteries in the vehicle under low-end static conditions can be analyzed, and based on this difference, the vehicle battery consistency assessment data can be obtained.

[0077] Vehicle type is a parameter used to describe the type or model of a vehicle. For example, it can be divided into types such as size, purpose, and performance.

[0078] Optionally, the operating condition data may also include operating condition data corresponding to at least one dimension of user behavior data and battery cell type. User behavior can be understood as operating condition data related to the user's driving behavior habits and / or charging behavior habits. For example, it includes but is not limited to: vehicle speed, driving time, charging habits, etc. Battery cell type is a parameter used to describe the type or model of battery cells, such as chemical composition, performance characteristics, etc.

[0079] The battery capacity label corresponds to the operating condition data of the battery operating condition and the vehicle operating condition, which can be obtained through any one or more acquisition methods such as real historical data, experimental measurement, database query from the database, and Internet of Vehicles.

[0080] Optionally, the cloud server can obtain the battery condition data and vehicle condition data saved or collected in real time by the vehicle terminal through the Internet of Vehicles. The vehicle terminal can also be understood as an on-board computer. The server can centrally manage the battery condition data and vehicle condition data from multiple vehicles, perform batch processing and analysis, and extract high-coverage condition indicator samples. Before training the battery capacity prediction model, the server can connect to the vehicle terminal through the Internet of Vehicles and remotely access the condition data.

[0081] The vehicle terminal can collect battery operating data and vehicle operating data through vehicle components such as vehicle sensors. Before uploading the data to the server, or after the server receives the data uploaded by the vehicle terminal, it can also pre-process the operating data, including but not limited to: Data cleaning: remove outliers and noise to improve data quality.

[0082] In this step, unlike the previous scheme that only uses the charging process conditions and cycle numbers, the battery mechanism is combined to obtain the operating condition data of the battery and vehicle conditions from the perspective of describing the influencing factors of battery life decay. The operating condition data may include the accumulated operating condition data at different time points in the entire life cycle of the vehicle, thereby providing a rich data foundation for the subsequent training of the battery capacity prediction model.

[0083] Step 230: Based on the operating condition data, a battery capacity prediction model is trained.

[0084] The battery capacity prediction model can be implemented through various machine learning models. Optionally, in some embodiments, the battery capacity prediction model can be implemented using an XGBoost (Extreme Gradient Boosting) model. The XGBoost model is a machine learning algorithm based on gradient boosting decision trees (GBDT). It can be understood that the XGBoost model is an optional implementation of the battery capacity prediction model, and the battery capacity prediction model can also be implemented through other types of machine learning models, such as the LightGBM model.

[0085] The technical solution of the embodiment of the present application, after obtaining the battery capacity label, obtains operating condition data of multiple dimensions including battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage. The operating condition data also includes: vehicle battery consistency evaluation data and / or vehicle model, and the vehicle battery consistency or vehicle model has a great influence on the battery capacity. For example, if the vehicle battery consistency is relatively good, the charging and discharging of each battery of the vehicle is relatively balanced, and the overall battery capacity of the vehicle is high. If the vehicle battery consistency is relatively poor, the charging and discharging of each battery of the vehicle is relatively unbalanced, resulting in a low overall battery capacity of the vehicle. Different models have different algorithms and strategies for battery management, and the differences in algorithms and strategies will also have a great impact on the battery capacity, so that the data used for training can cover various operating conditions of the vehicle. Therefore, based on the operating condition data, the battery capacity prediction model is trained, so that the data used for training can cover various operating conditions of the vehicle, and the trained battery capacity prediction model is obtained, which effectively improves the model's capacity prediction ability and prediction accuracy for vehicles under various conditions.

[0086] In some embodiments, Figure 3 The schematic flowchart of a battery capacity prediction model training method proposed in an embodiment of the present application is shown. Step 230 in other embodiments can be expressed as steps 330 to 370 in this embodiment. Among them: Step 330: extracting a target operating condition index sample corresponding to the battery capacity tag from the operating condition data.

[0087] Optionally, after obtaining the operating condition data, the operating condition data may be sorted out, such as analyzing the dimensions to which the data belongs, and classifying the data into corresponding types of target operating condition indicator samples according to the dimensions to which the data belongs, so as to extract different types of target operating condition indicator samples.

[0088] The target operating condition index samples can be understood as operating condition index samples of various dimensions that affect the battery capacity. For example, corresponding to the operating condition data of the above multiple dimensions, the types of target operating condition index samples may include but are not limited to: multiple index samples of battery storage index samples, battery cycle index samples, battery temperature index samples, SOC index samples and vehicle mileage index samples.

[0089] Among them, the battery storage index samples can be understood as index samples related to the battery storage dimension, such as but not limited to: battery storage conditions, storage time, etc.

[0090] Battery cycle index samples can be understood as index samples related to battery cycle dimensions, including but not limited to: number of cycles, charge and discharge depth, etc.

[0091] The battery temperature indicator sample can be understood as an indicator sample related to the battery temperature dimension, including but not limited to: the internal and external temperature of the battery.

[0092] The SOC indicator sample can be understood as an indicator sample related to the power level, including but not limited to: state of charge, state of charge fluctuation range, state of charge change rate, etc.

[0093] The vehicle mileage indicator sample can be understood as an indicator sample related to the vehicle cumulative mileage dimension, including but not limited to: cumulative mileage, average single mileage, etc.

[0094] The sample of vehicle battery consistency evaluation indicators can be understood as the sample of indicators related to the vehicle battery consistency evaluation data.

[0095] Vehicle type index samples can be understood as indicator samples related to vehicle types.

[0096] Step 340: By performing a correlation analysis on the battery capacity label and the corresponding target operating condition indicator sample, the target operating condition indicator samples that meet the correlation condition are screened out to obtain a training data set.

[0097] Correlation analysis may include, but is not limited to: correlation analysis between battery capacity labels and corresponding target operating condition indicator samples, monotonic analysis, correlation analysis between target operating condition indicator samples, etc. Through correlation analysis, target operating condition indicator samples that meet the correlation conditions can be screened, which can improve model training efficiency and model prediction accuracy.

[0098] Step 350: Train an initial battery capacity prediction model using the training data set, and analyze error influencing factors of the initial battery capacity prediction model, wherein the battery capacity prediction model is used to predict battery capacity.

[0099] The error influencing factor is a factor that causes an error between the prediction result of the battery capacity model prediction and the battery capacity label. The error influencing factor may include: the target operating condition indicator sample that causes the error, the parameter value of any one or more parameters of the sample, the battery capacity label corresponding to the sample, etc.

[0100] In the training process of step 350, the error of the battery capacity prediction model can be evaluated based on the battery capacity label corresponding to the target operating condition index sample and the capacity prediction result of the battery capacity prediction model, and any one or more error evaluation algorithms, and the error influencing factor can be analyzed based on the error. The error evaluation algorithm includes but is not limited to any one or more algorithms such as mean square error, mean absolute error, and determination coefficient.

[0101] Optionally, in some embodiments, based on the battery capacity label and capacity prediction results corresponding to the target operating condition index sample, the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation of the battery capacity prediction model are calculated. According to the distribution of the mean square error, mean absolute error, determination coefficient and each prediction deviation, for predictions with large errors (such as greater than a certain error threshold), the feature weights of the statistical model are calculated, and features with feature weights greater than a certain weight threshold (features that cause large model errors) are found as error influencing factors based on feature omissions.

[0102] In other embodiments, the mean square error of the battery capacity prediction model and the distribution of each prediction deviation are calculated based on the battery capacity label and capacity prediction result corresponding to the target operating condition index sample. According to the distribution of the mean square error and each prediction deviation, for the prediction with large error, the feature weight of the statistical model is calculated, and the feature ranking in the front preset position is found as the error influencing factor according to the feature missing weight sorting from large to small.

[0103] For example, if analysis finds that the vehicle mileage of a vehicle mileage index sample A is extremely small, and there is a large error in the correspondence between the extremely small vehicle mileage and the corresponding battery capacity label, based on this, it can be analyzed that the error influencing factor is the vehicle mileage of the vehicle mileage index sample or its correspondence with the battery capacity label.

[0104] In some embodiments, in order to better train the battery capacity prediction model, the training data set can also be divided into two parts according to the method of the frame number, and the ratio of the two parts can be set as needed, such as 7:3. Among them, a part of the training data set (such as 70%) is used as a training set and input into the battery capacity prediction model for training. Another part of the training data set (such as 30%) is used as a test set and input into the battery capacity prediction model for prediction testing. By dividing the training data set into a ratio of 7:3 according to the method of the frame number, the data in the training set and the test set can come from different vehicles, thereby avoiding data leakage and overfitting.

[0105] Step 360: Based on the error influencing factor, perform sample denoising on the training data set to obtain an updated training data set.

[0106] Sample noise reduction refers to adjusting the target operating condition index samples corresponding to the error influencing factors with larger error influence to reduce the error impact.

[0107] Optionally, the target operating condition index samples corresponding to the error influencing factors having a large impact on the error may be deleted from the training data set, and / or the index values ​​of the target operating condition index samples corresponding to the error influencing factors having a large impact on the error may be adjusted to eliminate the impact of the error.

[0108] Exemplarily, assuming that based on the aforementioned analysis steps, the error influencing factor with a large error impact is determined to be an index sample with a vehicle mileage less than a certain value, then the index sample can be deleted, or the vehicle mileage or the corresponding battery capacity label of the index sample can be reasonably increased or decreased to eliminate the error.

[0109] Step 370: Train the initial battery capacity prediction model using the updated training data set to obtain a trained battery capacity prediction model.

[0110] Optionally, the specific process of training the model based on the denoised training data may include: Determine the target operating condition index samples to be trained from the denoised training data set; Inputting the target operating condition indicator sample to be trained into the battery capacity prediction model to obtain the capacity prediction result output by the battery capacity prediction model; Adjust the model parameters based on the difference between the battery capacity label corresponding to the target operating condition indicator sample and the capacity prediction result; If the training stop condition is not met, return to the step of determining the target operating condition indicator sample to be trained from the denoised training data set, so as to continue training the battery capacity prediction model; When the training stop condition is met, a trained battery capacity prediction model is obtained.

[0111] After obtaining the trained battery capacity prediction model, the battery capacity prediction model can be applied to actual battery capacity prediction.

[0112] Optionally, when the battery capacity prediction method provided in the embodiment of the present application is applied to a server in the cloud, the battery capacity prediction model can be trained on the server, and the battery capacity prediction model can be deployed on the server. The server can provide a WEB application interface, receive a battery capacity prediction request through the WEB application interface, determine the operating condition index of the battery to be tested according to the battery capacity prediction request, input the operating condition index of the battery to be tested into the battery capacity prediction model, and obtain the capacity prediction result output by the battery capacity prediction model.

[0113] The technical solution of the embodiment of the present application, after obtaining the battery capacity label, extracts the operating condition data of multiple dimensions including battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, the operating condition data also includes: vehicle battery consistency evaluation data and / or vehicle model, extracts the target operating condition indicator samples corresponding to the battery capacity label from the operating condition data, so that the samples can cover various operating conditions of the vehicle, and performs correlation analysis on the target operating condition indicator samples, screens out the target operating condition indicator samples that meet the correlation conditions, realizes sample dimensionality reduction, and then trains the initial battery capacity prediction model through the training data set to analyze the error influence shadow, so that the training data set can be sample denoised based on the error influence factor to obtain an updated training data set, train the initial battery capacity prediction model, and obtain the trained battery capacity prediction model, which effectively improves the model's capacity prediction capability and prediction accuracy for vehicles under various conditions.

[0114] It should also be noted that although the relevant technology provides some battery capacity prediction methods, these methods are usually composed of a series of algebraic and differential equations to predict battery capacity, or use historical data to establish some simple mapping relationships, which cannot achieve good prediction. However, the embodiment of the present application drives the machine learning model through operating data in multiple dimensions, and adjusts the training samples by analyzing the error influencing factors. The optimized battery capacity prediction model is obtained based on the training samples after noise reduction, which has flexible application scenarios, is easy to use, and has accurate prediction.

[0115] In some embodiments, Figure 4 The schematic flowchart of a battery capacity prediction model training method proposed in an embodiment of the present application is shown. Step 220 in other embodiments can be expressed as step 420 in this embodiment. Among them: Step 420: Obtain the operating condition data of the battery operating condition and the vehicle operating condition corresponding to the battery capacity label, wherein the operating condition data include: operating condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes: user behavior data, vehicle battery consistency assessment data, and operating condition data corresponding to vehicle model and battery cell type.

[0116] It should be noted that Figure 4 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0117] In some embodiments, Figure 5 The schematic flowchart of a battery capacity prediction model training method proposed in an embodiment of the present application is shown. Step 210 in other embodiments can be expressed as step 510 in this embodiment. Among them: Step 510: Obtain a battery capacity tag that meets a tag selection condition, where the tag selection condition includes at least one of the following conditions: The battery corresponding to the battery capacity label should be discharged to the preset low-end capacity threshold before charging; The battery corresponding to the battery capacity label should reach the preset full-charge threshold before charging ends; After the battery corresponding to the battery capacity label is charged, the battery rest time should reach the preset rest time threshold; The battery corresponding to the battery capacity label does not experience any preset impact event during the charging process; The distribution ratio of battery capacity labels meets the distribution ratio condition.

[0118] Among them, the preset capacity low-end threshold, the preset capacity full threshold, the preset idle time threshold, and the preset impact event can be obtained based on experience or historical data statistics.

[0119] Before charging, the battery should be discharged to a preset low-end capacity threshold. The significance of this condition setting is to constrain the charging process to start from a relatively stable starting point and reduce the uncertainty of the initial state. For example, the preset low-end capacity threshold can be set to the state of charge dropping to 20%.

[0120] Before the battery is fully charged, it should reach the preset full-capacity threshold. The purpose of setting this condition is to constrain the battery to be fully charged to reach the maximum capacity and reduce the interference of insufficiently charged data on battery capacity prediction.

[0121] After the battery is charged, the battery rest time should reach the preset rest time threshold. The significance of setting this condition is to reduce the polarization phenomenon during the charging process and improve the accuracy of the data. For example, the preset rest time threshold can be set to 1 hour.

[0122] During the battery charging process, no preset influencing event occurs. The significance of setting this condition is to avoid interference of external factors on the charging process and ensure the purity of data. For example, the preset influencing event may include events such as backhaul and discharge.

[0123] The distribution ratio condition is a screening condition set to improve the rationality of the sample distribution in each capacity interval. For example, the distribution ratio condition can be expressed as the distribution ratio of each capacity interval is equal, or the distribution ratio of each capacity interval is equal to a preset ratio.

[0124] Through the technical solution of this implementation method, it is possible to effectively select battery capacity labels whose corresponding data are pure, accurate, certain, with little interference and reasonable distribution, thereby improving the prediction accuracy of the model.

[0125] It should be noted that Figure 5 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0126] In some embodiments, Figure 6 The schematic flowchart of a battery capacity prediction model training method proposed in the embodiment of the present application is shown. Step 510 in other embodiments can be expressed as steps 611 to 613 in this embodiment. Among them: Step 611: Obtain the battery capacity label of the corresponding battery before charging, when the discharge reaches the preset capacity low-end threshold, and before the end of charging, the battery reaches the preset capacity full-charge threshold, and after the end of charging, the battery rest time reaches the preset rest time threshold, and no preset influencing event occurs during the charging process, to obtain an initial battery capacity label set.

[0127] The specific implementation of step 611 can refer to the relevant description of step 510 in other embodiments, and will not be repeated here.

[0128] Step 612: Count the distribution ratios of the initial battery capacity tag set in multiple capacity intervals.

[0129] The capacity intervals are obtained by dividing the capacity tag data range of the initial battery capacity tag set into different intervals. The distribution ratio is the ratio of the number of battery capacity tags distributed in each capacity interval.

[0130] Exemplarily, the capacity tag data range of the initial battery capacity tag set is 0%-100%, which can be evenly divided into 10 capacity intervals. The data span of each capacity interval is 10%. The number of battery capacity tags distributed in the 10 capacity intervals in the initial battery capacity tag set is counted to obtain the distribution ratio of multiple capacity intervals.

[0131] Step 613: According to the distribution ratio, the battery capacity labels that meet the distribution ratio condition are screened out from the initial battery capacity labels.

[0132] Through the technical solution of this implementation method, reliable battery capacity labels are first screened out, and then screened through distribution percentage statistics, which can improve the credibility and distribution rationality of samples in each capacity interval and avoid too many or too few interval samples that do not meet the prediction requirements.

[0133] It should be noted that Figure 6 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0134] In some embodiments, Figure 7 The schematic flowchart of a battery capacity prediction model training method proposed in an embodiment of the present application is shown. Step 340 in other embodiments can be expressed as steps 741-742 in this embodiment. Among them: Step 741: Perform a first correlation analysis between the battery capacity tag and the corresponding target operating condition index sample to obtain a first correlation analysis result between the battery capacity tag and the corresponding target operating condition index sample.

[0135] It is understandable that by analyzing the dimensions to which the data belongs, more target operating condition indicator samples will be sorted out, and these target operating condition indicator samples have different degrees of correlation with the battery capacity prediction, and some are even not correlated. Therefore, in order to improve the model training efficiency and training effect, the embodiment of the present application will perform a first correlation analysis to screen out target operating condition indicator samples that can improve the model training efficiency and training effect.

[0136] The first correlation analysis refers to the analysis of the correlation between the battery capacity label and the corresponding target operating condition indicator sample. The analysis algorithm of the first correlation analysis is not limited. Exemplarily, the first correlation analysis may include: any one or more of monotonicity analysis and correlation degree analysis. Monotonicity analysis is an analysis of whether there is monotonicity in the relationship between the battery capacity label and any target operating condition indicator sample (e.g., when the value of any target operating condition indicator sample increases, the battery capacity label value also increases or decreases monotonically). Correlation degree analysis is an analysis of the degree of influence of any target operating condition indicator sample on the battery capacity label.

[0137] The first correlation analysis result is used to describe the correlation between the battery capacity label and the target operating condition index sample. For example, the first correlation analysis result may be manifested as: the battery capacity label and the battery storage index sample are monotonic, the battery capacity label and the vehicle driving behavior index sample are strongly correlated, and so on.

[0138] Step 742: Based on the first correlation analysis result, filter out target operating condition indicator samples that meet the first correlation condition for the battery capacity tag.

[0139] The first correlation condition may be set according to actual application needs. For example, the first correlation condition may include but is not limited to: screening out operating condition indicator samples that are monotonically and / or strongly correlated with the battery capacity label as target operating condition indicator samples.

[0140] Through the technical solution of this implementation method, based on the correlation between the battery capacity label and the target operating condition indicator sample, it is possible to effectively exclude the operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity label, thereby improving the model training efficiency and model prediction accuracy.

[0141] It should be noted that Figure 7 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0142] In some embodiments, such as Figure 8 The schematic flowchart of a battery capacity prediction model training method proposed in the embodiment of the present application is shown. Step 340 in other embodiments can be expressed as steps 841-842 in this embodiment. Among them: Step 841: Perform a second correlation analysis on a plurality of target operating condition indicator samples corresponding to any of the battery capacity tags to obtain a second correlation analysis result between the plurality of target operating condition indicator samples.

[0143] It is understandable that when multiple target operating condition indicator samples have a high correlation with the battery capacity label, it will be difficult to distinguish the degree of influence of various indicators on the prediction results during model training, thereby increasing the complexity of the model, low computational efficiency, and possibly leading to overfitting. Based on this, in order to improve the model training efficiency and training effect, the embodiment of the present application will conduct a second correlation analysis on multiple candidate operating condition indicator samples to eliminate highly correlated and redundant operating condition indicator samples, thereby screening out target operating condition indicator samples that can improve the model training efficiency and training effect.

[0144] The second correlation analysis is an analysis of the correlation between multiple target operating condition index samples. The analysis algorithm of the second correlation analysis is not limited. For example, the following Pearson correlation coefficient calculation formula can be used for analysis:

[0145] The calculation formula of the Pearson correlation coefficient is used to measure the linear correlation between two variables. The two variables can be understood as any two target working condition indicator samples. The value range of the Pearson correlation coefficient is -1 to 1, where: 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no correlation. , represents the (i)th observation value of variable (x). represents the (i)th observation value of variable (y). Among them, represents the mean of all observed values ​​of a variable ( x ), Represents the mean of all observed values ​​of a variable (y).

[0146] Step 842: Based on the second correlation analysis result, filter out target operating condition indicator samples that meet the second correlation condition for any of the battery capacity tags.

[0147] The second correlation analysis result is used to describe the correlation between multiple target operating condition index samples. For example, the second correlation analysis result may be expressed as: target operating condition index sample A is completely positively correlated with target operating condition index sample B. In this case, any target operating condition index sample can be selected from target operating condition index sample A and target operating condition index sample B as the target operating condition index sample, and the other target operating condition index sample is eliminated.

[0148] Through the technical solution of this implementation method, it is possible to effectively eliminate redundant samples based on the correlation between multiple target operating condition indicator samples, reduce data complexity, and improve model training efficiency and model prediction accuracy.

[0149] It should be noted that Figure 8 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0150] In some embodiments, such as Fig. 9 The schematic flowchart of a battery capacity prediction model training method proposed in an embodiment of the present application is shown. Step 340 in other embodiments can be expressed as step 940 in this embodiment. Among them: Step 940: By performing a correlation analysis on the battery capacity label and the corresponding target operating condition indicator sample, the target operating condition indicator samples that meet the correlation conditions are screened out to obtain a training data set, and the correlation conditions include: the correlation between any of the battery capacity labels and the corresponding target operating condition indicator sample is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples corresponding to any of the battery capacity labels is less than or equal to a second preset correlation threshold.

[0151] The first preset correlation threshold is a threshold used to constrain the strength of the correlation between the battery capacity label and the candidate operating condition indicator sample. The first preset correlation threshold can be expressed as a value between 0 and 1, indicating the absolute value of the correlation. For example, if the first preset correlation threshold is set to 0.5, the initial operating condition indicator samples whose correlation with the battery capacity label is greater than or equal to 0.5 can be screened out as candidate operating condition indicator samples.

[0152] The second preset correlation threshold is a threshold used to constrain the strength of the correlation between multiple candidate operating condition indicator samples. The second preset correlation threshold can be expressed as a numerical value, indicating the absolute value of the correlation. For example, if the second preset correlation threshold is set to 1, candidate operating condition indicator samples with a correlation less than 1, such as 0 and -1, can be screened out, and only one of the multiple candidate operating condition indicator samples with a correlation of 1 can be selected.

[0153] Optionally, in some embodiments, a first correlation analysis can be performed between the battery capacity tag and the corresponding target operating condition indicator sample to obtain a first correlation analysis result between the battery capacity tag and the corresponding target operating condition indicator sample, and based on the first correlation analysis result, a target operating condition indicator sample that meets the first correlation condition is screened out for the battery capacity tag; then a second correlation analysis is performed between multiple target operating condition indicator samples corresponding to any battery capacity tag to obtain a second correlation analysis result between the multiple target operating condition indicator samples, and based on the second correlation analysis result, the target operating condition indicator samples whose correlation does not exceed a second preset correlation threshold are screened out, and, for multiple target operating condition indicator samples whose correlation exceeds the second preset correlation threshold, only one candidate operating condition indicator sample is screened out as the target operating condition indicator sample.

[0154] The technical solution of this implementation method can screen out battery capacity labels and corresponding target operating condition indicator samples whose correlation is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples is less than or equal to a second preset correlation threshold, thereby effectively excluding operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity labels, and eliminating redundant samples from multiple strongly correlated operating condition indicator samples, reducing data complexity, and improving model training efficiency and model prediction accuracy.

[0155] It should be noted that Fig. 9 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0156] In some embodiments, such as Fig.10 The schematic flowchart of a battery capacity prediction model training method proposed in the embodiment of the present application is shown. The difference from other embodiments is that after the target operating condition indicator samples that meet the correlation condition are screened out in step 340, before the model is trained in step 350, the embodiment further includes steps 1041 to 1043. Among them: Step 1041: Calculate the covariance matrix corresponding to the set of target operating condition index samples.

[0157] The covariance matrix is ​​a matrix that includes the covariances between each indicator sample.

[0158] Step 1042: Based on the covariance matrix, determine the feature contribution corresponding to each target operating condition index sample in the set of target operating condition index samples.

[0159] The feature contribution is the degree of influence of the target operating condition index sample on the battery capacity label. The feature contribution and the eigenvector are attributes of the covariance matrix and can be obtained by solving the equation. Optionally, each target operating condition index sample can obtain the corresponding feature contribution and eigenvector by solving the equation.

[0160] Step 1043: Based on the feature contribution corresponding to each target operating condition index sample, the target operating condition index samples whose feature contribution does not meet the contribution condition are removed from the set of target operating condition index samples to obtain an updated target operating condition index sample set, and based on the updated target operating condition index sample set, the training data set is obtained.

[0161] The contribution condition is a condition that uses the feature contribution as the judgment object and is used to screen the target working condition indicator samples. This condition can be set according to actual needs. Optionally, the contribution condition can be set based on experience, such as the feature contribution should be greater than the contribution threshold set based on experience; the contribution condition can be set based on the accumulation of contribution, such as the cumulative feature contribution should reach a certain percentage (for example, 99% or 90%).

[0162] For example, the characteristic contribution of each target operating condition index sample to the battery capacity label can be sorted from high to low, and the index samples with cumulative contribution of 99% are retained, and the rest are removed from the sample set.

[0163] The technical solution of this implementation method calculates the covariance matrix of the set of target operating condition index samples, decomposes the feature contribution, screens the target operating condition index samples that meet the contribution conditions based on the feature contribution, and eliminates the target operating condition index samples that do not meet the conditions, thereby improving the sample quality of the target operating condition index sample set, achieving sample dimensionality reduction, and effectively improving the model training efficiency and training effect.

[0164] It should be noted that Fig.10 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0165] In some embodiments, such as Fig.11 The schematic flowchart of a battery capacity prediction model training method proposed in an embodiment of the present application is shown. Step 350 in other embodiments can be expressed as steps 1151 to 1155 in this embodiment. Among them: Step 1151: input the target operating condition index samples in the training data set into the initial battery capacity prediction model to obtain the capacity prediction result output by the initial battery capacity prediction model.

[0166] Optionally, the initial battery capacity prediction model may be an untrained battery capacity prediction model or a pre-trained battery capacity prediction model.

[0167] Step 1152: Based on the battery capacity label corresponding to the target operating condition index sample and the capacity prediction result, calculate the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation of the initial battery capacity prediction model.

[0168] The mean squared error (MSE) is the average of the squares of the difference between the actual value of the battery capacity label corresponding to the sample and the predicted value output by the model. MSE is calculated using the following formula:

[0169] in, is the actual value, is the predicted value, m is the number of samples, and i represents the i-th sample.

[0170] It is understandable that MSE is sensitive to larger errors because the error is squared. The smaller the value of MSE, the better the predictive performance of the model.

[0171] Mean Absolute Error (MAE) is the average of the absolute values ​​of the difference between the actual value of the battery capacity label corresponding to the sample and the model predicted value. MAE is calculated using the following formula:

[0172] The smaller the MAE value, the better the prediction performance of the model.

[0173] The coefficient of determination (R²) is the proportion of the variation explained by the model to the total variation. The coefficient of determination is calculated using the following formula:

[0174] in, It is the average of the actual values ​​of all battery capacity labels.

[0175] The value range of R² is [-∞, 1]. The closer the value is to 1, the better the model fit is. If R² is 0, it means that the prediction effect of the model is the same as the average value. If R² is less than 0, it means that the prediction effect of the model is worse than the average value.

[0176] The distribution of each prediction deviation is the distribution of the deviation (also known as the difference) between each prediction value of the model and the actual value. By analyzing the deviation distribution, the prediction performance and potential problems of the model can be analyzed.

[0177] For example, the range of each prediction deviation of the model can be counted, the range can be divided into multiple sections, the number of deviations falling into each section can be counted, and the distribution of each prediction deviation can be obtained.

[0178] Step 1153: Analyze the error influencing factors of the initial battery capacity prediction model based on the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation.

[0179] Optionally, the feature weights of the model can be calculated based on the distribution of the mean square error, mean absolute error, coefficient of determination and deviations of each prediction, for predictions with large errors (such as greater than a certain error threshold). Based on the feature omissions, features with feature weights greater than a certain weight threshold that lead to large model errors can be found as error influencing factors.

[0180] Optionally, the target operating condition index samples corresponding to the error influencing factors having a large impact on the error may be deleted from the training data set, and / or the index values ​​of the target operating condition index samples corresponding to the error influencing factors having a large impact on the error may be adjusted to eliminate the impact of the error.

[0181] Exemplarily, assuming that based on the aforementioned analysis steps, the error influencing factor with a large error impact is determined to be an index sample with a vehicle mileage less than a certain value, the index sample can be deleted, or the vehicle mileage or the corresponding battery capacity label of the index sample can be reasonably adjusted to eliminate the error.

[0182] Based on the error influencing factor, sample denoising is performed on the training data set to obtain an updated training data set. The battery capacity prediction model is then trained with the updated training data set and the model parameters are adjusted to obtain an optimized battery capacity prediction model. The model training process can refer to the relevant description of the model training process in other embodiments, which will not be repeated here.

[0183] For example, the battery capacity prediction model optimized by this embodiment, in some prediction and test set experiments of the embodiments of this application, obtained the following experimental data:

[0184] The deviation between the prediction and the test set shows the intervals of different deviation ranges. For example, [-0.015, 0.015] means that the deviation between the predicted value and the actual value is between -0.015 and 0.015.

[0185] Cumulative proportion shows the proportion of samples in the corresponding deviation range and before the interval to the total samples. For example, 68.93% means that the samples with deviations in the range of [-0.015, 0.015] and before account for 68.93% of the total samples.

[0186] The percentage of each interval shows the proportion of samples in the corresponding deviation interval to the total samples. For example, 68.93% means that the samples with deviations in the range of [-0.015, 0.015] account for 68.93% of the total samples.

[0187] The above experimental data show that the model has good prediction effect, with MSE, MAE, and R2_Score of 0.0003, 0.0122, and 0.8079 respectively. Combined with multiple real vehicles, the deviation is within 5%, which can effectively predict the battery capacity.

[0188] The technical solution of this implementation method analyzes the distribution of the mean square error, mean absolute error, determination coefficient and each prediction deviation to analyze the error influencing factor, and performs sample denoising on the training data set based on the error influencing factor, thereby reducing the error caused by abnormal samples, improving the model training effect, and obtaining a more optimized battery capacity prediction model.

[0189] It should be noted that Fig.11 The specific implementation methods of other steps in the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0190] In some embodiments, such as Fig.12 A schematic flowchart of a battery capacity prediction model training method proposed in an embodiment of the present application is shown, and the method includes: Step 1210: Obtain battery capacity tags that meet the trusted tag screening condition as an initial battery capacity tag set.

[0191] Among them, the trusted label screening conditions include: before charging, the battery should discharge to a preset low-end capacity threshold; before the end of charging, the battery should charge to a preset full-capacity threshold; after charging, the battery should rest for a preset time threshold; during the charging process, no preset impact events occur.

[0192] Step 1211: Count the distribution ratios of the initial battery capacity tag set in multiple capacity intervals.

[0193] Step 1212: According to the distribution ratio, filter out the battery capacity labels that meet the distribution ratio condition from the initial battery capacity labels.

[0194] Through the above steps 1210 to 1212, it is possible to screen out battery capacity labels that are discharged to the low end, fully charged, fully rested, and not affected by discharge and recharge, and count the number of vehicles and the cumulative number of charging processes of a single vehicle, so as to determine whether the distribution of battery capacity labels in each capacity interval is uniform and reasonable, and finally obtain battery capacity labels that meet the sample distribution screening conditions as objects participating in testing and training.

[0195] Step 1220: Obtain the operating condition data of the battery operating condition and the vehicle operating condition during the charging process corresponding to the battery capacity tag.

[0196] Step 1231: extracting an initial operating condition indicator sample corresponding to the battery capacity tag from the operating condition data.

[0197] Through the above steps 1220-1231, the cumulative vehicle operating condition data from the start time of each charging of each vehicle to the start time of the vehicle data can be obtained, the noise in the operating condition data can be analyzed and filtered, and the relevant operating condition indicators can be sorted out based on the influencing factors of the battery capacity life decay, and the distribution rationality of each indicator can be statistically analyzed to finally determine the set of initial operating condition indicator samples corresponding to the battery capacity label.

[0198] Step 1241: Perform a first correlation analysis on the battery capacity tag and the corresponding initial operating condition indicator sample to obtain a first correlation analysis result between the battery capacity tag and the corresponding initial operating condition indicator sample.

[0199] Step 1242: Based on the first correlation analysis result, initial operating condition indicator samples whose correlation exceeds a first preset correlation threshold are selected for the battery capacity tag as candidate operating condition indicator samples, and a candidate operating condition indicator sample set corresponding to the battery capacity tag is obtained.

[0200] Step 1243: Perform a second correlation analysis on the candidate operating condition index samples in the candidate operating condition index sample set to obtain a second correlation analysis result between the candidate operating condition index samples.

[0201] Step 1244: Based on the second correlation analysis result, select candidate operating condition indicator samples whose correlation does not exceed the second preset correlation threshold from the candidate operating condition indicator sample set as target operating condition indicator samples, and select only one candidate operating condition indicator sample from multiple candidate operating condition indicator samples whose correlation exceeds the second preset correlation threshold as the target operating condition indicator sample, to obtain the target operating condition indicator sample set corresponding to the battery capacity label.

[0202] Step 1245: Calculate the covariance matrix corresponding to the target operating condition index sample set.

[0203] Step 1246: Based on the covariance matrix, determine the feature contribution corresponding to each operating condition index sample in the target operating condition index sample set.

[0204] Step 1247: Based on the feature contribution corresponding to each operating condition index sample, the operating condition index samples whose feature contribution does not meet the contribution condition are removed from the target operating condition index sample set to obtain an updated target operating condition index sample set.

[0205] Through the above steps 1231 to 1247, the correlation between the battery capacity label and each target operating condition indicator sample and whether there is monotonicity are analyzed, and the target operating condition indicator samples with correlation or monotonicity are screened out. Then, the correlation between each indicator sample is analyzed to see whether there is a strong correlation or repeated function, and the target operating condition indicator samples are further screened out to form a target operating condition indicator sample set. Finally, the target operating condition indicator sample set is subjected to principal component analysis to filter out features with low contribution rates that can be ignored, and the remaining target operating condition indicator samples with cumulative contribution rates exceeding 99% are retained to obtain a training data set, thereby achieving dimensionality reduction of the data set, which can improve the training efficiency of the model and the model prediction accuracy.

[0206] Step 1251: Input the target operating condition index samples in the training data set into the initial battery capacity prediction model to obtain the capacity prediction result output by the initial battery capacity prediction model.

[0207] Step 1252: Based on the battery capacity labels and capacity prediction results corresponding to the target operating condition index samples, calculate the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation of the initial battery capacity prediction model.

[0208] Step 1253: Analyze the error influencing factors of the initial battery capacity prediction model based on the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation.

[0209] Step 1254: Based on the error influence factor, perform sample denoising on the training data set to obtain an updated training data set.

[0210] Step 1255: Train the initial battery capacity prediction model using the updated training data set to adjust model parameters and obtain a trained battery capacity prediction model.

[0211] Through the above steps 1251-1255, the battery capacity prediction model is trained using the training data set, and the MSE and MAE indicators are output, the model prediction effect score is scored, and the single calculation deviation between the training set and the prediction set is statistically analyzed, and the distribution of each deviation is statistically analyzed. The influencing factors of the error vehicle are analyzed, the error influencing factors are sorted out, and the training data set is denoised based on the error influencing factors, so that the battery capacity prediction model can be continued to be trained using the denoised training data set, and the model parameters can be adjusted to achieve model optimization and obtain the trained battery capacity prediction model.

[0212] It should be noted that Fig.12 For specific implementation methods not described in detail in the illustrated embodiment, reference can be made to the relevant descriptions of other embodiments and will not be repeated here.

[0213] Fig.13 1 shows a schematic flow chart of a battery capacity prediction method proposed in an embodiment of the present application. Optionally, the method in the embodiment of the present application may be applied to the above Figure 1 Server 120 is shown.

[0214] like Fig.13 As shown, the battery capacity prediction method may include the following steps.

[0215] Step 1310: Receive a battery capacity prediction request and determine the operating condition index of the battery to be tested.

[0216] Step 1320: input the operating condition index of the battery to be tested into a battery capacity prediction model to obtain a capacity prediction result output by the battery capacity prediction model.

[0217] Among them, the battery capacity prediction model is pre-trained based on operating condition data, and the operating condition data includes: operating condition data corresponding to multiple dimensions in battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model.

[0218] Optionally, a training data set can be obtained based on the operating condition data, and the initial battery capacity prediction model can be trained based on the training data set. The training data set can be obtained by performing sample denoising through error influence factors, and the error influence factors can be obtained by training and analyzing the initial battery capacity prediction model through the training data set before sample denoising, and the training data set before sample denoising includes target operating condition indicator samples that meet the correlation conditions, and the target operating condition indicator samples that meet the correlation conditions are screened by performing correlation analysis on the battery capacity label and the corresponding target operating condition indicator samples, and the target operating condition indicator samples corresponding to the battery capacity label are extracted from the operating condition data.

[0219] Through the technical solution of this implementation method, the battery capacity prediction model can effectively avoid the interference of error factors, learn the mapping relationship between different battery capacities and different operating condition data, and then accurately predict the battery capacity based on the operating condition indicators in the application, effectively improving the capacity prediction ability and prediction accuracy of the vehicle under various conditions.

[0220] It should be noted that Fig.13 The specific implementation of the illustrated embodiment can refer to the relevant descriptions of other embodiments and will not be repeated here.

[0221] The above describes the specific implementation of the battery charging method provided in the embodiment of the present application. Figure 14 to Figure 15 The specific embodiments of the related devices provided in the present application are described. It can be understood that the related descriptions in the following device embodiments can refer to the aforementioned method embodiments, and for the sake of brevity, they are not repeated here.

[0222] Fig.14 FIG. 1 is a schematic structural block diagram of a battery capacity prediction model training device 1400 according to an embodiment of the present application. Fig.14 As shown, the battery capacity prediction model training device 1400 includes: A battery capacity label acquisition module 1410 is used to acquire a battery capacity label; The operating condition data acquisition module 1420 is used to acquire the operating condition data of the battery operating condition and the vehicle operating condition corresponding to the battery capacity label, wherein the operating condition data includes: operating condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle cumulative mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model; The model training module 1430 is used to train the battery capacity prediction model based on the operating condition data.

[0223] The technical solution of the embodiment of the present application, after obtaining the battery capacity label, obtains operating condition data of multiple dimensions including battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and trains the battery capacity prediction model based on the operating condition data, so that the data used for training can cover various operating conditions of the vehicle, and a trained battery capacity prediction model is obtained, which effectively improves the model's capacity prediction capability and prediction accuracy for the vehicle under various conditions.

[0224] Optionally, the model training module includes: An index sample extraction submodule, used to extract a target operating condition index sample corresponding to the battery capacity tag from the operating condition data; A training data set acquisition submodule is used to perform correlation analysis on the battery capacity label and the corresponding target operating condition indicator sample, filter out the target operating condition indicator sample that meets the correlation condition, and obtain the training data set; An error analysis submodule, used to train an initial battery capacity prediction model using the training data set and analyze error influencing factors of the initial battery capacity prediction model, wherein the battery capacity prediction model is used to predict battery capacity; A sample denoising submodule 1 is used to perform sample denoising on the training data set based on the error influencing factor to obtain an updated training data set; The training submodule is used to train the initial battery capacity prediction model using the updated training data set to obtain a trained battery capacity prediction model.

[0225] The technical solution of this implementation method, after obtaining the battery capacity label, extracts the operating condition data of multiple dimensions including battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, extracts the target operating condition indicator samples corresponding to the battery capacity label from the operating condition data, so that the samples can cover various operating conditions of the vehicle, and performs correlation analysis on the target operating condition indicator samples to screen out the target operating condition indicator samples that meet the correlation conditions to achieve sample dimensionality reduction, and then trains the initial battery capacity prediction model through the training data set to analyze the error influence shadow, so that the training data set can be subjected to sample denoising based on the error influence factor to obtain an updated training data set, train the initial battery capacity prediction model, and obtain the trained battery capacity prediction model, which effectively improves the model's capacity prediction capability and prediction accuracy for vehicles under various conditions.

[0226] In a possible implementation, the battery capacity tag acquisition module is used to acquire a battery capacity tag that meets a tag selection condition, where the tag selection condition includes at least one of the following conditions: The battery corresponding to the battery capacity label should be discharged to the preset low-end capacity threshold before charging; The battery corresponding to the battery capacity label should reach the preset full-charge threshold before charging ends; After the battery corresponding to the battery capacity label is charged, the battery rest time should reach the preset rest time threshold; The battery corresponding to the battery capacity label does not experience any preset impact event during the charging process; The distribution ratio of battery capacity labels meets the distribution ratio condition.

[0227] Through the technical solution of this implementation method, it is possible to effectively select battery capacity labels whose corresponding data are pure, accurate, certain, with little interference and reasonable distribution, thereby improving the prediction accuracy of the model.

[0228] In a possible implementation, the battery capacity tag acquisition module includes: The initial label set acquisition submodule is used to acquire the battery capacity label of the corresponding battery before charging, when the discharge reaches the preset capacity low-end threshold, and before the end of charging, the battery reaches the preset capacity full-charge threshold, and after the end of charging, the battery rest time reaches the preset rest time threshold, and no preset influencing event occurs during the charging process, so as to obtain the initial battery capacity label set; A proportion statistics submodule, used to count the distribution proportions of the initial battery capacity tag set in multiple capacity intervals; The proportion screening submodule is used to screen out the battery capacity labels that meet the distribution proportion conditions from the initial battery capacity labels according to the distribution proportion.

[0229] Through the technical solution of this implementation method, reliable battery capacity labels are first screened out, and then screened through distribution percentage statistics, which can improve the credibility and distribution rationality of samples in each capacity interval and avoid too many or too few interval samples that do not meet the prediction requirements.

[0230] In some embodiments, the training data set acquisition submodule includes: A first correlation calculation submodule, used to perform a first correlation analysis between the battery capacity tag and the corresponding target operating condition index sample, to obtain a first correlation analysis result between the battery capacity tag and the corresponding target operating condition index sample; The first correlation screening submodule is used to screen target operating condition indicator samples that meet a first correlation condition for the battery capacity tag based on the first correlation analysis result.

[0231] Through the technical solution of this implementation method, based on the correlation between the battery capacity label and the target operating condition indicator sample, it is possible to effectively exclude the operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity label, thereby improving the model training efficiency and model prediction accuracy.

[0232] In some embodiments, the training data set acquisition submodule includes: A second correlation calculation submodule, configured to perform a second correlation analysis between a plurality of target operating condition index samples corresponding to any of the battery capacity tags, and obtain a second correlation analysis result between the plurality of target operating condition index samples; The second correlation screening submodule is used to screen out target operating condition indicator samples that meet the second correlation condition for any of the battery capacity tags based on the second correlation analysis result.

[0233] Through the technical solution of this implementation method, it is possible to effectively eliminate redundant samples based on the correlation between multiple target operating condition indicator samples, reduce data complexity, and improve model training efficiency and model prediction accuracy.

[0234] In some embodiments, the relevance condition includes: The correlation between any of the battery capacity tags and the corresponding target operating condition indicator samples is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples corresponding to any of the battery capacity tags is less than or equal to a second preset correlation threshold.

[0235] The technical solution of this implementation method can screen out battery capacity labels and corresponding target operating condition indicator samples whose correlation is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples is less than or equal to a second preset correlation threshold, thereby effectively excluding operating condition indicator samples that are irrelevant or weakly correlated with the battery capacity labels, and eliminating redundant samples from multiple strongly correlated operating condition indicator samples, reducing data complexity, and improving model training efficiency and model prediction accuracy.

[0236] In some embodiments, the apparatus further comprises: A covariance matrix calculation module, used to calculate the covariance matrix corresponding to the set of target operating condition index samples; A feature contribution acquisition submodule, used to determine the feature contribution corresponding to each target operating condition index sample in the set of target operating condition index samples based on the covariance matrix; A removal submodule, configured to remove target operating condition index samples whose feature contributions do not meet the contribution condition from the set of target operating condition index samples based on the feature contributions corresponding to the respective target operating condition index samples, so as to obtain an updated target operating condition index sample set; The training set acquisition submodule is used to obtain the training data set based on the updated target operating condition index sample set.

[0237] The technical solution of this implementation method calculates the covariance matrix of the set of target operating condition index samples, decomposes the feature contribution, screens the target operating condition index samples that meet the contribution conditions based on the feature contribution, and eliminates the target operating condition index samples that do not meet the conditions, thereby improving the sample quality of the target operating condition index sample set, achieving sample dimensionality reduction, and effectively improving the model training efficiency and training effect.

[0238] In some embodiments, the error analysis submodule includes: An input submodule, used to input the target operating condition index samples in the training data set into the initial battery capacity prediction model to obtain the capacity prediction result output by the initial battery capacity prediction model; A calculation submodule, used to calculate the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation of the initial battery capacity prediction model based on the battery capacity label corresponding to the target operating condition index sample and the capacity prediction result; The analysis submodule is used to analyze the error influencing factors of the initial battery capacity prediction model based on the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation.

[0239] The technical solution of this implementation method analyzes the distribution of the mean square error, mean absolute error, determination coefficient and each prediction deviation to analyze the error influencing factor, and performs sample denoising on the training data set based on the error influencing factor, thereby reducing the error caused by abnormal samples, improving the model training effect, and obtaining a more optimized battery capacity prediction model.

[0240] Fig.15 FIG. 1 is a schematic structural block diagram of a battery capacity prediction device 1500 according to an embodiment of the present application. Fig.15 As shown, the battery capacity prediction device 1500 includes: A prediction request receiving module 1510 is used to receive a battery capacity prediction request and determine an operating condition indicator of a battery to be tested; A prediction result output module 1520, used to input the working condition index of the battery to be tested into a battery capacity prediction model to obtain a capacity prediction result output by the battery capacity prediction model; The battery capacity prediction model is pre-trained based on operating condition data, and the operating condition data includes operating condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage.

[0241] Through the technical solution of this implementation method, the battery capacity prediction model can effectively avoid the interference of error factors, and the working condition data of the pre-trained model also includes the vehicle battery consistency evaluation data and / or vehicle model, and the vehicle battery consistency or vehicle model has a great influence on the battery capacity. For example, if the vehicle battery consistency is relatively good, the charging and discharging of each battery in the vehicle is relatively balanced, and the overall battery capacity of the vehicle is high. If the vehicle battery consistency is relatively poor, the charging and discharging of each battery in the vehicle is relatively unbalanced, resulting in a low overall battery capacity of the vehicle. Different models have different algorithms and strategies for battery management, and the differences in algorithms and strategies will also have a great impact on the battery capacity. Therefore, the data used for training can cover various working conditions of the vehicle, learn the mapping relationship between different battery capacities and different working condition indicators, and then accurately predict the battery capacity based on the working condition indicators in the application, effectively improving the capacity prediction capability and prediction accuracy of the vehicle under various conditions.

[0242] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application, and are devices corresponding to the above-mentioned battery detection method. All implementation methods in the above-mentioned method embodiment are applicable to the embodiments of the device. Its specific functions and technical effects can be found in the method embodiment part, which will not be repeated here.

[0243] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0244] Fig.16 A schematic diagram of the hardware structure of an electronic device provided in yet another embodiment of the present application is shown.

[0245] The electronic device 1600 may include a processor 1601 and a memory 1602 storing programs or instructions. When the processor 1601 executes the program, the steps in any of the above method embodiments are implemented.

[0246] Exemplarily, the program may be divided into one or more modules / units, one or more modules / units are stored in the memory 1602, and executed by the processor 1601 to complete the present application. One or more modules / units may be a series of program instruction segments capable of completing a specific function, and the instruction segments are used to describe the execution process of the program in the device.

[0247] Specifically, the processor 1601 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0248] The memory 1602 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 1602 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. Where appropriate, the memory 1602 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 1602 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 1602 is a non-volatile solid-state memory.

[0249] The memory may include read-only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical or other physical / tangible memory storage devices. Thus, typically, the memory includes one or more tangible (non-transitory) readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to an aspect of the present disclosure.

[0250] The processor 1601 implements any one of the methods in the above embodiments by reading and executing the program or instruction stored in the memory 1602 .

[0251] In one example, the electronic device may further include a communication interface 1603 and a bus 1604. The processor 1601, the memory 1602, and the communication interface 1603 are connected via the bus 1604 and communicate with each other.

[0252] The communication interface 1603 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0253] Bus 1604 includes hardware, software or both, and couples the components of online data traffic billing equipment to each other. For example, but not limitation, the bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front-side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnect (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. Where appropriate, bus 1604 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the present application considers any suitable bus or interconnection.

[0254] In addition, in combination with the method in the above embodiment, the embodiment of the present application may provide a machine-readable storage medium for implementation. The machine-readable storage medium stores a program or instruction; when the program or instruction is executed by a processor, any one of the methods in the above embodiment is implemented. The machine-readable storage medium can be read by a machine such as a computer.

[0255] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0256] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0257] An embodiment of the present application provides a computer program product, which is stored in a machine-readable storage medium. The program product is executed by at least one processor to implement the various processes of the above-mentioned method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0258] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0259] The functional modules shown in the structural block diagram described above can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier. "Machine-readable medium" may include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROMs, flash memories, erasable ROMs (EROMs), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer grid such as the Internet, an intranet, etc.

[0260] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0261] Aspects of the present disclosure are described above with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by a computer program or instruction. These programs or instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the functions / actions specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0262] Although the present application has been described with reference to preferred embodiments, various modifications may be made thereto and parts thereof may be replaced with equivalents without departing from the scope of the present application. In particular, the various technical features mentioned in the various embodiments may be combined in any manner as long as there is no structural conflict. The present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A battery capacity prediction model training method, characterized in that: include: Get the battery capacity label; Acquire the operating condition data of the battery operating condition and the vehicle operating condition corresponding to the battery capacity label, wherein the operating condition data includes: operating condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle cumulative mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model; Based on the operating condition data, a battery capacity prediction model is trained.

2. The method according to claim 1, characterized in that The training of the battery capacity prediction model based on the operating condition data includes: Extracting a target operating condition index sample corresponding to the battery capacity tag from the operating condition data; By performing correlation analysis on the battery capacity labels and the corresponding target operating condition indicator samples, target operating condition indicator samples that meet the correlation conditions are screened out to obtain a training data set; Using the training data set, an initial battery capacity prediction model is trained, and an error influencing factor of the initial battery capacity prediction model is analyzed, wherein the battery capacity prediction model is used to predict the battery capacity; Based on the error influencing factor, performing sample denoising on the training data set to obtain an updated training data set; Based on the updated training data set, the initial battery capacity prediction model is trained to obtain a trained battery capacity prediction model.

3. The method according to any one of claims 1 or 2, characterized in that: The obtaining of the battery capacity tag includes: obtaining a battery capacity tag that satisfies a tag selection condition, wherein the tag selection condition includes at least one of the following conditions: The battery corresponding to the battery capacity label should be discharged to the preset low-end capacity threshold before charging; The battery corresponding to the battery capacity label should reach the preset full-charge threshold before charging ends; After the battery corresponding to the battery capacity label is charged, the battery rest time should reach the preset rest time threshold; The battery corresponding to the battery capacity label does not experience any preset impact event during the charging process; The distribution ratio of battery capacity labels meets the distribution ratio condition.

4. The method according to claim 3, characterized in that The step of obtaining a battery capacity tag that meets a tag selection condition includes: Obtain the battery capacity label of the corresponding battery before charging, when the discharge reaches the preset capacity low-end threshold, and before the end of charging, the battery reaches the preset capacity full-charge threshold, and after the end of charging, the battery rest time reaches the preset rest time threshold, and no preset influencing event occurs during the charging process of the battery, to obtain an initial battery capacity label set; Counting the distribution ratio of the initial battery capacity tag set in multiple capacity intervals; According to the distribution ratio, the battery capacity labels that meet the distribution ratio condition are screened out from the initial battery capacity labels.

5. The method according to claim 2, characterized in that: The method of performing correlation analysis on the battery capacity label and the corresponding target operating condition indicator sample to screen out the target operating condition indicator sample that meets the correlation condition includes: Performing a first correlation analysis between the battery capacity tag and the corresponding target operating condition index sample to obtain a first correlation analysis result between the battery capacity tag and the corresponding target operating condition index sample; Based on the first correlation analysis result, target operating condition indicator samples that meet the first correlation condition are screened out for the battery capacity label.

6. The method according to claim 2 or 5, characterized in that: The method of performing correlation analysis on the battery capacity label and the corresponding target operating condition indicator sample to screen out the target operating condition indicator sample that meets the correlation condition includes: Performing a second correlation analysis between a plurality of target operating condition indicator samples corresponding to any of the battery capacity tags to obtain a second correlation analysis result between the plurality of target operating condition indicator samples; Based on the second correlation analysis result, a target operating condition indicator sample that meets the second correlation condition is screened out for any of the battery capacity tags.

7. The method according to claim 2 or 5, characterized in that: The relevance conditions include: The correlation between any of the battery capacity tags and the corresponding target operating condition indicator samples is greater than or equal to a first preset correlation threshold, and the correlation between multiple target operating condition indicator samples corresponding to any of the battery capacity tags is less than or equal to a second preset correlation threshold.

8. The method according to claim 2 or 5, characterized in that: After selecting the target operating condition index samples that meet the correlation condition, the method further includes: Calculating the covariance matrix corresponding to the set of target operating condition index samples; Determining, based on the covariance matrix, a feature contribution corresponding to each target operating condition index sample in the set of target operating condition index samples; Based on the characteristic contribution corresponding to each target operating condition index sample, the target operating condition index samples whose characteristic contribution does not meet the contribution condition are removed from the set of target operating condition index samples to obtain an updated target operating condition index sample set; The training data set is obtained based on the updated target operating condition index sample set.

9. The method according to claim 2 or 5, characterized in that: The step of training the initial battery capacity prediction model by using the training data set and analyzing the error influencing factors of the initial battery capacity prediction model includes: Inputting the target operating condition index samples in the training data set into an initial battery capacity prediction model to obtain a capacity prediction result output by the initial battery capacity prediction model; Based on the battery capacity label corresponding to the target operating condition index sample and the capacity prediction result, calculating the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation of the initial battery capacity prediction model; Based on the mean square error, mean absolute error, determination coefficient and distribution of each prediction deviation, the error influencing factor of the initial battery capacity prediction model is analyzed.

10. A battery capacity prediction method, characterized in that: include: Receive a battery capacity prediction request and determine the operating condition index of the battery to be tested; Inputting the operating condition index of the battery to be tested into a battery capacity prediction model to obtain a capacity prediction result output by the battery capacity prediction model; Among them, the battery capacity prediction model is pre-trained based on operating condition data, and the operating condition data includes: operating condition data corresponding to multiple dimensions in battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model.

11. A battery capacity prediction model training device, characterized in that: include: A battery capacity label acquisition module is used to acquire the battery capacity label; A working condition data acquisition module, used to acquire working condition data of the battery working condition and the vehicle working condition corresponding to the battery capacity label, wherein the working condition data includes: working condition data corresponding to multiple dimensions of battery storage, battery cycle, battery temperature, battery state of charge and vehicle cumulative mileage, and the working condition data also includes: vehicle battery consistency assessment data and / or vehicle model; The model training module is used to train the battery capacity prediction model based on the operating condition data.

12. A battery capacity prediction device, characterized in that: include: A prediction request receiving module, used to receive a battery capacity prediction request and determine the working condition index of the battery to be tested; A prediction result output module, used for inputting the working condition index of the battery to be tested into a battery capacity prediction model to obtain a capacity prediction result output by the battery capacity prediction model; Among them, the battery capacity prediction model is pre-trained based on operating condition data, and the operating condition data includes: operating condition data corresponding to multiple dimensions in battery storage, battery cycle, battery temperature, battery state of charge and vehicle accumulated mileage, and the operating condition data also includes: vehicle battery consistency assessment data and / or vehicle model.

13. An electronic device, characterized in that: It includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the battery capacity prediction model training method as described in any one of claims 1 to 9, or implements the steps of the battery capacity prediction method as described in claim 10.

14. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, it implements the battery capacity prediction model training method as described in any one of claims 1 to 9, or implements the steps of the battery capacity prediction method as described in claim 10.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the battery capacity prediction model training method as described in any one of claims 1 to 9 is implemented, or the steps of the battery capacity prediction method as described in claim 10 are implemented.

Images