Training method of battery life prediction model, battery life prediction method and equipment

Abstract

The embodiment of the invention provides a training method of a battery life prediction model, a battery life prediction method and equipment. According to the method, static sample data and dynamic sample data of a sample battery are obtained to construct a sample data set, the sample data set is adopted to train a prediction model to obtain a battery life prediction model, and then battery data of a target battery are input into the battery life prediction model to obtain the predicted life of the target battery. Wherein the static sample data comprises the solid state attribute of the sample battery, and the dynamic sample data comprises the operation data of the sample battery under various working conditions, so that the battery life prediction model is trained through the method, and the battery life prediction model can capture the influence of the battery type and the various working conditions on the battery life; the generalization ability of the prediction model is improved, so that the battery life prediction model is adopted to perform life prediction on different types of batteries under various working conditions, and the accuracy of battery life prediction is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and particularly to a method for training a battery life prediction model, an electronic device, and a storage medium. Background Art

[0002] As a core energy storage device, batteries are widely used in energy storage power stations, electric vehicles, and various mechanical and electronic devices. With the increase in the number of charge and discharge cycles, the capacity of the battery decreases, and the internal resistance increases or decreases, resulting in a reduction in battery life. In order to make reasonable plans during the use of the battery to reduce the loss of battery life and ensure the safe and reliable operation of the battery, the prior art usually adopts model-based methods and data-driven methods to estimate the battery life.

[0003] Among them, the model-based method establishes a non-linear system of battery life attenuation through a physical model or an empirical model. This method requires a large number of cross-calibration experiments, measures data under laboratory operating conditions, and calibrates the attenuation curve through complex modeling, which is time-consuming and computationally intensive. In addition, it is difficult for the experimental test environment to cover the complex working conditions in actual application scenarios. In addition, the data-driven method mines the relationship between historical measurement data and battery life through statistics or machine learning to indirectly estimate the battery life. However, the existing data-driven methods are usually designed for specific types of batteries and cannot predict the battery life with high accuracy for multiple types and multiple usage scenarios.

[0004] Therefore, the existing battery life estimation methods have the problem of low battery life prediction accuracy for multiple types of batteries under multiple working conditions. Summary of the Invention

[0005] Embodiments of the present application provide a method for training a battery life prediction model, a battery life prediction method, and a device, so as to achieve the effect of predicting the life of multiple types of batteries under multiple working conditions and obtaining accurate predicted battery life.

[0006] In a first aspect, an embodiment of the present application provides a method for training a battery life prediction model, including:

[0007] Obtain static sample data and dynamic sample data of a sample battery, where the static sample data is used to indicate the solid state attributes of the sample battery, and the dynamic sample data is used to indicate the operating data of the sample battery under multiple working conditions;

[0008] Determine a sample data set based on the static sample data and the dynamic sample data;

[0009] Use the sample data set to perform training processing on a prediction model to obtain a trained battery life prediction model.

[0010] In a possible implementation, the static sample data includes: rated capacity, and the dynamic sample data includes: currents under various working conditions, discharge durations corresponding to the currents, initial state-of-charge parameters corresponding to the current discharge cycle, and current state-of-charge parameters corresponding to the discharge durations. Determining a sample data set based on the static sample data and the dynamic sample data includes:

[0011] Determining the discharged electricity based on the discharge current and the corresponding discharge duration in the dynamic sample data;

[0012] Determining the change amount of the state-of-charge parameter of the sample battery according to the initial state-of-charge parameter and the current state-of-charge parameter;

[0013] Determining the current capacity of the sample battery based on the discharged electricity and the change amount of the state-of-charge parameter, and taking the ratio of the current capacity to the rated capacity as the actual health degree;

[0014] Taking the static sample data, the dynamic sample data, and the actual health degree as the sample data set.

[0015] In a possible implementation, the prediction model includes a dynamic characterization layer and a static characterization layer. Training the prediction model using the sample data set to obtain a trained battery life prediction model includes:

[0016] Performing feature extraction processing on the dynamic sample data through the dynamic characterization layer to obtain multiple first features of the sample battery, and determining a first life feature based on the correlation between the multiple first features and the battery health degree;

[0017] Determining a second life feature through the static characterization layer based on the correlation between the static sample data and the battery health degree;

[0018] Determining the predicted health degree of the sample battery based on the first life feature and the second life feature;

[0019] Determining the prediction accuracy corresponding to the sample battery based on the predicted health degree and the actual health degree;

[0020] In the case where the prediction accuracy does not reach the accuracy threshold, optimizing the prediction model based on the prediction accuracy and re-determining the new predicted health degree until the prediction accuracy reaches the accuracy threshold or the number of training times reaches the upper limit value to obtain the battery life prediction model.

[0021] In a possible implementation, the static sample data further includes: battery type and usage scenario. Optimizing the prediction model based on the prediction accuracy includes:

[0022] Adjusting the dynamic representation parameters corresponding to the dynamic representation layer and the static representation parameters corresponding to the static representation layer according to the prediction accuracy, where the dynamic representation parameters are shared by multiple battery types and usage scenarios, and the static representation parameters correspond to specific battery types and / or usage scenarios.

[0023] In a second aspect, a battery life prediction method includes:

[0024] Obtaining battery data of a target battery, where the battery data is used to indicate the solid-state parameters of the battery and the operating data under the current working condition;

[0025] Inputting the battery data into a battery life prediction model to obtain the predicted life output by the battery life prediction model;

[0026] Wherein, the battery life prediction model is trained by the training method of the battery life prediction model as described in the first aspect above.

[0027] In a possible implementation, the step of inputting the battery data into the battery life prediction model to obtain the predicted life output by the battery life prediction model includes:

[0028] According to the solid-state parameters in the battery data, calling the static representation parameters corresponding to the battery type and / or usage scenario in the solid-state parameters;

[0029] Based on the operating data, determining the third life characteristic of the target battery under the current working condition;

[0030] Based on the static representation parameters and the solid-state parameters, determining the fourth life characteristic of the target battery;

[0031] Performing fusion processing on the third life characteristic and the fourth life characteristic to obtain the predicted health degree of the target battery, and determining the remaining life of the target battery based on the predicted health degree.

[0032] In a possible implementation, the method further includes:

[0033] Real-time monitoring the predicted life of the target battery. When the predicted life reaches a preset threshold, generating a warning message and sending the warning message to a user terminal, where the warning message includes the predicted life.

[0034] In a third aspect, an embodiment of the present application provides a training device for a battery life prediction model, including:

[0035] A first acquisition module, configured to acquire static sample data and dynamic sample data of a sample battery, where the static sample data is used to indicate the solid state attributes of the sample battery, and the dynamic sample data is used to indicate the operation data of the sample battery under multiple working conditions;

[0036] A determination module, configured to determine a sample data set based on the static sample data and the dynamic sample data;

[0037] A training module, configured to use the sample data set to perform training processing on a prediction model to obtain a trained battery life prediction model.

[0038] In a possible implementation manner, the static sample data includes: rated capacity, and the dynamic sample data includes: currents under multiple working conditions, discharge durations corresponding to the currents, initial state of charge parameters corresponding to the current discharge cycle, and current state of charge parameters corresponding to the discharge durations. The determination module is further configured to determine a discharge power based on the discharge current and the corresponding discharge duration in the dynamic sample data;

[0039] The determination module is further configured to determine a change amount of the state of charge parameter of the sample battery according to the initial state of charge parameter and the current state of charge parameter;

[0040] The determination module is further configured to determine the current capacity of the sample battery based on the discharge power and the change amount of the state of charge parameter, and use the ratio of the current capacity to the rated capacity as the actual health degree;

[0041] The determination module is further configured to use the static sample data, the dynamic sample data, and the actual health degree as the sample data set.

[0042] In a possible implementation manner, the prediction model includes a dynamic characterization layer and a static characterization layer, and the device further includes: a feature extraction module and an optimization module;

[0043] The feature extraction module is configured to perform feature extraction processing on the dynamic sample data through the dynamic characterization layer to obtain multiple first features of the sample battery;

[0044] The determination module is further configured to determine a first life feature based on the association relationship between the multiple first features and the battery health degree;

[0045] The determination module is further configured to determine a second life feature through the static characterization layer based on the association relationship between the static sample data and the battery health degree;

[0046] The determining module is further configured to determine the predicted health degree of the sample battery based on the first life characteristic and the second life characteristic;

[0047] The determining module is further configured to determine the prediction accuracy corresponding to the sample battery based on the predicted health degree and the actual health degree;

[0048] The optimizing module is configured to, when the prediction accuracy does not reach the accuracy threshold, optimize the prediction model based on the prediction accuracy, and re-determine the new predicted health degree until the prediction accuracy reaches the accuracy threshold or the number of training times reaches the upper limit value, so as to obtain the battery life prediction model.

[0049] In a possible implementation manner, the static sample data further includes: battery type and usage scenario, and the device further includes: an adjustment module;

[0050] The adjustment module is configured to adjust the dynamic representation parameters corresponding to the dynamic representation layer and the static representation parameters corresponding to the static representation layer according to the prediction accuracy, where the dynamic representation parameters are shared by multiple battery types and usage scenarios, and the static representation parameters correspond to specific battery types and / or usage scenarios.

[0051] In a fourth aspect, an embodiment of the present application provides a battery life prediction device, including:

[0052] A second acquisition module, configured to acquire battery data of a target battery, where the battery data is used to indicate the solid-state parameters of the battery and the operation data under the current working condition;

[0053] An input module, configured to input the battery data into a battery life prediction model to obtain a predicted life output by the battery life prediction model;

[0054] Wherein, the battery life prediction model is trained by the battery life prediction model training method as described in the first aspect above.

[0055] In a possible implementation manner, the device further includes: a calling module, a determining module, and a processing module;

[0056] The calling module is configured to call the static representation parameters corresponding to the battery type and / or usage scenario in the solid-state parameters according to the solid-state parameters in the battery data;

[0057] The determining module is configured to determine a third life characteristic of the target battery under the current working condition based on the operation data;

[0058] The determining module is further configured to determine a fourth life characteristic of the target battery based on the static characterization parameter and the solid state parameter;

[0059] The processing module is configured to perform a fusion process on the third life characteristic and the fourth life characteristic to obtain a predicted health degree of the target battery;

[0060] The determining module is further configured to determine a remaining life of the target battery based on the predicted health degree.

[0061] In a fifth aspect, an embodiment of the present application provides an electronic device, including: a memory and a processor;

[0062] The memory stores computer execution instructions;

[0063] The processor executes the computer execution instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementation manners of the first aspect, and the above first aspect and / or various possible implementation manners of the first aspect.

[0064] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer execution instructions are stored, and when the computer execution instructions are executed by a processor, they are used to implement the above first aspect and / or various possible implementation manners of the first aspect, and the above first aspect and / or various possible implementation manners of the first aspect.

[0065] In a seventh aspect, an embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the above first aspect and / or various possible implementation manners of the first aspect, and the above first aspect and / or various possible implementation manners of the first aspect.

[0066] The training method, battery life prediction method, and device for a battery life prediction model provided by an embodiment of the present application obtain static sample data and dynamic sample data of a sample battery, determine a sample data set based on the above static sample data and dynamic sample data, and use the sample data set to perform a training process on the prediction model to obtain a trained battery life prediction model. Then, the battery data of a target battery is input into the battery life prediction model to obtain the predicted life of the target battery. Among them, the static sample data is used to indicate the solid state attributes of the sample battery, and the dynamic sample data is used to indicate the operating data of the sample battery under various working conditions. By training the battery life prediction model with the above static sample data and dynamic sample data, the battery life prediction model can capture the influence of battery types and various working conditions on battery life, improving the generalization ability of the prediction model. Therefore, using the above battery life prediction model to predict the life of different types of batteries under various working conditions improves the accuracy of the predicted battery life. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings are incorporated herein and form a part of this specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0068] Figure 1 Schematic flow chart of a training method for a battery life prediction model provided by the present application Figure 1 ;

[0069] Figure 2 Schematic flow chart of a training method for a battery life prediction model provided by the present application Figure 2 ;

[0070] Figure 3 Schematic flow chart of a battery life prediction method provided by the present application;

[0071] Figure 4 Schematic structural diagram of a training device for a battery life prediction model provided by the present application;

[0072] Figure 5 Schematic structural diagram of a battery life prediction device provided by the present application;

[0073] Figure 6 Schematic structural diagram of an electronic device provided by the present application.

[0074] Through the above accompanying drawings, specific embodiments of the present application have been shown, and there will be more detailed descriptions hereinafter. These accompanying drawings and the written description are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0075] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0076] With the development of new energy technologies, batteries are widely used in various scenarios such as electric vehicles and energy storage power stations. The service life of a battery directly affects the operating efficiency and stability of electronic devices. Therefore, accurately predicting the battery life can not only effectively improve the safety guarantee during the operation of the device but also ensure the maximization of the utilization of battery resources.

[0077] Existing battery life prediction methods are mainly divided into model-based methods and data-driven methods. Among them, model-based methods, for example, predict battery life based on empirical models or electrochemical models. However, the method using an empirical model predicts the battery life by statistically analyzing the usage and life of the battery and establishing a mathematical model. This method requires a large amount of battery operation data, and there are certain errors in the prediction results obtained based on statistical data. In addition, the prediction method using an electrochemical model has high requirements for computing resources and algorithms, and the experimental test environment cannot cover various working conditions in actual application scenarios. Therefore, it is impossible to accurately estimate the battery life state for various types of batteries under various working conditions. In addition, data-driven methods mine the relationship between historical measurement data and battery life through statistical or machine learning methods to predict the remaining battery life; however, existing data-driven methods are usually designed for specific types of batteries and have low prediction accuracy for different types of batteries under different working conditions.

[0078] Therefore, existing battery life prediction methods have the problem of low battery life prediction accuracy for various types of batteries under various working conditions.

[0079] Based on the above technical problems, the training method of the battery life prediction model provided in the present application trains a prediction model through static sample data and dynamic sample data to obtain a battery life prediction model, and then inputs the battery data of the target battery into the above battery life prediction model to obtain the predicted life of the above target battery. This method trains the model based on the above static sample data and dynamic sample data, enabling the obtained battery life prediction model to capture the influence of battery types and various working conditions on battery life, improving the generalization ability of the prediction model, and further improving the accuracy of the battery life prediction model in predicting the life of various types of batteries under various working conditions.

[0080] The following uses specific embodiments to elaborate in detail on the technical solution of the present application and how the technical solution of the present application solves the above technical problems. These several specific embodiments below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0081] Figure 1 Flow schematic of a method for training a battery life prediction model provided by the present application Figure 1 , such as Figure 1 shown, the method includes:

[0082] S101. Obtain static sample data and dynamic sample data of the sample battery.

[0083] Among them, the static sample data is used to indicate the solid state attributes of the sample battery, and the dynamic sample data is used to indicate the operating data of the above sample battery under various working conditions. Exemplarily, the above static sample data includes: battery model, chemical composition, usage scenario, rated capacity, rated power, etc.; the above dynamic sample data includes: battery external characteristic data and battery internal state data. The battery external characteristic data can be, for example, the voltage, current, and temperature of the sample battery under different working conditions, and the battery internal state data can be, for example, the charge parameter (State of Charge, SOC) of the sample battery at different times.

[0084] Optionally, this method is applied to the battery management system. During the operation of the above sample battery, the operation data of the sample battery is collected at a preset cycle to obtain the above dynamic sample data. The dynamic sample data and the above static sample data can be stored in the database of the battery management system or stored in the cloud.

[0085] It can be understood that the above charge parameter can be estimated by an algorithm built into the battery management system, and the algorithm includes but is not limited to the ampere-hour integration method.

[0086] S102. Determine a sample data set based on the static sample data and the dynamic sample data.

[0087] Specifically, based on the static sample data and the dynamic sample data, determine the actual life labels corresponding to multiple sample batteries, and use the actual life labels, the above-mentioned static sample data, and the dynamic sample data as a sample data set. Among them, the actual life label is the actual state of health (SOH) of the sample battery. In addition, the sample data set can be stored in the battery management system in the form of a table. In this method, the static sample data characterizes the inherent properties of the sample battery. Combining the static sample data with the above-mentioned actual life label, the dynamic sample data includes the operation data of the battery under different working conditions. Using the static sample data and the dynamic sample data as the training data set is conducive to capturing the life attenuation characteristics of specific types of batteries under specific usage scenarios, as well as the influence of different working conditions on the life of multiple types of batteries.

[0088] It can be understood that before determining the above-mentioned actual life label, in order to ensure the integrity and accuracy of the training data, preprocess the above-mentioned static sample data and dynamic sample data to remove duplicate data, handle missing values, and enhance data consistency. Among them, the preprocessing methods include but are not limited to noise processing, missing value processing, and normalization processing.

[0089] S103. Use the sample data set to perform training processing on the prediction model to obtain a trained battery life prediction model.

[0090] Specifically, the above-mentioned prediction model includes a static representation layer, a dynamic representation layer, and a fusion representation layer. Input the sample data set into the above-mentioned prediction model. The dynamic representation layer extracts features from the dynamic sample data and constructs a first life feature based on the extracted dynamic features. In addition, the static representation layer maps the static sample data into corresponding static features, and constructs a second life feature based on the correlation between the static features and the actual life labels in the sample data set. Then, through the above-mentioned fusion representation layer, fuse the first life feature and the second life feature, and determine the predicted life of the above-mentioned sample battery based on the fused third life feature. Furthermore, fit the predicted life and the above-mentioned actual life label to adjust the model parameters of the prediction model to obtain the above-mentioned battery life prediction model.

[0091] This method trains the prediction model based on the sample data set constructed by the static sample data and the dynamic sample data, so that the obtained battery life prediction model can predict the life of multiple types of batteries under different working conditions. And this battery life prediction model captures both the solid state properties of the sample battery and the influence of the operation data under different working conditions on the battery life, which is conducive to improving the prediction accuracy of the battery life.

[0092] The training method of the battery life prediction model provided by the embodiments of the present application obtains the static sample data and dynamic sample data of the sample battery, determines the sample data set based on the above static sample data and dynamic sample data, and then uses the sample data set to perform training processing on the prediction model to obtain the trained battery life prediction model. Among them, the static sample data is used to indicate the solid state attributes of the sample battery, and the dynamic sample data is used to indicate the operating data of the sample battery under various working conditions. This method trains the battery life prediction model through the above static sample data and dynamic sample data, enabling the battery life prediction model to capture the influence of battery types and various working conditions on battery life, improving the generalization ability of the prediction model, and further improving the accuracy of the battery life prediction model in predicting the life of various types of batteries under various working conditions.

[0093] Figure 2 Schematic flow of a training method for a battery life prediction model provided by the present application Figure 2 , such as Figure 2 shown. Based on the Figure 1 embodiment, a possible implementation of the training method for the battery life prediction model is described in detail. The method includes:

[0094] S201. Obtain the static sample data and dynamic sample data of the sample battery.

[0095] The explanation of this step S201 is similar to that of the above step S101 and will not be elaborated here.

[0096] S202. Determine the discharge capacity based on the discharge current and the corresponding discharge duration in the dynamic sample data.

[0097] Among them, the dynamic sample data includes: the current under various working conditions, the discharge duration corresponding to the current, the initial state of charge parameter corresponding to the current discharge cycle, and the current state of charge parameter corresponding to the above discharge duration.

[0098] Specifically, based on the above current discharge duration, integrate the discharge current within this duration to obtain the discharge capacity within the current discharge duration. Exemplarily, the following formula is used to determine the above discharge capacity:

[0099] Q = ∫ T I(t)dt

[0100] where Q is the above discharge capacity, T is the above discharge duration, and I(t) is the discharge current of the sample battery at time t.

[0101] Optionally, the above dynamic sample data may further include operating parameters such as the number of charge and discharge cycles, the cumulative charge and discharge amount, and the cumulative static duration of the sample battery. The above operating parameters such as the number of charge and discharge cycles, the cumulative charge and discharge amount, and the cumulative static duration are used to assist in the feature extraction of the dynamic characterization layer, which is beneficial for the prediction model to obtain more accurate prediction results.

[0102] S203. Determine the change amount of the charge parameter of the sample battery according to the initial charge parameter and the current charge parameter.

[0103] Among them, the initial charge parameter is used to characterize the initial charge state corresponding to the current discharge cycle, and the current charge parameter is used to characterize the charge state of the sample battery at the end of the current discharge cycle.

[0104] Optionally, the above charge parameter is the SOC of the battery. The change amount of the above charge parameter can be determined according to the following formula:

[0105] ΔSOC = SOC T, start - SOC T, end

[0106] Among them, SOC T,开始 is the initial charge parameter of the current discharge cycle, SOC T,结束 is the current charge parameter corresponding to the sample battery at the end of the current discharge cycle, and ΔSOC is the change amount of the charge parameter corresponding to the current discharge cycle.

[0107] It can be understood that the SOC at a certain moment can be estimated based on theoretical model parameters such as the cell characteristic parameter model and the cell rate characteristic model.

[0108] S204. Determine the current capacity of the sample battery based on the discharge amount and the change amount of the charge parameter, and use the ratio of the current capacity to the rated capacity as the actual health degree.

[0109] Among them, the static sample data includes the rated capacity of the sample battery.

[0110] Specifically, after obtaining the discharge amount and the change amount of the charge parameter corresponding to the current discharge cycle, the ratio of the above discharge amount to the change amount of the charge parameter is determined as the above current capacity, and then the ratio of the current capacity to the rated capacity of the sample battery is determined, and this ratio is determined as the actual health degree.

[0111] Optionally, the actual health degree can be SOH for example. The actual health degree of the sample battery is determined by the following formula:

[0112]

[0113] Wherein, SOH is the actual battery health degree mentioned above, Q is the current capacity of the sample battery obtained in the above step S202, ΔSOC is the change amount of the charge parameter obtained in the above step S203, and C is the rated capacity of the sample battery.

[0114] S205. Use the static sample data, the dynamic sample data, and the actual health degree as the sample data set.

[0115] Specifically, integrate the static sample data, the dynamic sample data, and the actual health degree to construct the above sample data set. The sample data set includes the acquisition time of the dynamic sample data, that is, each piece of sample data in the sample data set includes time, static sample data, dynamic sample data, and actual health degree. Among them, the actual health degree can be characterized by the SOH of the battery, for example.

[0116] This sample data set not only contains the inherent attributes of the battery, but also includes various dynamic parameters during the operation of the battery, and constructs the correlation relationship between the above inherent attributes, various dynamic parameters and the actual health degree, which more comprehensively reflects the influence of various inherent attributes and dynamic parameters of the battery on the battery life.

[0117] S206. Through the dynamic characterization layer, perform feature extraction processing on the dynamic sample data to obtain multiple first features of the sample battery, and determine the first life feature based on the correlation relationship between the multiple first features and the battery health degree.

[0118] Specifically, the above prediction model includes a dynamic characterization layer and a static characterization layer. Input the sample data set in the above step S205 into the prediction model, and extract multiple first features of the sample battery from the dynamic sample data carrying the time series through the dynamic characterization layer of the prediction model. Among them, the first features are, for example, voltage change features, current change features, and temperature change features. The first features can reflect the dynamic behavior of the battery during use from different angles. After obtaining the above multiple first features, construct the first life feature according to the correlation relationship between the multiple first features and the battery health degree. The first life feature can, for example, characterize the influence of temperature change on the battery health degree, that is, as the temperature rises, the health degree of the battery decreases.

[0119] It should be noted that the dynamic characterization layer is a neural network architecture, and the specific structure of this neural network architecture is not limited in this application. In addition, the network parameters of the dynamic characterization layer correspond to multiple types of batteries in multiple scenarios.

[0120] The method trains the dynamic representation layer in the prediction model through a sample data set, so that the dynamic representation layer outputs the first life characteristics of the sample battery, and the first life characteristics are used to characterize the battery life characteristics under the current working conditions. That is, the method enables the dynamic representation layer to obtain the influence of different working conditions on the battery life.

[0121] S207. Determine the second life characteristics through the static representation layer based on the correlation between the static sample data and the battery health degree.

[0122] Among them, the static representation layer constructs corresponding battery life characteristics based on the static sample data of the sample battery. Specifically, based on the static sample data, positive sample pairs and negative sample pairs are constructed. Among them, the positive sample pairs are battery pairs with similar models, chemical compositions, and usage scenarios, and the negative sample pairs are battery pairs with different models, chemical compositions, and usage scenarios. Then, by reducing the static representation distance between the batteries in the positive sample pairs and increasing the static representation distance between the batteries in the negative sample pairs, a contrast loss function is constructed, and through this contrast loss function, the similarity between the same type of samples is maximized, and the similarity between different types of samples is minimized, so that the training objective includes the representational discrimination between different types of batteries, and further enables the finally obtained second life characteristics to reflect the life attenuation characteristics of specific types of batteries.

[0123] Among them, the loss function can be, for example, a triplet loss. Here, a possible method for determining the loss function is provided. The method includes: in the sample data set, using the first sample battery as a reference point to determine the corresponding static representation distance, and the static representation distance determines the loss function. Optionally, the above loss function is determined by the following formula:

[0124] L triplet = max(‖v a - v p ‖ 2 -‖v a - v n ‖ 2 + α, 0)

[0125] Among them, v a is the static representation vector of the first sample battery, v p is the static representation vector of the second sample battery with the same type as the first sample battery, v n is the static representation vector of the third sample battery with the same type as the first sample battery. Then, ‖v a - v p ‖ 2 is the static representation distance between the first sample battery and the positive sample, and ‖v a - v n ‖ 2is the static characterization distance between the first sample battery and the negative sample, and α is the boundary hyperparameter, which is the minimum difference controlling the distance between the positive and negative samples.

[0126] Understandably, different types of batteries in the static characterization layer correspond to different static characterization parameters.

[0127] S208. Determine the predicted health degree of the sample battery based on the first life characteristic and the second life characteristic.

[0128] Specifically, the above prediction model further includes a fusion characterization layer. Through this fusion characterization layer, the above first life characteristic and second life characteristic are fused, and the fused third life characteristic is input into the fully connected layer of the neural network for processing to obtain the predicted health degree of the above sample battery. Among them, the above fully connected layer of the neural network can be, for example, an MLP (Multilayer Perceptron).

[0129] This method combines the above first life characteristic and second life characteristic, taking into account both the influence of the current working condition on the battery life and the influence of the solid state property of the battery on the battery life, making the accuracy of the obtained predicted health degree higher.

[0130] S209. Determine the prediction accuracy corresponding to the sample battery based on the predicted health degree and the actual health degree.

[0131] Specifically, according to the predicted health degree and the actual health degree of the sample battery, the ratio of the predicted health degree to the actual health degree is used as the prediction accuracy, and it is judged whether the prediction accuracy reaches a preset accuracy threshold. When the prediction accuracy reaches the above accuracy threshold, the trained prediction model is used as the battery life prediction model. In addition, when the above prediction accuracy does not reach the accuracy threshold, the prediction model is trained according to the method shown in the following step S210 to obtain the above battery life prediction model.

[0132] S210. When the prediction accuracy does not reach the accuracy threshold, optimize the prediction model based on the prediction accuracy and re-determine the new predicted health degree until the prediction accuracy reaches the accuracy threshold or the number of training times reaches the upper limit value to obtain the battery life prediction model.

[0133] Specifically, in the case where the prediction accuracy fails to reach the accuracy threshold, based on the above prediction accuracy, the model parameters are adjusted, and the prediction model after parameter adjustment is trained based on the sample data set until the obtained prediction accuracy reaches the accuracy threshold, or the number of training times reaches the upper limit value. After the training is completed, the trained prediction model is determined as the battery life prediction model, and the corresponding model parameters are saved. Among them, the model parameters include the static representation parameters corresponding to the static representation layer and the static representation parameters corresponding to the dynamic representation layer.

[0134] Optionally, the static sample data further includes: battery type and usage scenario. Here, a possible method for optimizing the above static representation parameters and dynamic representation parameters based on the above prediction accuracy is given. The method includes: adjusting the dynamic representation parameters corresponding to the dynamic representation layer and the static representation parameters corresponding to the static representation layer according to the prediction accuracy, where the dynamic representation parameters are shared by multiple battery types and usage scenarios, and the static representation parameters correspond to specific battery types and / or usage scenarios.

[0135] The training method of the battery life prediction model provided by the embodiments of the present application obtains the static sample data and dynamic sample data of the sample battery, and calculates the actual health degree corresponding to each sample data based on the static sample data and the dynamic sample data. Among them, the dynamic sample data carries the acquisition time. Therefore, the sample data set constructed based on the static sample data, the dynamic sample data, and the above actual health degree is a time series data set. After obtaining the sample data set, the sample data set is used to train the prediction model. The dynamic sample data is subjected to feature extraction through the dynamic representation layer of the prediction model, and based on the correlation between the extracted first feature and the battery health degree, a first life feature is determined. The first life feature is used to characterize the influence of different working conditions on the life of the sample battery. In addition, through the static representation layer of the prediction model, the static sample data is subjected to feature extraction, and based on the correlation between the extracted feature and the battery health degree, a second life feature is constructed. The second life feature is used to characterize the influence of the solid state attributes of the battery on its life, where the above solid state attributes include battery type and usage scenario. Then, the above first life feature and second life feature are subjected to feature fusion, and the predicted health degree of the sample battery is determined based on the fused features. Furthermore, based on the predicted health degree and the actual health degree, the prediction accuracy of the prediction model is determined. When the prediction accuracy does not reach the accuracy threshold, the model parameters of the prediction model are adjusted based on the above prediction accuracy until the prediction accuracy of the prediction model reaches the above accuracy threshold, or the number of training times reaches the upper limit value. The trained prediction model is used as the above battery life prediction model. This method trains the prediction model through a sample data set with time series, so that the obtained battery life prediction model can, on the one hand, predict the health degree of each type of battery under different working conditions, and on the other hand, combine the influence of different working conditions on the battery life and the influence of solid state attributes such as battery type and usage scenario on the battery life, improving the accuracy of predicting the health degree.

[0136] Figure 3 is a schematic flow chart of a battery life prediction method provided by the present application, as Figure 3 shown. Based on the Figure 1-2 embodiment, on the basis of the Figure 1-2 embodiment, the method for predicting the battery life by the battery life prediction model trained based on the

[0137] S301. Obtain the battery data of the target battery.

[0138] Among them, the battery data is used to indicate the solid state parameters of the battery and the operation data under the current working condition.

[0139] This prediction method is applied to a battery management system, and the battery management system is configured as Figure 1-2The trained battery life prediction model. Specifically, the operation data of the target battery under the current working condition is collected, and the operation data includes but is not limited to current, voltage, and temperature. In addition, the solid-state parameters of the target battery are received, and the solid-state parameters include battery type and usage environment. After obtaining the above operation data, the state of charge of the battery is determined according to the operation parameters, for example, calculating the SOC of the battery, and storing the SOC value in the database of the battery management system or in the cloud for later calling the state of charge value when predicting the battery health.

[0140] S302. Input the battery data into the battery life prediction model to obtain the predicted life output by the battery life prediction model.

[0141] Among them, the battery life prediction model is trained by the training method of the battery life prediction model as Figure 1-2 shown.

[0142] Specifically, according to the solid-state parameters in the above battery data, the static characterization parameters corresponding to the battery type and / or usage scenario in the solid-state parameters are called, and based on the operation data of the above target battery, the third life characteristic of the target battery under the current working condition is determined. In addition, based on the above static characterization parameters and the above solid-state parameters, the fourth life characteristic of the target battery is determined, and then the third life characteristic and the fourth life characteristic are fused to obtain the predicted health degree of the above target battery, and based on the predicted health degree, the remaining life of the target battery is determined.

[0143] For example, according to the battery type of the target battery, the static characterization parameters corresponding to the battery type are called to the static characterization layer, and the solid-state parameters are feature-mapped through the static characterization layer to obtain the above fourth life characteristic, which is used to characterize the influence of the solid-state parameters on the battery life. In addition, the operation data of the target battery is feature-processed through the dynamic characterization layer to obtain the above third life characteristic, which is used to characterize the attenuation characteristic of the battery life under the current working condition. Finally, the above third life characteristic and the fourth life characteristic are fused, and the predicted health degree of the target battery is determined based on the fused feature vector, and then the remaining life of the target battery is calculated according to the predicted health degree.

[0144] It can be understood that this method takes into account the influence of both the solid-state parameters of the battery and the current working condition on the battery life. Therefore, the remaining life obtained by this method is more accurate. The remaining life obtained by this method is helpful for formulating a more reasonable battery management strategy in various fields. For example, in electric vehicles or energy storage systems, the charging and discharging strategies can be optimized according to the predicted remaining life of the battery, so as to extend the overall service life of the battery.

[0145] Optionally, to ensure the safe use of the battery and maximize the utilization of battery resources, the usage status of the battery is determined to prevent the service life of the battery from reaching a certain level. Here, a method for monitoring the predicted life of the battery and warning the user based on the predicted life is shown. The method includes: real-time monitoring of the predicted life of the target battery, generating a warning message when the predicted life reaches a preset threshold, and sending the warning message to the user terminal, where the warning message includes the predicted life of the above-mentioned target battery.

[0146] The battery life prediction method provided by the embodiments of the present application obtains battery data of the target battery and inputs the battery data into the battery life prediction model to obtain the predicted life output by the battery life prediction model. This method uses the above battery life prediction model to extract the influence of the solid-state parameters and current working conditions of the target battery on the battery life, improving the accuracy of the predicted life.

[0147] Figure 4 It is a schematic structural diagram of a training device for a battery life prediction model provided by the present application, as Figure 4 shown, the training device 40 for the battery life prediction model provided in this embodiment includes:

[0148] The first acquisition module 401 is used to acquire static sample data and dynamic sample data of the sample battery. The static sample data is used to indicate the solid-state attributes of the sample battery, and the dynamic sample data is used to indicate the operation data of the sample battery under various working conditions;

[0149] The determination module 402 is used to determine a sample data set based on the static sample data and the dynamic sample data;

[0150] The training module 403 is used to perform training processing on the prediction model using the sample data set to obtain a trained battery life prediction model.

[0151] In a possible implementation manner, the static sample data includes: rated capacity, and the dynamic sample data includes: current under various working conditions, the discharge duration corresponding to the current, the initial state of charge parameter corresponding to the current discharge cycle, and the current state of charge parameter corresponding to the above discharge duration. The determination module 402 is further used to determine the discharge power based on the discharge current and the corresponding discharge duration in the dynamic sample data;

[0152] The determination module 402 is further used to determine the change amount of the state of charge parameter of the sample battery according to the initial state of charge parameter and the current state of charge parameter;

[0153] The determining module 402 is further configured to determine the current capacity of the sample battery based on the discharge power and the change amount of the charge parameter, and use the ratio of the current capacity to the rated capacity as the actual health degree;

[0154] The determining module 402 is further configured to use the static sample data, the dynamic sample data, and the actual health degree as the sample data set.

[0155] In a possible implementation manner, the prediction model includes a dynamic characterization layer and a static characterization layer, and the device further includes: a feature extraction module 404 and an optimization module 405;

[0156] The feature extraction module 404 is configured to perform feature extraction processing on the dynamic sample data through the dynamic characterization layer to obtain a plurality of first features of the sample battery;

[0157] The determining module 402 is further configured to determine a first life feature based on the correlation between the plurality of first features and the battery health degree;

[0158] The determining module 402 is further configured to determine a second life feature through the static characterization layer based on the correlation between the static sample data and the battery health degree;

[0159] The determining module 402 is further configured to determine the predicted health degree of the sample battery based on the first life feature and the second life feature;

[0160] The determining module 402 is further configured to determine the prediction accuracy corresponding to the sample battery based on the predicted health degree and the actual health degree;

[0161] The optimization module 405 is configured to, when the prediction accuracy does not reach the accuracy threshold, optimize the prediction model based on the prediction accuracy and re-determine the new predicted health degree until the prediction accuracy reaches the accuracy threshold or the number of training times reaches the upper limit value, to obtain the battery life prediction model.

[0162] In a possible implementation manner, the static sample data further includes: battery type and usage scenario, and the device further includes: an adjustment module 406;

[0163] The adjustment module 406 is configured to adjust the dynamic characterization parameters corresponding to the dynamic characterization layer and the static characterization parameters corresponding to the static characterization layer according to the prediction accuracy, where the dynamic characterization parameters are shared by multiple battery types and usage scenarios, and the static characterization parameters correspond to specific battery types and / or usage scenarios.

[0164] The training device for the battery life prediction model provided in this embodiment can execute the training method for the battery life prediction model provided in the above method embodiment. The implementation principle and technical effects are similar, and will not be elaborated here in this embodiment.

[0165] Figure 5 As shown in the structural schematic diagram of a battery life prediction device provided in this application, Figure 5 the battery life prediction device 50 provided in this embodiment includes:

[0166] A second acquisition module 501, configured to acquire battery data of a target battery, where the battery data is used to indicate the solid-state parameters of the battery and the operation data under the current working condition;

[0167] An input module 502, configured to input the battery data into the battery life prediction model to obtain the predicted life output by the battery life prediction model;

[0168] Among them, the above battery life prediction model is trained by the training method of the battery life prediction model as shown in the above Figure 1-2 embodiment.

[0169] In a possible implementation manner, the device further includes: a calling module 503, a determining module 504, and a processing module 505;

[0170] The calling module 503 is configured to call the static characterization parameters corresponding to the battery type and / or usage scenario in the solid-state parameters according to the solid-state parameters in the battery data;

[0171] The determining module 504 is configured to determine a third life characteristic of the target battery under the current working condition based on the operation data;

[0172] The determining module 504 is further configured to determine a fourth life characteristic of the target battery based on the static characterization parameters and the solid-state parameters;

[0173] The processing module 505 is configured to perform a fusion process on the third life characteristic and the fourth life characteristic to obtain the predicted health degree of the target battery;

[0174] The determining module 504 is further configured to determine the remaining life of the target battery based on the predicted health degree.

[0175] The battery life prediction device provided in this embodiment can execute the battery life prediction method provided in the above method embodiment. The implementation principle and technical effects are similar, and will not be elaborated here in this embodiment.

[0176] Figure 6A schematic structural diagram of an electronic device provided by this application. As Figure 6 shown, the electronic device 60 provided in this embodiment includes: at least one processor 601 and a memory 602. Optionally, the device 60 further includes a communication component 603. Among them, the processor 601, the memory 602, and the communication component 603 are connected through a bus 604.

[0177] In a specific implementation process, at least one processor 601 executes computer-executable instructions stored in the memory 602, so that at least one processor 601 executes the above method.

[0178] For the specific implementation process of the processor 601, reference can be made to the above method embodiment, and its implementation principle and technical effects are similar, so they will not be elaborated here in this embodiment.

[0179] In the above embodiment, it should be understood that the processor may be a central processing unit (English: Central Processing Unit, abbreviated: CPU), or other general-purpose processors, digital signal processors (English: Digital Signal Processor, abbreviated: DSP), application-specific integrated circuits (English: Application Specific Integrated Circuit, abbreviated: ASIC), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the invention can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0180] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.

[0181] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0182] This application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0183] The present application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above method.

[0184] The above-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk or an optical disk. The readable storage medium can be any available medium accessible by a general-purpose or special-purpose computer.

[0185] An exemplary readable storage medium is coupled to the processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in a device.

[0186] The division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces, and the indirect coupling or communication connection of devices or units can be in electrical, mechanical or other forms.

[0187] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0188] In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0189] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, and other various media that can store program codes.

[0190] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When this program is executed, it executes the steps including the above method embodiments; and the aforementioned storage medium includes: ROMs, RAMs, magnetic disks, or optical discs, and other various media that can store program codes.

[0191] Finally, it should be noted that: After considering the specification and practicing the invention disclosed herein, those skilled in the art will easily think of other implementation manners of the present invention. The present invention is intended to cover any variations, uses, or adaptive changes of the present invention. These variations, uses, or adaptive changes follow the general principles of the present invention and include the common general knowledge or conventional technical means in the technical field of the present invention that is not disclosed in the present invention. It is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A method for training a battery life prediction model, characterized in that: include: Acquire static sample data and dynamic sample data of a sample battery, wherein the static sample data is used to indicate the solid-state properties of the sample battery, and the dynamic sample data is used to indicate the operating data of the sample battery under various working conditions; Determine a sample data set based on the static sample data and the dynamic sample data; The sample data set is used to train the prediction model to obtain a trained battery life prediction model.

2. The method according to claim 1, characterized in that The static sample data includes: rated capacity, the dynamic sample data includes: current under various working conditions, discharge duration corresponding to the current, initial charge parameters corresponding to the current discharge cycle, and current charge parameters corresponding to the discharge duration, and the sample data set is determined based on the static sample data and the dynamic sample data, including: Determine the discharge capacity based on the discharge current and the corresponding discharge duration in the dynamic sample data; Determining a change in the charge parameter of the sample battery according to the initial charge parameter and the current charge parameter; Determine the current capacity of the sample battery based on the discharged power and the charge parameter change, and use the ratio of the current capacity to the rated capacity as the actual health level; The static sample data, the dynamic sample data and the actual health status are used as the sample data set.

3. The method according to claim 2, characterized in that The prediction model includes a dynamic characterization layer and a static characterization layer. The sample data set is used to train the prediction model to obtain a trained battery life prediction model, including: By means of the dynamic characterization layer, feature extraction processing is performed on the dynamic sample data to obtain a plurality of first features of the sample battery, and a first life feature is determined based on a correlation between the plurality of first features and the battery health level; Determine, by the static characterization layer, a second life feature based on the correlation between the static sample data and the battery health level; Determining a predicted health of the sample battery based on the first life characteristic and the second life characteristic; Determining a prediction accuracy corresponding to the sample battery based on the predicted health status and the actual health status; When the prediction accuracy does not reach the accuracy threshold, the prediction model is optimized based on the prediction accuracy, and a new prediction health level is re-determined until the prediction accuracy reaches the accuracy threshold or the number of training times reaches an upper limit, thereby obtaining the battery life prediction model.

4. The method according to claim 3, characterized in that The static sample data further includes: battery type and usage scenario, and the optimizing the prediction model based on the prediction accuracy includes: The dynamic characterization parameters corresponding to the dynamic characterization layer and the static characterization parameters corresponding to the static characterization layer are adjusted according to the prediction accuracy, wherein the dynamic characterization parameters are shared by multiple battery types and usage scenarios, and the static characterization parameters correspond to specific battery types and / or usage scenarios.

5. A battery life prediction method, characterized in that: The method comprises: Acquire battery data of a target battery, wherein the battery data is used to indicate solid-state parameters of the battery and operating data under current working conditions; Inputting the battery data into a battery life prediction model to obtain a predicted life output by the battery life prediction model; The battery life prediction model is obtained by training using the battery life prediction model training method as described in any one of claims 1 to 5.

6. The method according to claim 5, characterized in that The step of inputting the battery data into a battery life prediction model to obtain a predicted life output by the battery life prediction model includes: According to the solid-state parameters in the battery data, calling static characterization parameters corresponding to the battery type and / or usage scenario in the solid-state parameters; Determining, based on the operating data, a third life characteristic of the target battery under the current operating condition; Determining a fourth life characteristic of the target battery based on the static characterization parameter and the solid-state parameter; The third life characteristic and the fourth life characteristic are fused to obtain a predicted health state of the target battery, and the remaining life of the target battery is determined based on the predicted health state.

7. The method according to claim 6, characterized in that The method further comprises: The predicted life of the target battery is monitored in real time, and when the predicted life reaches a preset threshold, an early warning message is generated and sent to a user terminal, wherein the early warning message includes the predicted life.

8. A training device for a battery life prediction model, characterized in that: include: A first acquisition module, used to acquire static sample data and dynamic sample data of a sample battery, wherein the static sample data is used to indicate the solid-state properties of the sample battery, and the dynamic sample data is used to indicate the operating data of the sample battery under various working conditions; A determination module, configured to determine a sample data set based on the static sample data and the dynamic sample data; The training module is used to use the sample data set to train the prediction model to obtain a trained battery life prediction model.

9. A battery life prediction device, characterized in that: include: A second acquisition module, used to acquire battery data of a target battery, wherein the battery data is used to indicate solid-state parameters of the battery and operating data under current working conditions; An input module, used for inputting the battery data into a battery life prediction model to obtain a predicted life output by the battery life prediction model; The battery life prediction model is obtained by training using the battery life prediction model training method as described in any one of claims 1 to 5.

10. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 7.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

12. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 7 when being executed by a processor.

Images