Battery life prediction method, system and equipment based on transfer learning and medium

Abstract

The invention discloses a battery life prediction method, system and device based on transfer learning and a medium, and relates to the technical field of battery monitoring. The method comprises the following steps: acquiring first battery operation data, vehicle driving data and environment data of a first energy storage battery; determining target domain data according to the first battery operation data, the vehicle driving data and the environment data, and constructing a corresponding initial battery life prediction model; determining a plurality of second energy storage batteries different from the first energy storage battery in battery type, and acquiring second battery operation data and battery life of each second energy storage battery; the second battery operation data and the battery life serve as source domain data, model parameters of the initial battery life prediction model are adjusted through a transfer learning algorithm and the source domain data, and a target battery life prediction model is obtained; and predicting the target battery life of the first energy storage battery based on the target battery life prediction model. By implementing the technical scheme provided by the invention, the effect of improving the battery life prediction efficiency is achieved.

Description

Technical Field

[0001] The present application relates to the field of battery monitoring technology, and specifically to a battery life prediction method, system, device and medium based on transfer learning. Background Art

[0002] With the widespread use of electric vehicles and renewable energy storage systems, battery technology has become one of the core technologies in the modern energy field. The performance and life of batteries directly affect the operating efficiency and maintenance cost of equipment. Therefore, accurate prediction of battery life is crucial for battery maintenance management and cost-effectiveness analysis. However, battery life prediction faces many challenges, such as the large differences in battery performance under different operating conditions and environmental factors, as well as differences in different types of battery materials and designs. At present, the existing battery life prediction method predicts the battery life by monitoring the battery data of the energy storage battery and establishing a separate model. However, in actual applications, when different types of energy storage batteries are operated in various environments, the degree of loss of the energy storage batteries varies. Therefore, when predicting the battery life of different energy storage batteries, a large number of models need to be established separately, and the amount of data processing is large, resulting in low efficiency in battery life prediction. Summary of the invention

[0003] The present application provides a battery life prediction method, system, device and medium based on transfer learning, which has the effect of improving the efficiency of battery life prediction.

[0004] In a first aspect, the present application provides a battery life prediction method based on transfer learning, comprising: Acquire first battery operation data of a first energy storage battery, vehicle driving data of a corresponding vehicle, and environmental data of the environment in which the vehicle is located; Determine target domain data of the first energy storage battery according to the first battery operation data, the vehicle driving data and the environmental data, and construct an initial battery life prediction model corresponding to the target domain data; Determine a plurality of second energy storage batteries of a different battery type from the first energy storage battery, and obtain second battery operation data and battery life of each of the second energy storage batteries; Using each of the second battery operation data and each of the battery life as source domain data, adjusting the model parameters of the initial battery life prediction model through a transfer learning algorithm and the source domain data, to obtain a current battery life prediction model; Based on the target battery life prediction model, a target battery life of the first energy storage battery is predicted.

[0005] By adopting the above technical solution, the operation data of the first energy storage battery, the corresponding vehicle driving data and environmental data are obtained, and the target domain data is determined accordingly, so that the various key factors affecting the battery life can be fully considered and the modeling quality of the prediction model can be improved. By constructing an initial battery life prediction model, and using the operation data and life data of other different types of second energy storage batteries as source domain data, and adjusting the initial model parameters through the transfer learning algorithm, the target battery life prediction model obtained can avoid the defect of modeling each battery separately, but use the transfer learning algorithm to combine the data of multiple types of batteries for migration, thereby improving the generalization ability of the target battery life prediction model, reducing data requirements and modeling complexity, and improving the efficiency of battery life prediction.

[0006] Optionally, extract battery aging characteristics from the first battery operation data; obtain first target domain characteristics based on the battery aging characteristics, vehicle mileage in the vehicle driving data, and driving time in the vehicle driving data; obtain second target domain characteristics based on the battery aging characteristics, ambient temperature in the environmental data, and ambient humidity in the environmental data; and use the first target domain characteristics and the second target domain characteristics as the target domain data.

[0007] By adopting the above technical solution, by extracting the battery aging characteristics from the first battery operation data, the internal state of the battery can be directly reflected, providing more accurate and critical input features for the subsequent prediction of the remaining battery life. The battery aging characteristics are combined with the vehicle driving data (mileage and duration) to obtain the first target domain characteristics. This approach fully considers the impact of the vehicle's operating conditions on battery aging, so that the prediction model can better capture the intrinsic connection between battery aging and vehicle use, and combines the battery aging characteristics with environmental data (temperature and humidity) to obtain the second target domain characteristics. Ambient temperature and humidity are another key factor affecting battery life. By incorporating them into feature engineering, the prediction model can better understand and quantify the mechanism of environmental effects on battery aging. The first and second target domain features are combined as the final target domain data input, so that the prediction model can simultaneously learn and weigh the comprehensive impact of battery aging, vehicle operating conditions and environmental conditions on battery life, and comprehensively improve the prediction ability of the model.

[0008] Optionally, a voltage floating rate is determined based on the operating voltage and a preset life cycle; a current floating rate is determined based on the operating current and the preset life cycle; an impedance loss rate is determined based on the battery impedance and the preset life cycle; if any of the voltage floating rate, the current floating rate and the impedance loss rate has a characteristic greater than a preset threshold, the characteristic is determined to be a battery aging characteristic.

[0009] By adopting the above technical solutions, using quantitative indicators such as voltage floating rate, current floating rate and impedance loss rate, the performance degradation of the battery in different working cycles can be intuitively reflected. Voltage and current are two key parameters that describe the charging and discharging performance of the battery, while impedance reflects the aging degree of the internal structure and medium of the battery. By tracking and monitoring these three indicators for a long time and conducting quantitative analysis, the aging level of the battery can be comprehensively evaluated.

[0010] Optionally, a first training data set is determined based on the first target domain feature and the timestamp of the first target domain feature, and a second training data set is determined based on the second target domain feature and the timestamp of the second target domain feature; a first battery life label corresponding to the first training data set and a second battery life label corresponding to the second training data set are obtained; based on the first training data set and the first battery life label, an initial neural network model is trained until a preset model training stop condition is reached to obtain a first battery model; based on the second training data set and the second battery life label, an initial neural network model is trained until the preset model training stop condition is reached to obtain a second battery model, and the preset model training stop condition includes that the model training times reaches a preset times or the error rate of the model output result is less than a preset error rate; the first battery model and the second battery model are fitted to obtain the initial battery life prediction model.

[0011] By adopting the above technical solution, the first target domain features (including battery aging features, vehicle mileage and driving time) and the second target domain features (including battery aging features, ambient temperature and ambient humidity) are respectively used as input to construct the first training data set and the second training data set. This separate processing of features from different sources is conducive to the model to better learn the intrinsic relationship between different features. For the two training data sets, the solution obtains the corresponding battery life labels (first and second life labels) from the battery life data as label data for supervised learning to guide the model training process. The initial prediction model is obtained by fitting the two battery models. This multi-model fusion approach can more comprehensively capture the impact of different features on battery life and improve the model's expressiveness and generalization performance.

[0012] Optionally, the layers whose similarity in the first battery model and the second battery model is greater than or equal to a preset similarity and the parameters corresponding to the layers are taken as the first fitting layer object; the layers whose similarity in the first battery model and the second battery model is less than the preset similarity and the parameters corresponding to the layers are weighted summed to obtain a second fitting layer object; and the initial battery life prediction model is constructed based on the first fitting layer object and the second fitting layer object.

[0013] By adopting the above technical solution, by calculating the similarity of each layer in the first and second models, the difference between the two models at different levels can be quantitatively evaluated. For highly similar layers, it means that the two models have learned a common pattern at this layer and can be directly fused. For layers with lower similarity, the solution uses a weighted summation method for fusion. This is more flexible and efficient than simple splicing or averaging. It can give different weights to different layers according to actual conditions, better retain the valuable parts of the two models, and through the layered fusion strategy, the initial prediction model can effectively inherit the excellent parts of the first and second models and make up for their respective shortcomings, thereby obtaining a more comprehensive and accurate modeling capability, avoiding the need to completely retrain a new model, but based on the fusion of existing models, saving computing resources and improving model building efficiency.

[0014] Optionally, based on the operating data of each of the second batteries and the battery life in the source domain data, a migration model of each of the second energy storage batteries is constructed; according to the target domain data, each of the migration models is trained to obtain the model parameters to be migrated after the training of each of the migration models; the target model parameters in each of the model parameters to be migrated are determined, and the model parameters of the layer to be migrated in the initial battery life prediction model are adjusted to the target model parameters.

[0015] By adopting the above technical solution, multiple migration models are constructed, which can learn the characteristic patterns of different second batteries respectively and fully tap the knowledge in the source domain data. Different migration models correspond to different migration paths, which increases the diversity of knowledge migration. Training the migration model on the target domain data can gradually adapt it to the target task and improve its effectiveness in the target domain. During the training process, the loss function can be customized according to actual needs, and the migration learning strategy can be flexibly controlled. By determining the model parameters to be migrated, the solution can finely adjust the parameter values ​​at a specific level of the initial prediction model to achieve incremental model learning.

[0016] Optionally, among the model parameters to be migrated, multiple model parameter combinations are randomly generated, and each of the model parameter combinations is used as multiple populations; the fitness of each of the populations and the target domain data is calculated, and the population corresponding to the maximum fitness among the populations is used as the target model parameter.

[0017] By adopting the above technical solution, multiple parameter combinations are randomly generated in the model parameters to be migrated, and the basic population for population initialization is constructed. This random combination method can maintain the diversity of the parameter search space to the greatest extent, laying a good foundation for subsequent optimization. The fitness score is calculated for each parameter combination (population) and the target domain data. The fitness reflects the degree of fit of the current parameter combination to the target task and is a key indicator for evaluating the quality of the population in the evolutionary algorithm. Through fitness evaluation, the performance of each parameter combination can be effectively measured, and the population with the highest fitness can be selected as the target model parameter, and applied to the initial prediction model to complete the fine-tuning of the parameters. This natural selection process ensures that the prediction model obtains the optimal parameter combination, so that the performance of the model in the target domain is optimal.

[0018] In a second aspect of the present application, a battery life prediction system based on transfer learning is provided, the system comprising: A data acquisition module, used to acquire first battery operation data of the first energy storage battery, vehicle driving data of the corresponding vehicle, and environmental data of the environment in which the vehicle is located; a model determination module, configured to determine the target domain data of the first energy storage battery according to the first battery operation data, the vehicle driving data and the environmental data, and to construct an initial battery life prediction model corresponding to the target domain data; determine a plurality of second energy storage batteries of different battery types from the first energy storage battery, and obtain second battery operation data and battery life of each of the second energy storage batteries; A model migration module, used to use each of the second battery operation data and each of the battery life as source domain data, and adjust the model parameters of the initial battery life prediction model through a migration learning algorithm and the source domain data to obtain a current battery life prediction model; A battery life prediction module is used to predict the target battery life of the first energy storage battery based on the target battery life prediction model.

[0019] In a third aspect of the present application, an electronic device is provided.

[0020] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein the program can implement a battery life prediction method based on transfer learning when loaded and executed by the processor.

[0021] In a fourth aspect of the present application, a computer-readable storage medium is provided.

[0022] A computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to implement a battery life prediction method based on transfer learning.

[0023] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: By adopting the technical solution of the present application, the operating data of the first energy storage battery, the corresponding vehicle driving data and environmental data are obtained, and the target domain data is determined accordingly, so that the various key factors affecting the battery life can be fully considered and the modeling quality of the prediction model can be improved. By constructing an initial battery life prediction model, and using the operating data and life data of other different types of second energy storage batteries as source domain data, and adjusting the initial model parameters through the transfer learning algorithm, the target battery life prediction model obtained can avoid the defect of modeling each battery separately, but use the transfer learning algorithm to combine the data of multiple types of batteries for migration, thereby improving the generalization ability of the target battery life prediction model, reducing data requirements and modeling complexity, and improving the efficiency of battery life prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flowchart of a battery life prediction method based on transfer learning provided in an embodiment of the present application; Figure 2 It is a structural schematic diagram of a battery life prediction system based on transfer learning disclosed in an embodiment of the present application; Figure 3 It is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present application.

[0025] Description of reference numerals: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0026] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0027] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0028] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0029] The present application embodiment provides a battery life prediction method based on transfer learning. In one embodiment, please refer to Figure 1 , Figure 1 This is a flow chart of a battery life prediction method based on transfer learning provided in an embodiment of the present application. The method can be implemented by a computer program, which can be integrated into an application or run as an independent tool application. The method can also be implemented by a single-chip microcomputer or run on a battery life prediction system based on transfer learning based on a von Neumann system. Specifically, the method can include the following steps: Step 101: Acquire first battery operation data of a first energy storage battery, vehicle driving data of a corresponding vehicle, and environmental data of the surrounding environment.

[0030] Among them, the first battery operation data refers to data reflecting the operating status of the target first energy storage battery during actual use. In the embodiment of the present application, it can be understood as time series data including parameters such as operating voltage, operating current and battery internal resistance, which is used to analyze the electrochemical changes of the battery and evaluate the health status and loss degree of the battery.

[0031] Vehicle driving data refers to data reflecting the actual usage of the vehicle installed with the target first energy storage battery. In the embodiment of the present application, it can be understood to include parameters such as the vehicle's cumulative mileage and driving time, which are used to evaluate the battery's usage environment and approximate degree of wear.

[0032] Environmental data refers to parameter data reflecting the use environment of the target first energy storage battery. In the embodiment of the present application, it can be understood to include parameters such as temperature and humidity, which are used to evaluate the impact of external environmental conditions on battery performance.

[0033] Specifically, in order to establish a battery life prediction model suitable for the first energy storage battery, it is necessary to obtain various data closely related to the performance of the battery, including the operating data of the battery itself, the driving data of the vehicle in which it is located, and the data of the external environment. The operating data of the first battery may include parameters such as operating voltage, operating current, and battery internal resistance. The time series data of these parameters reflect the electrochemical changes of the battery during use and are an important basis for evaluating the health status of the battery. The driving data of the vehicle includes information such as the cumulative mileage and driving time of the vehicle, which reflects the use environment and loss of the battery. Environmental data includes parameters such as temperature and humidity, which will affect the chemical reaction process of the battery, thereby affecting the health status of the battery. Acquiring multi-source heterogeneous data comprehensively reflects all aspects that affect the battery life, which is the basis for building an accurate battery health assessment and life prediction model.

[0034] Step 102: Determine target domain data of the first energy storage battery according to the first battery operation data, vehicle driving data and environmental data, and construct an initial battery life prediction model corresponding to the target domain data.

[0035] Among them, the target domain data refers to a data feature set that reflects the usage of the target first energy storage battery. In the embodiment of the present application, it can be understood as consisting of the first target domain features and the second target domain features, which are used to construct a battery life prediction model suitable for the first energy storage battery.

[0036] The initial battery life prediction model refers to a battery life prediction model that is constructed only based on the data of the target first energy storage battery and has not yet been optimized. In the embodiment of the present application, it can be understood as a neural network model for battery life prediction trained based on target domain data, which is used to make a preliminary prediction of the remaining battery life.

[0037] Specifically, after obtaining the operating data, vehicle driving data and environmental data of the first battery, it is necessary to determine the target domain features of the first battery based on these multi-source heterogeneous data to build a corresponding battery life prediction model. It is necessary to extract the voltage floating rate, current floating rate, battery internal impedance loss rate and other features from the operating data of the first battery. These features can reflect the health status of the battery. Then, by combining the vehicle's mileage and driving time, the first target domain features reflecting the battery usage environment can be obtained. At the same time, by combining the ambient temperature and humidity data, the second target domain features reflecting the usage environment can be obtained. The first target domain features and the second target domain features are combined to form the target domain data of the first energy storage battery. Based on the target domain data determined in this way, the feature data of different time periods and the corresponding battery life labels can be obtained to construct an initial battery life prediction model. This model establishes a mapping from battery health features to life by learning the relationship between the variables in the target domain data, and can predict the remaining life of the battery.

[0038] Based on the above embodiment, as an optional embodiment, in step 102: determining the target domain data of the first energy storage battery according to the first battery operation data, the vehicle driving data and the environmental data, this step may also include the following steps: Step 201: extracting battery aging characteristics from first battery operation data.

[0039] Among them, the battery aging characteristics refer to relevant characteristic parameters that can reflect the health decay status of the first energy storage battery. In the embodiment of the present application, they can be understood as parameters such as voltage floating rate, current floating rate and internal impedance loss rate, which are used to evaluate the health status and remaining life of the battery.

[0040] Specifically, the floating rate of battery voltage, the floating rate of current, and the loss rate of internal resistance within a preset life cycle are calculated. Among them, the voltage floating rate represents the degree of battery voltage decay over time, the current floating rate represents the degree of decrease in current output capacity, and the internal resistance loss rate represents the degree of active material loss. If the rate of change of these characteristics exceeds a preset threshold, it can be determined as a signal of battery performance degradation, so that these characteristics are extracted as battery aging characteristics. The purpose of extracting these aging characteristics is to more accurately evaluate the health status of the battery. The more changes in battery voltage stability, load capacity, and internal resistance directly reflect the remaining life of the battery.

[0041] Based on the above embodiment, as an optional embodiment, in step 201: extracting the battery aging characteristics in the first battery operation data, this step may also include the following steps: Step 211: Determine the voltage floating rate according to the operating voltage and the preset life cycle.

[0042] Among them, the preset life cycle refers to the theoretical use cycle of the first energy storage battery from leaving the factory to being scrapped. In the embodiment of the present application, it can be understood as the standard service life indicated in the technical parameter manual of the battery product, which is used as a reference benchmark for calculating the battery aging characteristics.

[0043] The operating voltage refers to the real-time voltage value at the output end of the target first energy storage battery during use. In the embodiment of the present application, it can be understood as the terminal voltage monitoring data when the battery is working, which is used to calculate the voltage floating rate and evaluate the health status of the battery.

[0044] The voltage floating rate refers to the ratio of the fluctuation range of the operating voltage relative to the initial voltage or the average voltage of the first energy storage battery within a preset life cycle. In the embodiment of the present application, it can be understood as the rate of change of the battery operating voltage data, which is used to evaluate the degree of decrease in battery voltage stability, thereby reflecting the health attenuation state of the battery.

[0045] Specifically, in order to evaluate the health decay of the first energy storage battery, it is necessary to determine the voltage floating rate as one of the battery aging characteristics based on the operating voltage, statistically analyze the operating voltage data of the first battery within the preset life cycle, and calculate the fluctuation range and change trend of its voltage value. Then, based on the ratio of the voltage fluctuation range to its initial value or average value, the voltage floating rate of the battery is obtained. The purpose of this is that the stability of the voltage directly affects the output performance of the battery. When the battery ages, the voltage fluctuation range will gradually increase. Therefore, the degree of health decay of the battery can be evaluated by calculating the voltage floating rate.

[0046] Step 221: Determine the current floating rate according to the operating current and the preset life cycle.

[0047] Among them, the operating current refers to the real-time current value of the target first energy storage battery when it is charged and discharged during use. In the embodiment of the present application, it can be understood as the current monitoring data when the battery is working, which is used to calculate the current floating rate and evaluate the battery's load capacity.

[0048] The current floating rate refers to the ratio of the fluctuation range of the operating current relative to the initial current or the average current of the first energy storage battery within the preset life cycle. In the embodiment of the present application, it can be understood as the rate of change of the battery operating current data, which is used to evaluate the degree of decline in the battery load capacity, thereby reflecting the health attenuation state of the battery.

[0049] Specifically, the operating current data of the first battery within the preset life cycle is statistically analyzed to calculate the fluctuation range and change trend of its current value. Then, based on the ratio of the current fluctuation range to its initial value or average value, the current floating rate of the battery is obtained. The purpose of this is that the stability of the current directly affects the output capacity of the battery. When the battery ages, the fluctuation range of the current will gradually increase. Therefore, by calculating the current floating rate, the attenuation degree of the battery load capacity can be evaluated.

[0050] Step 231: Determine the impedance loss rate according to the battery impedance and a preset life cycle.

[0051] Among them, the battery impedance refers to the equivalent series impedance inside the target first energy storage battery during use, which can be understood as the resistance of the chemical reaction inside the battery in the embodiment of the present application, and is used to calculate the impedance loss rate and evaluate the degree of attenuation of the internal performance of the battery.

[0052] The impedance loss rate refers to the growth ratio of the battery impedance relative to the initial impedance of the first energy storage battery within the preset life cycle. In the embodiment of the present application, it can be understood as the rate of change of the battery internal resistance data, which is used to evaluate the degree of degradation of the internal performance of the battery, thereby reflecting the health attenuation state of the battery.

[0053] Specifically, the battery impedance data of the first battery within the preset life cycle is statistically analyzed to calculate the change trend and growth rate of its impedance value. Then, based on the ratio of the impedance growth rate to its initial impedance, the impedance loss rate of the battery is obtained. The purpose of this is that the increase in the internal impedance of the battery will directly reduce the battery's efficacy and shorten its service life. Therefore, by calculating the impedance loss rate, the degree of internal performance degradation of the battery can be evaluated, providing key information for building a battery life prediction model, and obtaining an impedance loss rate indicator that reflects the growth of the battery's internal resistance, which is conducive to more accurately predicting the remaining life of the battery.

[0054] Step 241: If there is a feature among the voltage floating rate, the current floating rate, and the impedance loss rate that is greater than a preset threshold, then the feature is determined to be a battery aging feature.

[0055] Specifically, in order to screen out the features that can effectively reflect the health decay state of the first battery, it is necessary to perform threshold judgments on the voltage floating rate, current floating rate and impedance loss rate to extract the battery aging features. The normal variation range of the voltage floating rate, current floating rate and impedance loss rate is determined in advance based on the statistical analysis of the usage data of this model of battery, and the corresponding thresholds are set. When any of the characteristic indicators is greater than the preset threshold, it indicates that the feature has been significantly alienated and can be determined as a signal of battery performance decay. At this time, the feature is determined as a battery aging feature for evaluating the battery health state. The purpose of this is to extract only features closely related to battery aging through threshold filtering, prevent invalid features from interfering with the model prediction results, and improve the accuracy of battery life prediction.

[0056] Step 202: Obtain a first target domain feature according to the battery aging feature, the vehicle mileage in the vehicle driving data, and the driving time in the vehicle driving data.

[0057] Among them, vehicle mileage refers to the mileage recorded during the driving process of a vehicle equipped with a target first energy storage battery. In the embodiment of the present application, it can be understood as the cumulative mileage data of the vehicle, which is used to reflect the usage time and wear environment of the target battery.

[0058] The driving time refers to the total driving time recorded during the driving process of a vehicle equipped with a target first energy storage battery. In the embodiment of the present application, it can be understood as the cumulative driving time data of the vehicle, which is used to reflect the usage time and wear environment of the target battery.

[0059] The first target domain feature refers to a feature set that integrates the first energy storage battery operation data and vehicle driving data. In the embodiment of the present application, it can be understood as consisting of battery aging characteristics and vehicle mileage and driving time, and is used to construct a battery life prediction model that takes into account both the battery's own state and the usage environment.

[0060] Specifically, the battery's aging characteristics such as voltage floating rate, current floating rate, and internal resistance loss rate are combined with the vehicle's cumulative mileage and driving time to construct a first target domain feature that represents the battery's usage environment. The purpose of this is to allow the first target domain feature to reflect both the battery's own health attenuation state and the usage environment. Mileage and driving time can reflect the battery's usage time and loss environment. By combining it with the battery aging characteristics, the first target domain feature that represents both the battery's health state and the usage environment can be obtained. Obtaining such a first target domain feature that fully reflects the actual battery situation will provide richer and more effective information for the subsequent construction of a battery life prediction model. This enables the prediction model to make more accurate predictions of the remaining battery life based on the battery's health state and usage environment.

[0061] Step 203: Obtain a second target domain feature according to the battery aging feature, the ambient temperature in the environmental data, and the ambient humidity in the environmental data; and use the first target domain feature and the second target domain feature as target domain data.

[0062] Among them, the ambient temperature refers to the temperature parameter of the environment in which the target first energy storage battery is located during use. In the embodiment of the present application, it can be understood as the temperature monitoring data of the battery working environment, which is used to evaluate the impact of external temperature conditions on battery performance.

[0063] Ambient humidity refers to the humidity parameter of the environment in which the target first energy storage battery is located during use. In the embodiment of the present application, it can be understood as the relative humidity monitoring data of the battery working environment, which is used to evaluate the impact of external humidity conditions on battery performance.

[0064] The second target domain feature refers to a feature set that integrates the operating data and environmental data of the first energy storage battery. In the embodiment of the present application, it can be understood as consisting of battery aging characteristics and ambient temperature and humidity, and is used to construct a battery life prediction model that takes into account both the battery's own state and external environmental conditions.

[0065] Specifically, the battery's aging characteristics such as voltage floating rate, current floating rate, and internal resistance loss rate are combined with ambient temperature and humidity parameters to construct a second target domain feature that represents the impact of the external environment. The purpose of this is to allow the target domain feature to reflect the battery's own state and usage environment, as well as the influencing factors of external environmental conditions. Temperature and humidity conditions affect the battery's chemical reaction process, thereby affecting the battery's health state. The first target domain feature and the second target domain feature are combined to form the final target domain data. These data comprehensively reflect the factors that affect all aspects of battery life. The battery life prediction model constructed on this basis can more accurately evaluate the health decay and life of batteries in complex environments.

[0066] Based on the above embodiment, as an optional embodiment, in step 102: constructing an initial battery life prediction model corresponding to the target domain data, this step may also include the following steps: Step 204: Determine a first training data set according to the first target domain feature and the timestamp of the first target domain feature, and determine a second training data set according to the second target domain feature and the timestamp of the second target domain feature.

[0067] Among them, the first training data set refers to a training sample set containing the first target domain features and their timestamps. In the embodiment of the present application, it can be understood as a data set that only contains the aging characteristics of the battery itself, which is used to train the model to evaluate the battery health status.

[0068] The second training data set refers to a training sample set containing the second target domain features and their timestamps. In the embodiment of the present application, it can be understood as a data set that contains both battery aging characteristics and environmental characteristics, and is used to train a model to evaluate the impact of the external environment on the battery health status.

[0069] Specifically, the first target domain features reflecting the aging state of the battery itself are summarized with the corresponding acquisition timestamps to form the first training data set. At the same time, the second target domain features reflecting the impact of the external environment are summarized with the corresponding acquisition timestamps to form the second training data set. The purpose of this is that the training data set directly determines the learning quality of the model. Combining target domain features of different categories as training data can make the model's battery health assessment more comprehensive. The first training data set allows the model to learn the battery's own aging laws; the environmental impact information provided by the second training data set can improve the model's adaptability to complex application scenarios, and obtain training data for the use environment and aging characteristics of the specific battery, which will improve the generalization performance of the battery life prediction model and make reliable predictions of the remaining battery life.

[0070] Step 205: Obtain a first battery life label corresponding to the first training data set and a second battery life label corresponding to the second training data set; based on the first training data set and the first battery life label, train the initial neural network model until a preset model training stop condition is reached to obtain a first battery model.

[0071] Among them, the first battery life label refers to the actual service life of the battery corresponding to the first training data set. In the embodiment of the present application, it can be understood that only the battery usage cycle data under the battery's own state is considered to provide the target output corresponding to the first training data set, and the training model's ability to predict battery life.

[0072] The first battery model refers to a battery life prediction model trained using only the first training data set, which can be understood in the embodiment of the present application as a neural network trained based on the battery's own state characteristics, and is used to predict the battery's health status and remaining life.

[0073] Specifically, the actual battery life label corresponding to the first training data set and the battery life label considering environmental factors corresponding to the second training data set are obtained, and then based on the first training data set containing only the battery's own state characteristics and the corresponding life label, the initial neural network model is used for iterative training until the value of the model loss function is less than a preset threshold, so as to obtain a life prediction model for the first battery, that is, the first battery model. The purpose of this is to acquire the ability to evaluate the battery health status through training the model on the first training data set, and to preliminarily establish a mapping relationship between battery life and characteristic parameters for battery life prediction.

[0074] Step 206: Based on the second training data set and the second battery life label, the initial neural network model is trained until the preset model training stop condition is reached to obtain a second battery model. The preset model training stop condition includes that the model training times reach a preset number of times or the error rate of the model output result is less than a preset error rate; fit the first battery model and the second battery model to obtain an initial battery life prediction model.

[0075] Among them, the second battery life label refers to the battery life corresponding to the second training data set taking into account environmental factors. In the embodiment of the present application, it can be understood as considering the battery's own state and the battery usage cycle data under the external environment at the same time, and is used to provide the target output corresponding to the second training data set, and the training model's ability to comprehensively evaluate battery life.

[0076] Specifically, a battery life label considering environmental factors corresponding to the second training data set is obtained. Then, based on the second training data set containing both battery characteristics and environmental characteristics and the corresponding life label, the first battery model is used as the initial model to continue iterative training. The training will continue until the number of model training times or the model prediction error meets the preset requirements, and finally the second battery model is obtained. Then, the model fusion strategy is adopted to fit the first battery model and the second battery model to obtain an initial battery life prediction model that integrates the battery's own factors and external environmental factors.

[0077] Based on the above embodiment, as an optional embodiment, in step 206: fitting the first battery model and the second battery model to obtain an initial battery life prediction model, this step may also include the following steps: Step 216: taking the layers and parameters corresponding to the layers in the first battery model and the second battery model whose similarity is greater than or equal to the preset similarity as the first fitting layer objects.

[0078] Among them, similarity refers to the degree of matching between the parameters of the corresponding layers in the first battery model and the second battery model. In the embodiment of the present application, it can be understood as the numerical difference of the trainable parameters in the corresponding layers, which is used to evaluate the fusibility between the two model layers.

[0079] The first fitting layer object refers to the first battery model layer and the second battery model layer whose similarity is greater than or equal to the preset threshold, and the model parameters corresponding to these layers. In the embodiment of the present application, it can be understood as the common part of the two models used to evaluate the battery aging law, which is used for integration during model fusion to obtain a better prediction model.

[0080] Specifically, in order to effectively fuse the first battery model and the second battery model, it is necessary to first extract the layers with similar structures and parameters in the two models as fitting objects. The network structures of the two models are analyzed, and the parameter similarity of each layer is calculated in turn. The layers with similarity greater than or equal to the preset threshold and their corresponding parameters are determined as the first fitting layer objects. The purpose of this is that the similar layers reflect the common parts of the two models in modeling the battery aging law. Taking the similar layers as the fusion objects can retain the knowledge that is effective for battery life prediction to the greatest extent and improve the reliability of model fusion.

[0081] Step 217: performing weighted summation on the layers and parameters corresponding to the layers whose similarities in the first battery model and the second battery model are less than the preset similarity, to obtain a second fitting layer object.

[0082] Among them, the second fitting layer object refers to the first battery model layer and the second battery model layer whose similarity is less than a preset threshold, and the model parameters corresponding to these layers after weighted summation. In the embodiment of the present application, it can be understood as the heterogeneous parts of the two models used to evaluate the battery aging law, which are used for integration during model fusion to obtain a better prediction model.

[0083] Specifically, in order to effectively fuse the different knowledge of the first battery model and the second battery model, it is necessary to extract the heterogeneous layers in the two models as fitting objects, and perform weighted summation on the non-similar layers and their parameters in the two models whose similarity is less than the preset threshold. The weight coefficients of the weighted summation are added to 1, and the second fitting layer object is calculated. The purpose of this is that the different layers contain the supplementary knowledge of the two models for battery life prediction. Weighted fusion of the parameters of the non-similar layers can merge the different modeling perspectives of the two models. Compared with direct splicing, weighted summation can balance the contribution of the two models in the new model, avoid the knowledge of a certain model dominating the new model, and improve the stability of fusion.

[0084] Step 218: construct an initial battery life prediction model based on the first fitting layer object and the second fitting layer object.

[0085] Specifically, in order to obtain a life prediction model that combines the advantages of the two models and predicts the battery health status more accurately, it is necessary to build a model based on the extracted fitting layer objects. The first fitting layer object and the second fitting layer object are reorganized in the order of the original model structure to form the network framework of the initial battery life prediction model. And load the parameters in the two fitting objects into the corresponding layers to complete the model construction. The purpose of this is that the first fitting object provides the common part of the model for modeling the battery aging law, and the second fitting object provides the fusion of the heterogeneous knowledge of the two models. The rational use of the two types of fitting objects can build a new model that combines the advantages of the two models.

[0086] Step 103: Determine a plurality of second energy storage batteries of a different battery type from the first energy storage battery, and obtain second battery operation data and battery life of each second energy storage battery.

[0087] Among them, battery type refers to the different classifications of product attributes such as battery material composition and structural characteristics. In the embodiment of the present application, it can be understood as different battery products such as nickel-hydrogen batteries and lithium batteries, which are used to distinguish the differences in battery working mechanisms and aging modes.

[0088] The second battery operation data refers to performance parameters collected during the use of multiple second batteries that are different from the first battery type. In the embodiment of the present application, it can be understood as voltage, current, temperature and other data of different types of batteries, which are used to train the model to evaluate the health status of different types of batteries.

[0089] Battery life refers to the entire life cycle of a battery from the time it leaves the factory to the time it is scrapped. In the embodiments of the present application, it can be understood as the actual length of use of different types of batteries, which is used as label data for different batteries and the ability to train the model to predict battery life.

[0090] Specifically, multiple second batteries of different types from the first battery are determined. These second batteries can be batteries of different materials and different manufacturers. During the actual use of the second battery, its operating status is detected, and operating data, including voltage, current, temperature and other information, is obtained. The entire service life cycle of these second batteries from new to scrapped is counted. The purpose of this is that the operating mechanisms and aging modes of different types of batteries are different. Obtaining comprehensive data of multiple batteries can increase the diversity of training samples and improve the model's adaptability to new environments.

[0091] Step 104: Use the second battery operation data and the battery life as source domain data, adjust the model parameters of the initial battery life prediction model through the transfer learning algorithm and the source domain data, and obtain the target battery life prediction model; based on the target battery life prediction model, predict the target battery life of the first energy storage battery.

[0092] Among them, source domain data refers to sample data used for pre-training models in transfer learning. In the embodiment of the present application, it can be understood as operating parameters and life data of multiple second batteries, which is used to provide a rich data source required for model generalization capabilities.

[0093] The transfer learning algorithm refers to a machine learning method that enables the model to utilize source domain knowledge to improve the performance of target domain tasks. In the embodiments of the present application, it can be understood as algorithms such as model fine-tuning and sub-network activation, which are used to enhance the model's adaptability to new batteries through source domain data.

[0094] Model parameters refer to the adjustable variables such as coefficients and weights of each layer that constitute the machine learning model. In the embodiment of the present application, they can be understood as parameters such as connection weights and biases in the neural network, which are used to define the learning ability of the model.

[0095] Specifically, the collected operating data of various second batteries and the corresponding battery life constitute the source domain data in transfer learning. Then, based on these source domain data, the transfer learning algorithm is applied to adjust the parameters of the initial battery life prediction model through cross-domain training, sub-network activation, etc., to obtain a battery life prediction model suitable for the target domain. The purpose of this is that transfer learning can reduce the dependence on target domain data, and the model is pre-trained using the data in the source domain to adapt the model to the characteristic distribution of new batteries. The source domain data of various types of batteries provide the model with rich prior knowledge, which helps the model to quickly adapt to new batteries and achieve better generalization performance. Based on the tuned target domain model, the remaining life of the first battery can be predicted more accurately, and it can be reliably applied to the life evaluation of various new batteries.

[0096] Based on the above embodiment, as an optional embodiment, in step 104: adjusting the model parameters of the initial battery life prediction model by transferring the learning algorithm and the source domain data, this step may also include the following steps: Step 301: constructing a migration model for each second energy storage battery based on each second battery operation data and each battery life in the source domain data.

[0097] Among them, the migration model refers to a model trained using the corresponding source domain data for a specific battery type. In the embodiment of the present application, it can be understood as an exclusive model built based on the data of a certain type of second battery, which is used to extract the usage characteristics of this type of battery and improve model adaptability.

[0098] Specifically, the operating data and corresponding life labels of a certain type of second battery are extracted from the collected source domain data, and a migration model exclusive to this type of battery is constructed based on this data. The purpose of doing this is that the usage characteristics of different battery types are not exactly the same. Using data from a specific type of battery can train a model that is more sensitive to this type of battery. Compared with a unified model, this personalized migration model can more accurately extract the usage characteristics of the battery, making the life prediction more accurate for a specific battery type. A series of migration models optimized for the data distribution and usage characteristics of different battery types are obtained, which provides a basis for subsequent model fusion and improves the adaptability to multiple types of new batteries.

[0099] Step 302: Train each migration model according to the target domain data to obtain the model parameters to be migrated after the migration model is trained.

[0100] Among them, the model parameters to be migrated refer to the model parameters of each migration model after training and adjustment using the target domain data. In the embodiment of the present application, they can be understood as the parameters of various battery migration models that are adapted to the first battery data, and are used to provide candidate parameters that are adapted to the target domain for building an integrated model.

[0101] Specifically, for each migration model obtained by pre-training, further training is performed using the target domain operation data of the first battery, and the model parameters are adjusted to obtain a series of parameters of the model to be migrated that are adapted to the target domain. The purpose of this is that the target domain data can help the model capture the real usage characteristics of the target battery and prevent the model from being overly dependent on the source domain distribution. By fine-tuning each migration model using the first battery data, a set of model parameters that conform to both the source domain and the target domain can be obtained, providing candidates for subsequent parameter selection. Multiple sets of parameters to be migrated that are adapted to the target domain lay the foundation for building a more generalized model for new batteries.

[0102] Step 303: Determine the target model parameters among the model parameters to be migrated, and adjust the model parameters of the layer to be migrated in the initial battery life prediction model to the target model parameters.

[0103] Among them, the target model parameters refer to the best parameters selected from the various model parameters to be migrated. In the embodiment of the present application, they can be understood as the model parameters with the best effect in the target domain, which are used to load into the initial prediction model to improve its adaptability to the target domain.

[0104] The layer to be migrated refers to the layer in the initial prediction model whose parameters can be adjusted. In the embodiment of the present application, it can be understood as a trainable layer such as a fully connected layer and a convolutional layer, which is used to achieve adaptation of the model to the target domain through parameter replacement.

[0105] Specifically, set parameter selection conditions, such as minimizing errors, and select target model parameters that meet the conditions from the parameters to be migrated after adjustment of each migration model. And load the target model parameters to the fine-tunable layer in the initial prediction model to complete the model parameter migration. The purpose of doing this is that the optimization results of different models may not be exactly the same. It is necessary to select the best parameters that meet the characteristics of the target domain. By selecting the target model parameters with the best effect, a prediction model that absorbs the advantages of each model and conforms to the target domain distribution can be obtained. Parameter migration adjustment can make the initial model parameters further adapt to the characteristics of the target new battery while retaining the original advantages, and realize effective generalized life prediction for multiple types of batteries. Based on the above embodiment, as an optional embodiment, in step 303: determining the target model parameters among the model parameters to be migrated, this step may also include the following steps: Step 401: randomly generate multiple model parameter combinations among the model parameters to be migrated, and use the model parameter combinations as multiple populations.

[0106] Among them, the model parameter combination refers to a parameter mixture randomly selected from various model parameters to be migrated. In the embodiment of the present application, it can be understood as a parameter subset composed of different migration model parameters, which is used as a population individual to search for the optimal parameter solution.

[0107] A population refers to a set of candidate solutions stored in evolutionary optimization, and in the embodiments of the present application, it can be understood as a set of multiple model parameter combinations, which is used to represent different search areas in the parameter space.

[0108] Specifically, a certain number of parameter subsets are randomly selected from the parameters to be migrated after each migration model training and optimization, and multiple parameter combinations are constructed, and each combination constitutes a population. The purpose of this is that random combinations can obtain more parameter options, which is conducive to searching for the global optimal solution. Multiple populations are generated by random parameter combinations, which expands the optimization parameter range and provides more possibilities for finding the model parameters that best suit the target domain. The generation of multiple populations lays the foundation for the subsequent use of evolutionary algorithms to search for the global optimal parameters.

[0109] Step 402: Calculate the fitness of various populations and the target domain data, and use the population corresponding to the maximum fitness among various populations as the target model parameter.

[0110] Among them, fitness refers to an indicator for evaluating the performance of individuals in a population in a target domain. In the embodiment of the present application, it can be understood as the degree of effect of population parameters on target domain data, which is used to guide the selection of parameters in the target domain.

[0111] Specifically, load the model parameters corresponding to the population, test the performance on the first battery data in the target domain, calculate the fitness indicators of the population individuals and the target domain, such as minimizing the loss, and then select the population with the highest fitness among all populations, that is, the population with the best effect in the target domain, and its corresponding parameter combination is the target model parameter. The purpose of this is that fitness evaluation can guide the fitting effect of the parameters in the target domain and help find the optimal parameters. Selecting the population parameters with the highest fitness as the target model parameters can obtain an integrated parameter solution that optimizes the target domain. Fitness calculation can make the parameter selection process conform to the target domain distribution and achieve highly generalized life prediction for new batteries.

[0112] Reference Figure 2 , a battery life prediction system based on transfer learning provided by an embodiment of the present application, the system comprises: a data acquisition module, a model determination module, a model migration model, and a battery life prediction module, wherein: A data acquisition module, used to acquire first battery operation data of the first energy storage battery, vehicle driving data of the corresponding vehicle, and environmental data of the environment in which the vehicle is located; A model determination module is used to determine the target domain data of the first energy storage battery according to the first battery operation data, the vehicle driving data and the environmental data, and to construct an initial battery life prediction model corresponding to the target domain data; determine a plurality of second energy storage batteries of different battery types from the first energy storage battery, and obtain second battery operation data and battery life of each second energy storage battery; A model migration module, used to use each second battery operation data and each battery life as source domain data, and adjust the model parameters of the initial battery life prediction model through a migration learning algorithm and the source domain data to obtain a target battery life prediction model; The battery life prediction module is used to predict the target battery life of the first energy storage battery based on the target battery life prediction model.

[0113] Based on the above embodiments, the model determination module is also used to extract battery aging characteristics from the first battery operation data; obtain the first target domain characteristics based on the battery aging characteristics, the vehicle mileage in the vehicle driving data, and the driving time in the vehicle driving data; obtain the second target domain characteristics based on the battery aging characteristics, the ambient temperature in the environmental data, and the ambient humidity in the environmental data; and use the first target domain characteristics and the second target domain characteristics as target domain data.

[0114] Based on the above embodiments, the model determination module is also used to determine the voltage floating rate according to the operating voltage and the preset life cycle; determine the current floating rate according to the operating current and the preset life cycle; determine the impedance loss rate according to the battery impedance and the preset life cycle; if there is a feature greater than a preset threshold value among the voltage floating rate, the current floating rate and the impedance loss rate, the feature is determined to be a battery aging feature.

[0115] On the basis of the above embodiments, the model determination module is also used to determine the first training data set according to the first target domain feature and the timestamp of the first target domain feature, and determine the second training data set according to the second target domain feature and the timestamp of the second target domain feature; obtain the first battery life label corresponding to the first training data set and the second battery life label corresponding to the second training data set; based on the first training data set and the first battery life label, train the initial neural network model until the preset model training stop condition is reached to obtain the first battery model; based on the second training data set and the second battery life label, train the initial neural network model until the preset model training stop condition is reached to obtain the second battery model, and the preset model training stop condition includes that the model training times reach the preset times or the error rate of the model output result is less than the preset error rate; fit the first battery model and the second battery model to obtain the initial battery life prediction model.

[0116] Based on the above embodiments, the model determination module is also used to use the layers and parameters corresponding to the layers in the first battery model and the second battery model whose similarity is greater than or equal to the preset similarity as the first fitting layer object; perform weighted summation on the layers and parameters corresponding to the layers in the first battery model and the second battery model whose similarity is less than the preset similarity to obtain the second fitting layer object; and construct an initial battery life prediction model based on the first fitting layer object and the second fitting layer object.

[0117] On the basis of the above embodiments, the model migration module is also used to construct a migration model of each second energy storage battery based on the second battery operation data and each battery life in the source domain data; train each migration model according to the target domain data to obtain the model parameters to be migrated after each migration model is trained; determine the target model parameters in each model parameter to be migrated, and adjust the model parameters of the layer to be migrated in the initial battery life prediction model to the target model parameters.

[0118] Based on the above embodiments, the model migration module is also used to randomly generate multiple model parameter combinations among the model parameters to be migrated, and use each model parameter combination as multiple populations; calculate the fitness of each population with the target domain data, and use the population corresponding to the maximum fitness among the various populations as the target model parameters.

[0119] It should be noted that: when the device provided in the above embodiment realizes its function, only the division of the above functional modules is used as an example. In actual application, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0120] The present application also discloses an electronic device. Figure 3 , Figure 3 The electronic device 300 may include: at least one processor 301 , at least one network interface 304 , a user interface 303 , a memory 305 , and at least one communication bus 302 .

[0121] The communication bus 302 is used to realize the connection and communication between these components.

[0122] The user interface 303 may include a display interface and a camera interface. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0123] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0124] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface diagrams and applications, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0125] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may optionally also be at least one storage device located away from the aforementioned processor 301. Refer to Figure 3 , the memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program of a battery life prediction method based on transfer learning.

[0126] exist Figure 3 In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call the application program storing a battery life prediction method based on transfer learning in the memory 305, and when executed by one or more processors 301, the electronic device 300 executes one or more methods in the above-mentioned embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0127] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as 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 mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0129] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0131] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0132] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0133] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not recorded in the present disclosure. The description and examples are to be regarded as exemplary only.

Claims

1. A battery life prediction method based on transfer learning, characterized in that: include: Acquire first battery operation data of a first energy storage battery, vehicle driving data of a corresponding vehicle, and environmental data of the environment in which the vehicle is located; Determine target domain data of the first energy storage battery according to the first battery operation data, the vehicle driving data and the environmental data, and construct an initial battery life prediction model corresponding to the target domain data; Determine a plurality of second energy storage batteries of a different battery type from the first energy storage battery, and obtain second battery operation data and battery life of each of the second energy storage batteries; Using each of the second battery operation data and each of the battery life as source domain data, adjusting the model parameters of the initial battery life prediction model through a transfer learning algorithm and the source domain data, to obtain a current battery life prediction model; Based on the target battery life prediction model, a target battery life of the first energy storage battery is predicted.

2. The method for predicting battery life based on transfer learning according to claim 1, characterized in that: The determining the target domain data of the first energy storage battery according to the first battery operation data, the vehicle driving data and the environmental data includes: extracting battery aging characteristics from the first battery operation data; Obtaining a first target domain feature according to the battery aging feature, the vehicle mileage in the vehicle driving data, and the driving time in the vehicle driving data; Obtaining a second target domain feature according to the battery aging feature, the ambient temperature in the environmental data, and the ambient humidity in the environmental data; The first target domain feature and the second target domain feature are used as the target domain data.

3. The method for predicting battery life based on transfer learning according to claim 2, characterized in that: The first battery operation data includes an operation voltage, an operation current, and a battery impedance, and extracting a battery aging feature from the first battery operation data includes: Determining a voltage floating rate according to the operating voltage and a preset life cycle; Determining a current floating rate according to the operating current and the preset life cycle; Determining an impedance loss rate according to the battery impedance and the preset life cycle; If there is a feature among the voltage floating rate, the current floating rate, and the impedance loss rate that is greater than a preset threshold, it is determined that the feature is a battery aging feature.

4. The method for predicting battery life based on transfer learning according to claim 2, characterized in that: The constructing the initial battery life prediction model corresponding to the target domain data includes: Determine a first training data set according to the first target domain feature and the timestamp of the first target domain feature, and determine a second training data set according to the second target domain feature and the timestamp of the second target domain feature; Obtaining a first battery life label corresponding to the first training data set and a second battery life label corresponding to the second training data set; Based on the first training data set and the first battery life label, training the initial neural network model until a preset model training stop condition is reached to obtain a first battery model; Based on the second training data set and the second battery life label, the initial neural network model is trained until the preset model training stop condition is reached to obtain a second battery model, wherein the preset model training stop condition includes that the number of model training times reaches a preset number of times or the error rate of the model output result is less than a preset error rate; The first battery model and the second battery model are fitted to obtain the initial battery life prediction model.

5. The method for predicting battery life based on transfer learning according to claim 4, characterized in that: The fitting of the first battery model and the second battery model to obtain the initial battery life prediction model includes: taking the layers whose similarity in the first battery model and the second battery model is greater than or equal to a preset similarity and the parameters corresponding to the layers as the first fitting layer object; Performing weighted summation on the layers whose similarity in the first battery model and the second battery model is less than the preset similarity and the parameters corresponding to the layers to obtain a second fitting layer object; The initial battery life prediction model is constructed according to the first fitting layer object and the second fitting layer object.

6. The method for predicting battery life based on transfer learning according to claim 1, characterized in that: The adjusting the model parameters of the initial battery life prediction model by using the transfer learning algorithm and the source domain data includes: constructing a migration model of each of the second energy storage batteries based on each of the second battery operation data and each of the battery lifespans in the source domain data; According to the target domain data, each of the migration models is trained to obtain the model parameters to be migrated after the migration model is trained; The target model parameters in each of the model parameters to be migrated are determined, and the model parameters of the layer to be migrated in the initial battery life prediction model are adjusted to the target model parameters.

7. The method for predicting battery life based on transfer learning according to claim 6, characterized in that: The determining of the target model parameters among the model parameters to be migrated includes: Randomly generate a plurality of model parameter combinations among the model parameters to be migrated, and use the model parameter combinations as a plurality of populations; The fitness of each of the populations and the target domain data is calculated, and the population corresponding to the maximum fitness among the populations is used as the target model parameter.

8. A battery life prediction system based on transfer learning, characterized in that: The system comprises: A data acquisition module, used to acquire first battery operation data of the first energy storage battery, vehicle driving data of the corresponding vehicle, and environmental data of the environment in which the vehicle is located; a model determination module, configured to determine the target domain data of the first energy storage battery according to the first battery operation data, the vehicle driving data and the environmental data, and to construct an initial battery life prediction model corresponding to the target domain data; determine a plurality of second energy storage batteries of different battery types from the first energy storage battery, and obtain second battery operation data and battery life of each of the second energy storage batteries; A model migration module, used to use each of the second battery operation data and each of the battery life as source domain data, and adjust the model parameters of the initial battery life prediction model through a migration learning algorithm and the source domain data to obtain a current battery life prediction model; A battery life prediction module is used to predict the target battery life of the first energy storage battery based on the target battery life prediction model.

9. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the battery life prediction method based on transfer learning as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the battery life prediction method based on transfer learning as described in any one of claims 1-7 is executed.

Images