Method and device for predicting service life of power battery

Abstract

The invention provides a power battery life prediction method and prediction device. The method comprises the following steps: acquiring charge and discharge data of a target battery; extracting at least two battery capacity characteristics according to the charging and discharging data; fusing the at least two battery capacity characteristics to obtain a target battery capacity characteristic; and inputting the target battery capacity characteristics into a pre-trained life prediction model, and determining the residual life of the target battery. At least two battery capacity characteristics are extracted from original charging and discharging data, so that the limitation of traditional single-parameter analysis can be broken through, and the characterization capability of a microscopic aging mechanism is remarkably enhanced; and fusing the at least two battery capacity features to obtain the target battery capacity feature, so that the life prediction model can mine the spatial correlation between the features to obtain a more accurate life prediction result.

Description

Technical Field

[0001] The present application relates to the technical field of power batteries, and in particular to a method and device for predicting the life of a power battery. Background Art

[0002] Existing battery life prediction methods often rely on a single electrochemical parameter, making it difficult to capture microscopic aging mechanisms. Furthermore, traditional BMS systems are limited by sampling accuracy (≥1mV / 1°C), resulting in a high failure rate of up to 78% for detecting voltage anomalies as small as 0.1mV caused by early micro-shorts. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method and device for predicting the life of a power battery, so as to solve the problem that the relevant technology is difficult to capture microscopic anomalies and has low accuracy in life prediction.

[0004] An embodiment of the present application provides a method for predicting the life of a power battery, the method comprising: Obtain the charge and discharge data of the target battery; extracting at least two battery capacity characteristics based on the charge and discharge data; fusing the at least two battery capacity characteristics to obtain a target battery capacity characteristic; The target battery capacity characteristics are input into a pre-trained life prediction model to determine the remaining life of the target battery.

[0005] Furthermore, the battery capacity characteristics include: a first battery capacity characteristic for characterizing intrinsic electrochemical properties, a second battery capacity characteristic for characterizing kinetic process variations, and a third battery capacity characteristic for characterizing cycle-to-cycle degradation accumulation.

[0006] Furthermore, at least two battery capacity characteristics are extracted based on the charge and discharge data, including: constructing a three-dimensional tensor of cycle number-voltage point-capacity value according to the voltage-capacity sequence data of the constant current discharge stage in the charge and discharge data to obtain a first battery capacity feature; performing a numerical differentiation operation on the first battery capacity characteristic, and determining a second battery capacity characteristic of an electrode phase transition critical point offset according to the differential capacity value at each voltage point in each cycle; Based on the first battery capacity characteristic, a capacity difference at the same voltage point between adjacent cycles is determined to obtain a third battery capacity characteristic.

[0007] Furthermore, the at least two battery capacity characteristics are fused to obtain a target battery capacity characteristic, including: Convert each battery capacity characteristic into a battery capacity characteristic graph in a unified format; The converted battery capacity characteristic graphs of each type are stacked along the channel dimension of the life prediction model, and feature processing is performed on the stacking results to obtain the target battery capacity characteristics.

[0008] Furthermore, after acquiring the charge and discharge data of the target battery, the prediction method further includes: Performing time domain filtering on the charge and discharge data to obtain filtered charge and discharge data; The filtered charge and discharge data of different cycle periods are aligned using a dynamic time warping algorithm.

[0009] Furthermore, the life prediction model is an improved ResNet50 backbone network; the life prediction model includes a convolutional layer, a global average pooling layer and a fully connected layer; the convolutional layer includes a shallow convolutional layer in which an attention module is embedded.

[0010] Furthermore, the method of training the life prediction model includes: Load the pre-trained weights into the constructed original lifespan prediction model; The model parameters of the shallow network structure in the original life prediction model are fixed, and the high-level network in the original life prediction model is fine-tuned and trained using a dynamic learning rate scheduling method to obtain the trained life prediction model.

[0011] The present application also provides a device for predicting the life of a power battery, the device comprising: An acquisition module is used to obtain the charge and discharge data of the target battery; an extraction module, configured to extract at least two battery capacity characteristics based on the charge and discharge data; a fusion module, configured to fuse the at least two battery capacity characteristics to obtain a target battery capacity characteristic; The prediction module is used to input the target battery capacity characteristics into a pre-trained life prediction model to determine the remaining life of the target battery.

[0012] An embodiment of the present application also provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the prediction method as described above are performed.

[0013] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned prediction method are executed.

[0014] An embodiment of the present application provides a method and device for predicting the life of a power battery, which obtains charge and discharge data of a target battery; extracts at least two battery capacity characteristics based on the charge and discharge data; fuses the at least two battery capacity characteristics to obtain a target battery capacity characteristic; and inputs the target battery capacity characteristic into a pre-trained life prediction model to determine the remaining life of the target battery.

[0015] By extracting at least two battery capacity characteristics from the original charge and discharge data, it is possible to break through the limitations of traditional single-parameter analysis and significantly enhance the ability to characterize microscopic aging mechanisms; by fusing at least two battery capacity characteristics to obtain the target battery capacity characteristics, the life prediction model can explore the spatial correlation between the characteristics and obtain more accurate life prediction results.

[0016] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A flowchart showing a method for predicting the life of a power battery provided in an embodiment of the present application is shown; Figure 2 A schematic diagram showing the structure of a device for predicting the life of a power battery provided in an embodiment of the present application is shown; Figure 3 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for which protection is claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work falls within the scope of protection of the present application.

[0020] Research has found that existing battery life prediction methods rely heavily on a single electrochemical parameter, making it difficult to capture microscopic aging mechanisms. Furthermore, traditional BMS systems are limited by sampling accuracy (≥1mV / 1°C), resulting in a high failure rate of up to 78% for detecting voltage anomalies as small as 0.1mV, which can be caused by early micro-shorts.

[0021] Based on this, an embodiment of the present application provides a method for predicting the life of a power battery to solve the problem that related technologies have difficulty in capturing microscopic anomalies and have low accuracy in life prediction.

[0022] See also Figure 1 , Figure 1 This is a flow chart of a method for predicting the life of a power battery provided in an embodiment of the present application. Figure 1 As shown in , the prediction method provided in the embodiment of the present application includes: S101: Acquire charge and discharge data of a target battery.

[0023] In this step, the vehicle-mounted battery management system may obtain charge and discharge data of the power battery in multiple charge and discharge cycles, where the charge and discharge data includes voltage-capacity data during the charge and discharge process.

[0024] S102: Extract at least two battery capacity characteristics based on the charge and discharge data.

[0025] In this step, at least the battery capacity feature is extracted from the charge and discharge data through feature engineering.

[0026] Among them, the battery capacity characteristics include: a first battery capacity characteristic for characterizing the intrinsic electrochemical characteristics, a second battery capacity characteristic for characterizing the abnormality of the kinetic process, and a third battery capacity characteristic for characterizing the accumulation of decay between cycles.

[0027] Intrinsic electrochemical characteristics refer to the inherent physical and chemical properties within a battery that are directly related to electrochemical reactions. Kinetic process anomalies refer to abnormal changes in key kinetic parameters related to ion transport, electron conduction, and chemical reaction rates during the battery's charge and discharge processes. These changes are typically caused by degradation of the battery's internal microstructure and are directly reflected in the decline in battery performance. Cycle-to-cycle degradation refers to the cumulative effect of performance degradation caused by gradual degradation of the battery's internal microstructure over multiple charge and discharge cycles. This cumulative effect is an important indicator for predicting the remaining useful life (RUL).

[0028] In one example, the first battery capacity characteristic is a capacity-voltage curve, the second battery capacity characteristic is an incremental capacity curve, and the third battery capacity characteristic is a capacity difference curve.

[0029] In a possible implementation, step S102 may include: S1021. Construct a three-dimensional tensor of cycle number-voltage point-capacity value based on the voltage-capacity sequence data of the constant current discharge stage in the charge and discharge data to obtain a first battery capacity feature.

[0030] In this step, by extracting the voltage-capacity sequence data in different charge and discharge cycles during the constant current discharge stage, a three-dimensional tensor of cycle number-voltage point-capacity value is constructed.

[0031] In practice, continuous, time-stamped raw sequence data is organized into a structured tensor indexed by cycle, voltage, and capacity. The first dimension of this three-dimensional tensor is the cycle number, representing the Nth charge-discharge cycle the battery has undergone, reflecting the evolution of battery performance over time / lifecycle. The second dimension is the voltage point, representing the discrete points along the voltage axis where data was collected during a single charge-discharge cycle, reflecting the battery's behavior at different states of charge. The third dimension is the capacity value, representing the capacity measured at a specific voltage point in a specific cycle.

[0032] S1022: Perform a numerical differentiation operation on the first battery capacity characteristic, and determine a second battery capacity characteristic of electrode phase transition critical point offset according to the differential capacity value at each voltage point in each cycle.

[0033] In this step, based on the first battery capacity characteristic, a numerical differentiation operation can be performed on each cycle. Specifically, the capacity Q is differentiated with respect to the voltage V (dQ / dV). This yields the differential capacity value at each voltage point in each cycle, forming a differential capacity curve. The peak characteristics in the differential capacity curve are then extracted to investigate the shift characteristics of the electrode phase transition critical point during cycling. Generally speaking, as the battery ages, the dQ / dV curve may exhibit a peak shift, reflecting the shift of the electrode phase transition point.

[0034] S1023. Determine the capacity difference at the same voltage point between adjacent cycles based on the first battery capacity characteristic to obtain a third battery capacity characteristic.

[0035] In this step, based on the first battery capacity characteristic, the capacity difference between adjacent cycles at the same SOC point can be calculated to quantify the capacity difference curve, i.e., the third battery capacity characteristic, to quantify the active lithium loss rate. The F3 curve of the third battery capacity characteristic quantifies the capacity difference between adjacent cycles at the same SOC point, enabling dynamic tracking of the accumulated degradation between cycles. Experiments have shown that abnormally degraded batteries can be identified 300 cycles in advance, with a false alarm rate of ≤3.5%.

[0036] The composite graphical feature set constructed in the examples of this application breaks through the limitations of traditional single-parameter analysis by integrating three types of features: the capacity-voltage curve (the first battery capacity feature F1), the incremental capacity curve (the second battery capacity feature F2), and the capacity difference curve (the third battery capacity feature F3). This model incorporates the coupling effect of the electrode phase transition critical point offset (F2) and the active lithium loss rate (F3), significantly enhancing the ability to characterize microscopic aging mechanisms such as SEI film thickening and lithium dendrite growth. Experiments have shown that the prediction error can be reduced by more than 60% compared to traditional capacity decay models.

[0037] S103: Fusing the at least two battery capacity characteristics to obtain a target battery capacity characteristic.

[0038] In this step, at least two battery capacity features are fused to construct a target battery capacity feature, so that the subsequent life prediction model can mine the spatial correlation between the features.

[0039] In a possible implementation, step S103 may include: S1031. Convert each battery capacity characteristic into a battery capacity characteristic graph in a unified format.

[0040] First, spatial mapping is performed to convert each selected battery capacity feature into a 100×100 pixel grayscale image, and the grayscale value is normalized to the [0,255] interval.

[0041] S1032: stacking each converted battery capacity feature graph along the channel dimension of the life prediction model, and performing feature processing on the stacking result to obtain the target battery capacity feature.

[0042] In this step, the three battery capacity feature maps are stacked along the channel dimension, generating a standard 224×224×3 input format, with each pixel containing multidimensional degradation information. The stacked result is then subjected to feature processing, using a bicubic interpolation algorithm to preserve the feature topology and an edge-filling strategy to minimize loss of valid information.

[0043] In this way, combined with three-channel fusion technology (224×224×3), the model can simultaneously capture the spatial correlation characteristics of electrochemical intrinsic characteristics (F1), kinetic changes (F2) and decay accumulation (F3). In the experiment, the sensitivity of micro-short circuit detection is improved to 0.05mV, which is 20 times higher than that of traditional BMS systems.

[0044] In addition, the target battery capacity characteristics can be further integrated with multimodal battery characteristic data such as EIS spectrum and thermal imaging to reflect the target power battery's richer multi-dimensional characteristics, and then input into the life prediction model to obtain more accurate prediction results.

[0045] S104: Input the target battery capacity characteristics into a pre-trained life prediction model to determine the remaining life of the target battery.

[0046] The lifespan prediction model can be a convolutional neural network (CNN). In one example, the lifespan prediction model is a modified ResNet50 backbone network; the lifespan prediction model includes a convolutional layer, a global average pooling layer, and a fully connected layer; the convolutional layer includes a shallow convolutional layer with an embedded attention module.

[0047] Here, the SE attention module is embedded in the shallow convolutional layers (such as the first three convolutional layers). The SE attention module enhances the weight distribution of the voltage inflection point area, improves the feature extraction efficiency by 40%, reduces the cloud training energy consumption by 55%, and enhances the sensitivity to the voltage inflection point area; a global average pooling layer is added before the fully connected layer to suppress the risk of overfitting.

[0048] The method of training the life prediction model includes: Step 1: Load the pre-trained weights into the constructed original lifespan prediction model.

[0049] Step 2: Fix the model parameters of the shallow network structure in the original life prediction model, and use a dynamic learning rate scheduling method to fine-tune the high-level network in the original life prediction model to obtain the trained life prediction model.

[0050] This embodiment of the application uses a transfer learning strategy for model training. First, the ImageNet pre-trained weights are loaded. Then, the parameters of the shallow network structure (e.g., the first three convolutional blocks) are frozen to preserve the general feature extraction capabilities of the pre-trained model. Dynamic learning rate scheduling (e.g., initial lr = 1e-4, decaying by 50% every 10 rounds) is used to fine-tune the high-level network to adapt to the characteristics of the battery degradation pattern. This approach, by freezing the low-level convolutional feature extractor and dynamically tuning the high-level degradation feature adaptation layer, enables efficient transfer of general visual knowledge to battery degradation patterns and accelerates the training process.

[0051] For training data, the embodiment of the present application adopts mirror symmetry and random cropping strategies to expand the original data set by 300%, solving the problem of scarcity of early cycle data and the problem of model overfitting under small sample conditions.

[0052] In experiments, a dataset of 123 NMC battery aging models was divided into 80% training and 20% test sets. An early stopping mechanism (patience = 15) was introduced to prevent overfitting. The resulting model achieved a RMSE of 2.1% (cycle life) and an R² of 0.963 on the test set, representing a 63.2% reduction in error compared to traditional LSTM models.

[0053] Through the above method, the training data requirement of the life prediction model in the embodiment of the present application is reduced by 70%, and the edge model size is compressed to 35MB, meeting the real-time processing requirements of the on-board ECU (10ms level response).

[0054] Furthermore, after step S101, the prediction method further includes: The charge and discharge data are subjected to time domain filtering to obtain filtered charge and discharge data; and the filtered charge and discharge data of different cycle periods are aligned using a dynamic time warping algorithm.

[0055] To further enhance prediction performance, this embodiment of the present application also preprocesses the raw charge and discharge data. Specifically, the raw charge and discharge data is filtered in the time domain, using a Savitzky-Golay filter to eliminate high-frequency noise, and the sampling rate is increased to 1Hz. The dynamic time warping (DTW) algorithm is then used to align data from different cycles to eliminate systematic errors introduced by differences in charge and discharge depth.

[0056] In one embodiment, the prediction method provided by the present invention seamlessly integrates with existing systems and can be deployed via an OTA update. For example, a lightweight prediction model can be deployed to meet the real-time monitoring needs of an onboard battery management system (BMS). During vehicle operation, charge and discharge voltage-capacity data is acquired at a 1Hz sampling rate. A Savitzky-Golay filter is used to eliminate high-frequency noise, and a dynamic time warping (DTW) algorithm is used to align data from different cycles to eliminate offset errors caused by SOC differences. In the vehicle's ECU, the three curves are converted into 100×100 pixel grayscale images, which are then bicubic interpolated to generate a 224×224×3 input tensor. A pre-trained ResNet50 improved model (with an embedded SE attention module) is loaded to achieve 10ms-level lifetime prediction with an RMSE of ≤2.5%. Compared to traditional BMS systems, the micro-short detection sensitivity is increased to 0.05mV, shortening the warning response time by 90%, meeting ASIL-B functional safety requirements.

[0057] For example, a cloud-edge collaborative architecture was built, with the cloud-based battery health management platform enabling multi-vehicle data aggregation and model iterative optimization. Data augmentation: Mirror symmetry and random cropping were applied to 123 sets of NMC battery data, expanding the dataset by 300% to cover various operating temperatures (-20°C to 55°C) and charge / discharge rates (0.5°C to 3°C). Using a transfer learning strategy, pre-trained ImageNet weights were loaded, the first three convolutional blocks of ResNet50 were frozen, the learning rate was dynamically adjusted (initially 1e-4, decaying by 50% every 10 epochs), and the last four convolutional blocks were fine-tuned to adapt to battery characteristics. For model deployment, the fully connected layers were replaced with global average pooling layers, reducing model parameters by 82% and requiring only 35MB of storage space at the edge. The cloud platform received feature images from 100,000 vehicles and continuously optimized the model, increasing the R² metric from 0.92 to 0.963. The cloud platform can support concurrent analysis of millions of battery packs, maintaining a stable lifetime prediction error within 2.1%, and reducing cloud-based training energy consumption by 55%.

[0058] Furthermore, this prediction method can be applied to cascade recycling sorting systems, enabling the sorting of retired batteries and accurate assessment of their state of health (SOH). For early prediction, based on data from the first 100 cycles, the inter-cycle capacity decay slope is quantified using the F3 curve. Combined with a CNN model, this method predicts the remaining cycle life, enabling identification of abnormal batteries 300 cycles in advance (with a false positive rate ≤ 3.5%). EIS spectrum data is fused with the F1-F3 curves to construct a four-channel input image (224×224×4). An improved TCNN model is used to extract multi-band impedance features, improving SOH assessment accuracy to 95%. Based on the prediction results, batteries are classified into three categories: A (SOH ≥ 80%), B (60%-79%), and C (<60%). Monte Carlo simulation is used to optimize the sorting threshold, increasing the return on cascade recycling by 30%. This method achieves a sorting efficiency of 2,000 batteries per hour, reducing costs by 70% compared to manual inspection. The carbon footprint of recycled batteries per GWh is reduced by 12,000 tons of CO2, demonstrating significant economic benefits.

[0059] An embodiment of the present application provides a method for predicting the life of a power battery, which comprises obtaining charge and discharge data of a target battery; extracting at least two battery capacity characteristics based on the charge and discharge data; fusing the at least two battery capacity characteristics to obtain a target battery capacity characteristic; and inputting the target battery capacity characteristic into a pre-trained life prediction model to determine the remaining life of the target battery.

[0060] By extracting at least two battery capacity characteristics from the original charge and discharge data, it is possible to break through the limitations of traditional single-parameter analysis and significantly enhance the ability to characterize microscopic aging mechanisms; by fusing at least two battery capacity characteristics to obtain the target battery capacity characteristics, the life prediction model can explore the spatial correlation between the characteristics and obtain more accurate life prediction results. See also Figure 2 , Figure 2 This is a schematic diagram of the structure of a power battery life prediction device provided by an embodiment of the present application. Figure 2 As shown in , the prediction device 200 includes: An acquisition module 210 is used to acquire charge and discharge data of a target battery; an extraction module 220, configured to extract at least two battery capacity characteristics based on the charge and discharge data; a fusion module 230, configured to fuse the at least two battery capacity characteristics to obtain a target battery capacity characteristic; The prediction module 240 is configured to input the target battery capacity characteristics into a pre-trained life prediction model to determine the remaining life of the target battery.

[0061] Furthermore, the battery capacity characteristics include: a first battery capacity characteristic for characterizing intrinsic electrochemical properties, a second battery capacity characteristic for characterizing kinetic process variations, and a third battery capacity characteristic for characterizing cycle-to-cycle degradation accumulation.

[0062] Furthermore, when the extraction module 220 is used to extract at least two battery capacity characteristics based on the charge and discharge data, the extraction module 220 is used to: constructing a three-dimensional tensor of cycle number-voltage point-capacity value according to the voltage-capacity sequence data of the constant current discharge stage in the charge and discharge data to obtain a first battery capacity feature; performing a numerical differentiation operation on the first battery capacity characteristic, and determining a second battery capacity characteristic of an electrode phase transition critical point offset according to the differential capacity value at each voltage point in each cycle; Based on the first battery capacity characteristic, a capacity difference at the same voltage point between adjacent cycles is determined to obtain a third battery capacity characteristic.

[0063] Furthermore, when the fusion module 230 is used to fuse the at least two battery capacity characteristics to obtain the target battery capacity characteristics, the fusion module 230 is used to: Convert each battery capacity characteristic into a battery capacity characteristic graph in a unified format; The converted battery capacity characteristic graphs of each type are stacked along the channel dimension of the life prediction model, and feature processing is performed on the stacking results to obtain the target battery capacity characteristics.

[0064] Furthermore, the prediction device further includes a preprocessing module; the preprocessing module is used to: Performing time domain filtering on the charge and discharge data to obtain filtered charge and discharge data; The filtered charge and discharge data of different cycle periods are aligned using a dynamic time warping algorithm.

[0065] Furthermore, the life prediction model is an improved ResNet50 backbone network; the life prediction model includes a convolutional layer, a global average pooling layer and a fully connected layer; the convolutional layer includes a shallow convolutional layer in which an attention module is embedded.

[0066] Furthermore, the prediction device further includes a training module; the training module is used to: Load the pre-trained weights into the constructed original lifespan prediction model; The model parameters of the shallow network structure in the original life prediction model are fixed, and the high-level network in the original life prediction model is fine-tuned and trained using a dynamic learning rate scheduling method to obtain the trained life prediction model.

[0067] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 3 As shown in FIG, the electronic device 300 includes a processor 310 , a memory 320 and a bus 330 .

[0068] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 is running, the processor 310 communicates with the memory 320 through the bus 330. When the machine-readable instructions are executed by the processor 310, the steps of the power battery life prediction method in the above-mentioned method embodiment can be executed. The specific implementation method can be found in the method embodiment and will not be repeated here.

[0069] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for predicting the life of a power battery as in the above-mentioned method embodiment can be executed. For specific implementation methods, please refer to the method embodiment and will not be described in detail here.

[0070] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0071] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. There may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0072] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0073] 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.

[0074] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0075] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The scope of protection of the present application is not limited thereto. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present application, or perform equivalent replacements for some of the technical features thereof. These modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for predicting the life of a power battery, characterized in that: The prediction method comprises: Obtain the charge and discharge data of the target battery; extracting at least two battery capacity characteristics based on the charge and discharge data; fusing the at least two battery capacity characteristics to obtain a target battery capacity characteristic; The target battery capacity characteristics are input into a pre-trained life prediction model to determine the remaining life of the target battery.

2. The prediction method according to claim 1, characterized in that The battery capacity characteristics include: a first battery capacity characteristic for characterizing intrinsic electrochemical properties, a second battery capacity characteristic for characterizing kinetic process variations, and a third battery capacity characteristic for characterizing cycle-to-cycle degradation accumulation.

3. The prediction method according to claim 1 or 2, characterized in that: Extracting at least two battery capacity characteristics based on the charge and discharge data, including: constructing a three-dimensional tensor of cycle number-voltage point-capacity value according to the voltage-capacity sequence data of the constant current discharge stage in the charge and discharge data to obtain a first battery capacity feature; performing a numerical differentiation operation on the first battery capacity characteristic, and determining a second battery capacity characteristic of an electrode phase transition critical point offset according to the differential capacity value at each voltage point in each cycle; Based on the first battery capacity characteristic, a capacity difference at the same voltage point between adjacent cycles is determined to obtain a third battery capacity characteristic.

4. The prediction method according to claim 3, characterized in that The at least two battery capacity characteristics are integrated to obtain a target battery capacity characteristic, including: Convert each battery capacity characteristic into a battery capacity characteristic graph in a unified format; The converted battery capacity characteristic graphs of each type are stacked along the channel dimension of the life prediction model, and feature processing is performed on the stacking results to obtain the target battery capacity characteristics.

5. The prediction method according to claim 1, wherein: After acquiring the charge and discharge data of the target battery, the prediction method further includes: Performing time domain filtering on the charge and discharge data to obtain filtered charge and discharge data; The filtered charge and discharge data of different cycle periods are aligned using a dynamic time warping algorithm.

6. The prediction method according to claim 1, characterized in that The life prediction model is an improved ResNet50 backbone network; the life prediction model includes a convolutional layer, a global average pooling layer and a fully connected layer; the shallow convolutional layer included in the convolutional layer is embedded with an attention module.

7. The prediction method according to claim 1 or 6, characterized in that: The method of training the life prediction model includes: Load the pre-trained weights into the constructed original lifespan prediction model; The model parameters of the shallow network structure in the original life prediction model are fixed, and the high-level network in the original life prediction model is fine-tuned and trained using a dynamic learning rate scheduling method to obtain the trained life prediction model.

8. A device for predicting the life of a power battery, characterized in that: The prediction device comprises: An acquisition module is used to obtain the charge and discharge data of the target battery; an extraction module, configured to extract at least two battery capacity characteristics based on the charge and discharge data; a fusion module, configured to fuse the at least two battery capacity characteristics to obtain a target battery capacity characteristic; The prediction module is used to input the target battery capacity characteristics into a pre-trained life prediction model to determine the remaining life of the target battery.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus. When the processor is running, the machine-readable instructions execute the steps of the method for predicting the life of a power battery as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for predicting the life of a power battery as claimed in any one of claims 1 to 7 are executed.

Images