A battery life prediction method and device, electronic equipment and storage medium
Patent Information
- Application Number
- CN202311664427.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-06
- Publication Date
- 2026-08-18
- Estimated Expiration
- 2043-12-06
AI Technical Summary
[0004]但是为了平衡模型计算效率与预测结果准确度，基于老化机理的电池物理模型存在假设与简化，且使用中需要识别复杂的参数，导致其预测结果在可能出现“边缘工况失真”现象，难以给出定量结果用于直接指导电池管理
[0010]The battery life prediction method of this application addresses the issue that the battery capacity loss curve includes both linearly and non-linearly decreasing components. Since the linearly decreasing component has significant interference factors when predicting battery life, this application not only considers the impact of the IC curve on battery life and extracts the battery's capacity increment characteristics from the IC curve, but also extracts the battery's capacity drop characteristics from the non-linearly decreasing component of the capacity loss curve. Therefore, the battery life prediction model trained on an LSTM model based on at least one battery capacity drop characteristic and capacity increment characteristic has a better prediction effect than one that only considers a single capacity increment characteristic. This is mainly reflected in its high prediction accuracy and high prediction efficiency. By inputting the predicted battery capacity increment and capacity drop characteristics into the trained battery life prediction model, the battery life prediction result can be obtained. Furthermore, since the battery life prediction model in this application is obtained by training an LSTM model, it does not require the participation of a battery physical model, thus eliminating the need to identify complex physical parameters. Even under extreme operating conditions, it can provide accurate prediction results. In addition, this application can use only the data of one battery as the training set during the training process, and can achieve good battery capacity prediction results even under marginal conditions with a small amount of training data required.
Smart Images
Figure CN117930011B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of battery technology, and in particular to a battery life prediction method, apparatus, electronic device, and storage medium. Background Technology
[0002] In the global effort to reduce carbon emissions, electric vehicles have garnered significant attention due to their zero-carbon emission characteristics. Lithium-ion batteries, with their high energy density, are widely used in electric vehicle power batteries. However, lithium-ion batteries suffer from capacity decay, causing the driving range provided by the power battery to decrease over time. To ensure drivers have accurate knowledge of their vehicle's range and alleviate "range anxiety" for owners of new energy vehicles, accurate estimation of the battery's State of Health (SOH), i.e., battery lifespan, is crucial. The SOH decay process is highly nonlinear due to its strong correlation with the accumulation of complex by-reaction products within the battery. Furthermore, the degree of nonlinearity varies with changes in external stimuli, significantly increasing the difficulty of SOH prediction and posing a challenge to the refined management of batteries.
[0003] There are two main approaches to predicting the lifespan of lithium-ion batteries: One is based on physical models, such as SLIDE or PyBAMM models, to simulate the aging process during battery operation and describe the aging process from a physical perspective, thus predicting the lifespan of the lithium-ion battery after aging. The other is based on data-driven approaches, which intuitively extract the latent features of battery aging trends to predict battery lifespan. For example, Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Gaussian Process Regression (GPR) can be combined into data-driven models for battery lifespan prediction.
[0004] However, to balance computational efficiency and prediction accuracy, battery physics models based on aging mechanisms involve assumptions and simplifications, and require the identification of complex parameters during use. This can lead to "edge condition distortion" in predictions, making it difficult to provide quantitative results for direct guidance of battery management. The predictive power of data-driven models depends on the quality and scale of training data, but the amount of battery data for edge conditions such as high and low temperatures is often insufficient, resulting in poor prediction performance under these conditions. Therefore, none of the above solutions can achieve accurate prediction of lithium-ion battery life under edge conditions. Summary of the Invention
[0005] The purpose of this application is to provide a battery life prediction method, device, electronic device, and storage medium, which can accurately predict the life of lithium-ion batteries under edge conditions, thereby achieving refined battery management.
[0006] To address the aforementioned technical problems, embodiments of this application provide a battery life prediction method, comprising the following steps: extracting the battery's capacity drop features based on the battery's capacity loss curve, and extracting the battery's capacity increment features based on the battery's IC curve; wherein, the capacity loss curve includes a component where the battery capacity loses linearly and a component where the battery capacity loses non-linearly, and the capacity drop features are extracted from the component where the battery capacity loses non-linearly; inputting the capacity increment features and the capacity drop features into a pre-trained battery life prediction model to obtain the battery life prediction result, wherein the battery life prediction model is obtained by training an LSTM model using at least one battery's capacity drop features and capacity increment features.
[0007] An embodiment of this application provides a battery life prediction device, comprising: a feature extraction module, configured to extract capacity drop features of the battery based on a capacity loss curve, and extract capacity increment features of the battery based on an IC curve of the battery; wherein the capacity loss curve includes a component of battery capacity loss with linear loss and a component of battery capacity loss with nonlinear loss, and the capacity drop features are extracted from the component of battery capacity loss with nonlinear loss; and a life prediction module, configured to input the capacity increment features and the capacity drop features into a pre-trained battery life prediction model to obtain a battery life prediction result, wherein the battery life prediction model is obtained by training an LSTM model using at least one battery capacity drop feature and capacity increment feature.
[0008] Embodiments of this application also provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described battery life prediction method.
[0009] Embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described battery life prediction method.
[0010] The battery life prediction method of this application addresses the issue that the battery capacity loss curve includes both linearly and non-linearly decreasing components. Since the linearly decreasing component has significant interference factors when predicting battery life, this application not only considers the impact of the IC curve on battery life and extracts the battery's capacity increment characteristics from the IC curve, but also extracts the battery's capacity drop characteristics from the non-linearly decreasing component of the capacity loss curve. Therefore, the battery life prediction model trained on an LSTM model based on at least one battery capacity drop characteristic and capacity increment characteristic has a better prediction effect than one that only considers a single capacity increment characteristic. This is mainly reflected in its high prediction accuracy and high prediction efficiency. By inputting the predicted battery capacity increment and capacity drop characteristics into the trained battery life prediction model, the battery life prediction result can be obtained. Furthermore, since the battery life prediction model in this application is obtained by training an LSTM model, it does not require the participation of a battery physical model, thus eliminating the need to identify complex physical parameters. Even under extreme operating conditions, it can provide accurate prediction results. In addition, this application can use only the data of one battery as the training set during the training process, and can achieve good battery capacity prediction results even under marginal conditions with a small amount of training data required.
[0011] In addition, the battery life prediction model is trained through the following steps: using the capacity increment feature of the at least one battery as the main feature and the capacity drop feature as the auxiliary feature, the LSTM model is trained to generate the battery life prediction model.
[0012] Furthermore, the capacity drop characteristic is obtained by normalizing the components where the battery capacity experiences a non-linear loss. The normalization formula is as follows:
[0013]
[0014] in, The component representing the non-linear loss of battery capacity is -R(t), which represents the percentage of the remaining number of cycles of the battery out of the total number of cycles. The value of -R(t) ranges from [0, 1].
[0015] In addition, the capacity increment feature includes the peak value of the IC curve and the voltage corresponding to the peak value.
[0016] In addition, before extracting the capacity increment features of the battery based on the IC curve of the battery, the method further includes: using the SG filtering method to filter the IC curve to smooth the IC curve and denoise the signal.
[0017] In addition, the LSTM model includes an attention mechanism layer, which is used to automatically adjust the peak value of the IC curve, the position of the peak value, and the weight of the capacity drop feature, so as to improve the model's ability to resist abnormal capacity changes and its transfer generalization ability.
[0018] In addition, the LSTM model includes a first LSTM neural network layer, a second LSTM neural network layer, a first dropout layer, a second dropout layer, an attention mechanism layer, and a fully connected layer; wherein the first LSTM neural network layer, the first dropout layer, the second LSTM neural network layer, and the second dropout layer are connected in series in sequence to reduce the probability of data overfitting during model training. Attached Figure Description
[0019] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.
[0020] Figure 1 This is a flowchart of a battery life prediction method provided according to an embodiment of this application. Figure 1 ;
[0021] Figure 2 This is a schematic diagram of a battery capacity curve and a capacity loss curve provided according to an embodiment of this application;
[0022] Figure 3 This is a KPP provided according to an embodiment of the present application. exp A schematic diagram of the (t) curve;
[0023] Figure 4 This is a schematic diagram of IC curves of a battery at different cycle numbers according to an embodiment of this application;
[0024] Figure 5 This is a schematic diagram of the IC curves of a battery before and after SG filtering, according to an embodiment of this application;
[0025] Figure 6 This is a schematic diagram of the structure of an LSTM model according to an embodiment of this application;
[0026] Figure 7 This is a schematic diagram illustrating the prediction effect of a battery life prediction model provided according to an embodiment of this application;
[0027] Figure 8 This is a schematic diagram of a battery life prediction device according to an embodiment of this application;
[0028] Figure 9 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation
[0029] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.
[0030] One embodiment of this application relates to a battery life prediction method, wherein the battery in this embodiment specifically refers to a lithium battery, and the specific process of the battery life prediction method in this embodiment can be as follows: Figure 1 As shown, it includes:
[0031] Step 101: Extract the capacity drop characteristics of the battery based on the battery capacity loss curve, and extract the capacity increment characteristics of the battery based on the battery IC curve; wherein, the capacity loss curve includes the component of battery capacity loss with linear loss and the component of battery capacity loss with nonlinear loss, and the capacity drop characteristics are extracted from the component of battery capacity loss with nonlinear loss.
[0032] Step 102: Input the capacity increment feature and capacity drop feature into the pre-trained battery life prediction model to obtain the battery life prediction result. The battery life prediction model is obtained by training the LSTM model with at least one battery capacity drop feature and capacity increment feature.
[0033] In this embodiment, since the battery capacity loss curve includes components with linear capacity loss and components with non-linear capacity loss, the component with linear capacity loss has a significant interference factor when predicting battery life. Therefore, this application not only considers the influence of the IC curve on battery life and extracts the battery capacity increment feature from the battery's IC curve, but also extracts the battery capacity drop feature from the component with non-linear capacity loss in the battery capacity loss curve. Therefore, the battery life prediction model trained on the LSTM model based on at least one battery capacity drop feature and capacity increment feature has a better prediction effect than considering only a single capacity increment feature. This is mainly reflected in the high prediction accuracy and high prediction efficiency. By inputting the predicted battery capacity increment feature and capacity drop feature into the trained battery life prediction model, the battery life prediction result can be obtained. Furthermore, since the battery life prediction model in this application is obtained by training an LSTM model, it does not require the participation of a battery physical model, and therefore does not need to identify complex physical parameters. Even if the battery is under extreme operating conditions, it can still provide accurate prediction results. In addition, this application can use only the data of one battery as the training set during the training process, and can achieve good battery capacity prediction results even under marginal conditions with a small amount of training data required.
[0034] The following is a detailed description of the implementation details of the battery life prediction method in this embodiment. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing this solution.
[0035] In step 101, various battery operating data that affect battery life are first collected. This battery is the one whose battery life needs to be predicted. The battery operating data includes the battery capacity loss curve and the battery IC curve. The capacity drop characteristics of the battery can be extracted from the capacity loss curve, and the capacity increment characteristics of the battery can be extracted from the IC curve.
[0036] First, we will introduce how to extract capacity drop features from the capacity loss curve:
[0037] See Figure 2 It shows the battery's capacity curve and capacity loss curve. The capacity curve describes the battery's capacity during each discharge cycle, and the capacity loss curve describes the capacity loss during each cycle. The capacity loss curve includes two parts: the component where the battery capacity loses linearly (…). Figure 2 The linear region in the battery and the components that exhibit non-linear capacity loss (in the battery) Figure 2In the nonlinear region of the battery capacity loss curve, the capacity loss of components with linear capacity loss remains at a relatively flat level, while the capacity loss of components with nonlinear capacity loss begins to rise rapidly, forming a "capacity plunge" phenomenon. The capacity plunge characteristic can be specifically extracted from the components with nonlinear capacity loss in the capacity loss curve.
[0038] In the specific implementation, the steps for extracting the capacity plunge feature are as follows:
[0039] (1) Obtain the battery's State of Health (SOH). SOH can be obtained from the battery's current cycle count and its initial capacity: Where SOH(t) represents the battery's state of health at cycle number t, and Q... t Q0 is the battery capacity at cycle number t, and Q0 is the initial battery capacity.
[0040] (2) Obtain the components in the battery capacity loss curve where the battery capacity exhibits non-linear loss. Components in the capacity loss curve where the battery capacity decreases linearly. It can be obtained by one-dimensional linear fitting, specifically by: for the linear fitting equation The error between the fitted result and the actual SOH is To make the fitting results closely resemble the true values, the following loss function should be minimized.
[0041]
[0042] The parameter sets m and b can be obtained, thus revealing the components where battery capacity exhibits a linear loss. The formula for calculating the component whose battery capacity exhibits a non-linear loss is:
[0043] (3) Extracting capacity drop characteristics. The point where the capacity loss begins to rise rapidly in the battery's capacity loss curve is called the "knee point". In this embodiment, the capacity drop characteristic can be called the knee point probability (KPP), and KPP is specifically obtained through KPP. exp The capacity plunge characteristic (i.e., KPP) is represented by a (t) curve. exp (t) curve) specifically describes the components where battery capacity loss is non-linear. The normalized calculation yields the following formula:
[0044]
[0045] Where -R(t) represents the percentage of remaining battery cycles out of the total number of cycles, and the value range of -R(t) is [0, 1]. Since the battery capacity loss curve ranges from [0.8, 1], in some embodiments, the value range of -R(t) is consistent with the value range of the capacity loss curve, which is [0.8, 1]. In geometric theory, KPP... exp (t) reaches its maximum value at the inflection point, therefore KPP can be increased after the inflection point. exp To obtain KPP, the value of (t) is set to 1. exp (t) curve, KPP exp (t) curve as Figure 3 As shown.
[0046] Since the components in the capacity loss curve that exhibit non-linear capacity loss have a significant impact on battery life prediction, this embodiment extracts capacity drop features from the components exhibiting non-linear capacity loss for battery life prediction.
[0047] Secondly, the extraction of capacity increment features from IC curves is introduced as follows:
[0048] The IC curve of a battery represents the rate of change of capacity with voltage during constant current charging / charging / discharging. It analyzes the aging principle of the battery at the electrode level and has a good descriptive ability for the phase transition characteristics during the insertion and stripping of lithium-ion active materials. Therefore, this embodiment extracts capacity increment characteristics from the battery's IC curve to describe the degradation process of the battery's internal materials. Taking the battery in a low-temperature operating condition at the edge of its operating range as an example, since the constant current process of fast charging is very short and it quickly enters constant voltage charging, the IC curve can be obtained using the constant current range during the discharging process. The calculation formula is as follows:
[0049]
[0050] Where Q represents the battery capacity at time t, V represents the voltage, and t represents the sampling time.
[0051] In a specific implementation, the capacity increment characteristic includes the peak value of the battery's IC curve and the voltage corresponding to that peak value. Specifically, see [link to relevant documentation]. Figure 4The diagram illustrates the IC curves of a battery at different cycle counts, including 1, 12, 31, 41, and 51 cycles. It shows that the IC curves exhibit similar shapes across different cycle counts, gradually decreasing in height and showing a clear trend as the cycle count increases. The peaks in the IC curves possess unique heights and positions, reflecting the changes in electrode materials during charging and discharging. Furthermore, the changes in the peak value and position of the IC curve are relatively linear in the early stages, but as the battery capacity begins to decline, the peak value of the IC curve drops rapidly, shifting towards lower voltage levels. This verifies that the IC curve has the ability to identify the inflection point of sudden battery degradation. Therefore, in this embodiment, the peak value of the IC curve during battery discharge and the corresponding voltage coordinate are used as capacity increment features for battery life prediction.
[0052] In some embodiments, since the parameter values of the IC curve are easily disturbed by battery-to-cycle noise in the charging platform region, it is difficult to collect the peak value of the IC curve, making it difficult to extract subsequent capacity enhancement features. Therefore, in this embodiment, after obtaining the IC curve, before extracting the battery's capacity increment features based on the battery's IC curve, it is first filtered. Specifically, the SG filtering method is used to filter the IC curve. Figure 5 The IC curves before and after SG filtering are shown. It can be seen that the IC curve after filtering is smooth and can be used to extract capacity increment features.
[0053] SG filtering is a digital signal processing method used for smoothing and denoising signals. The formula for SG filtering is as follows:
[0054]
[0055] Among them, y i It is the IC signal value after SG filtering, x i+j It is the (i+j)th sampling point of the original IC signal, c j These are the coefficients used in the SG filter, and n is the order of the SG filter.
[0056] As can be seen, this embodiment not only considers the IC curve reflecting the battery aging characteristics and extracts the battery capacity increment characteristics from it to determine the degree of battery aging and thus predict the battery life, but also considers the battery capacity loss curve that interferes with the prediction of battery life and collects multiple factors that affect battery life.
[0057] In step 102, the collected battery capacity drop features and capacity increment features are used as input to the pre-trained battery life prediction model. Since the battery life prediction model is obtained by training the LSTM model based on the capacity drop features and capacity increment features of at least one battery, the output of the battery life prediction model is the battery life prediction result.
[0058] In the specific implementation, the battery life prediction model is trained through the following steps: Using the capacity increment feature of at least one battery as the primary feature and the capacity drop feature as the auxiliary feature, an LSTM model is trained to generate the battery life prediction model. The IC curve of the battery analyzes the battery aging mechanism from the electrode level, and can be combined with the external signal rheology of the battery from the mechanism level. The capacity increment feature extracted from it can be used as a separate feature to complete model training; therefore, in this embodiment, it is used as the primary feature in the battery life prediction model training process. The capacity drop feature does not describe the internal aging mechanism of the battery, so it can be used as an auxiliary feature in the battery life prediction model training process. The two features are fused as training samples, resulting in a battery life prediction model with better prediction performance. Furthermore, this embodiment uses LSTM as the basic network architecture. LSTM, by introducing input gates, forget gates, and output gates, can effectively control the flow of information and retain important historical information, thereby improving the model's performance and stability.
[0059] In some embodiments, the LSTM model includes an attention mechanism layer, which is used to automatically adjust the peak value of the IC curve, the location of the peak value, and the weights of the capacity drop feature.
[0060] Attention mechanism (AM) is a machine learning technique that helps neural networks focus more on important features, thereby improving model performance and generalization ability. In neural networks, each feature has a different impact on the result, but typically only one set of features plays a dominant role in the output (i.e., capacity increment features), while the remaining features play a supporting role (i.e., capacity drop features). The attention mechanism learns based on the attention level of individual features in the sequence and integrates features according to that attention level. Even if there is only one feature input to the LSTM model, when processing time-series data, the attention mechanism can still capture key information in the time-series data better by weighting different time steps. The expression of the attention mechanism is as follows:
[0061] u (t) =tanh(W (uH) *h (t) +b (u) )
[0062]
[0063] s (t) =a (t) h (t)
[0064] Where W represents the weight, b represents the bias term, a represents the weight of each attribute, and s represents the prediction result after weighted summation.
[0065] Furthermore, since the IC curve amplifies the noise in the measurement during processing, and filtering is also unable to eliminate the impact of noise on the training results, abnormal capacity mutations may occur during model prediction. Therefore, the attention mechanism introduced in this embodiment can improve the model's ability to resist abnormal capacity mutations and its transfer and generalization capabilities by automatically adjusting the value and position of the IC curve peak in the capacity increment feature, as well as the weight of the subsequently added capacity drop feature, and optimizing the dependency relationship between each feature and the time step in a timely manner.
[0066] In some embodiments, the LSTM model includes a first LSTM neural network layer, a second LSTM neural network layer, a first dropout layer, a second dropout layer, an attention mechanism layer, and a fully connected layer; wherein the first LSTM neural network layer, the first dropout layer, the second LSTM neural network layer, and the second dropout layer are connected in series in sequence. Figure 6 As shown, both the first and second LSTM neural network layers have 64 neurons, and the dropout rate of the first and second dropout layers is 0.4. In this embodiment, by concatenating the first LSTM neural network layer and the first dropout layer, and concatenating the second LSTM neural network layer and the second dropout layer, the probability of data overfitting when training the battery life prediction model can be reduced.
[0067] To facilitate understanding, the battery life prediction methods of the above embodiments are now illustrated through the following experiments:
[0068] Taking a low-temperature operating condition (-20℃) as an example, battery data for six batteries at -20℃ was first obtained through bench testing, including capacity loss curves and IC curves for each battery. Based on these curves, the capacity drop characteristics and capacity increase characteristics of the six batteries were extracted. The six batteries are battery 1, battery 2, battery 3, battery 4, battery 5, and battery 6. Batteries 1 and 2 were charged at 15A, batteries 3 and 4 at 10A, and batteries 5 and 6 at 5A.
[0069] A battery life prediction model was trained using battery 1. Specifically, the corresponding battery life prediction models were trained in the following ways: (1) The LSTM model was trained using the capacity increment features of battery 1 to generate the corresponding battery life prediction model. At this time, the structure of the LSTM model did not include the attention mechanism layer; (2) The LSTM model was trained using the capacity drop features and capacity increment features of battery 1 to generate the corresponding battery life prediction model. At this time, the structure of the LSTM model did not include the attention mechanism layer; (3) The LSTM model was trained using the capacity increment features of battery 1 to generate the corresponding battery life prediction model. At this time, the structure of the LSTM model included the attention mechanism layer; (4) The LSTM model was trained using the capacity drop features and capacity increment features of battery 1 to generate the corresponding battery life prediction model. At this time, the structure of the LSTM model included the attention mechanism layer. Finally, the battery life prediction models trained in the above four ways were verified using the working data corresponding to batteries 2 to 6: capacity loss curves and IC curves (i.e., capacity drop features and capacity increment features).
[0070] Figure 7 a) to Figure 7 f) shows the validation results of the above battery life prediction models for batteries 2 to 6 respectively. Table 1 shows the life prediction results of batteries 2 to 6 using the above battery life prediction models.
[0071] Table 1
[0072]
[0073] Among them, ①LSTM+IC represents a battery life prediction model trained on an LSTM model using the capacity increment features of battery 1, and the structure of the LSTM model in this case does not include an attention mechanism layer; ②LSTM+IC+KPP represents a battery life prediction model trained on an LSTM model using the capacity drop and capacity increment features of battery 1, and the structure of the LSTM model in this case also does not include an attention mechanism layer; ③LSTM+IC+AM represents a battery life prediction model trained on an LSTM model using the capacity increment features of battery 1, and the structure of the LSTM model in this case includes an attention mechanism layer (i.e., AM); ④LSTM+IC+KPP+AM represents a battery life prediction model trained on an LSTM model using the capacity drop and capacity increment features of battery 1, and the structure of the LSTM model in this case includes an attention mechanism layer (i.e., AM).
[0074] Combination Figure 7 a) to Figure 7f) and Table 1 show that the battery life prediction model trained on the LSTM model without the attention mechanism layer using the capacity drop and capacity increment features of battery 1 (i.e., ②LSTM+IC+KPP) and the battery life prediction model trained on the LSTM model with the attention mechanism layer using the capacity increment features of battery 1 (i.e., ③LSTM+IC+AM) have better prediction performance than the battery life prediction model trained on the LSTM model without the attention mechanism layer using only the capacity increment features of battery 1 (i.e., ①LSTM+IC). The battery life prediction model trained on the LSTM model with the attention mechanism layer by combining the capacity drop and capacity increment features of battery 1 has the best performance.
[0075] It should be noted that the examples described above in this embodiment are merely illustrative for ease of understanding and do not constitute a limitation on the technical solution of this application.
[0076] The steps of the various methods described above are only for clarity. In practice, they can be combined into one step or some steps can be split into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, but without changing the core design of the algorithm and process, are also within the scope of protection of this patent.
[0077] Another embodiment of this application relates to a battery life prediction device. The implementation details of this embodiment's battery life prediction device are described below. The following implementation details are provided for ease of understanding and are not essential for implementing this solution. A schematic diagram of this embodiment's battery life prediction device can be seen as follows: Figure 8 As shown, it includes: a feature extraction module 801 and a lifetime prediction module 802.
[0078] Specifically, the feature extraction module 801 is used to extract the capacity drop feature of the battery based on the battery capacity loss curve, and to extract the capacity increment feature of the battery based on the battery IC curve; wherein, the capacity loss curve includes the component of battery capacity loss with linear loss and the component of battery capacity loss with nonlinear loss, and the capacity drop feature is extracted from the component of battery capacity loss with nonlinear loss.
[0079] The life prediction module 802 is used to input the capacity increment feature and capacity drop feature into the pre-trained battery life prediction model to obtain the battery life prediction result. The battery life prediction model is obtained by training the LSTM model with the capacity drop feature and capacity increment feature of at least one battery.
[0080] In this embodiment, since the battery capacity loss curve includes components with linear capacity loss and components with non-linear capacity loss, the component with linear capacity loss has a significant interference factor when predicting battery life. Therefore, this application not only considers the influence of the IC curve on battery life and extracts the battery capacity increment feature from the battery's IC curve, but also extracts the battery capacity drop feature from the component with non-linear capacity loss in the battery capacity loss curve. Therefore, the battery life prediction model trained on the LSTM model based on at least one battery capacity drop feature and capacity increment feature has a better prediction effect than considering only a single capacity increment feature. This is mainly reflected in the high prediction accuracy and high prediction efficiency. By inputting the predicted battery capacity increment feature and capacity drop feature into the trained battery life prediction model, the battery life prediction result can be obtained. Furthermore, since the battery life prediction model in this application is obtained by training an LSTM model, it does not require the participation of a battery physical model, and therefore does not need to identify complex physical parameters. Even if the battery is under extreme operating conditions, it can still provide accurate prediction results. In addition, this application can use only the data of one battery as the training set during the training process, and can achieve good battery capacity prediction results even under marginal conditions with a small amount of training data required.
[0081] It is not difficult to see that this embodiment is a device embodiment corresponding to the above method embodiments, and this embodiment can be implemented in conjunction with the above method embodiments. The relevant technical details and technical effects mentioned in the above embodiments are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above embodiments.
[0082] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.
[0083] Another embodiment of this application relates to an electronic device, such as... Figure 9 As shown, it includes: at least one processor 901; and a memory 902 communicatively connected to the at least one processor 901; wherein the memory 902 stores instructions executable by the at least one processor 901, the instructions being executed by the at least one processor 901 to enable the at least one processor 901 to perform the battery life prediction method in the above embodiments.
[0084] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.
[0085] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.
[0086] Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.
[0087] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0088] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.
Claims
1. A method for predicting battery life, characterized in that, include: The capacity drop characteristics of the battery are extracted based on the battery's capacity loss curve, and the capacity increment characteristics of the battery are extracted based on the battery's IC curve; wherein, the capacity loss curve includes components with linear capacity loss and components with nonlinear capacity loss, and the capacity drop characteristics are extracted from the components with nonlinear capacity loss. The capacity increment feature and the capacity drop feature are input into the pre-trained battery life prediction model to obtain the battery life prediction result. The battery life prediction model is obtained by training the LSTM model with at least one battery capacity drop feature and capacity increment feature. The extraction steps for the capacity drop feature are as follows: Obtain the health status of the battery In the formula, The health status of the battery at cycle number t. The capacity of the battery at cycle number t. This refers to the initial capacity of the battery; The components in the capacity loss curve that exhibit linear capacity loss are obtained through one-dimensional linear fitting. To obtain the components in which the battery capacity exhibits a non-linear loss. = ; According to the components where the battery capacity exhibits a non-linear loss Through the normalization formula Extract the capacity drop characteristics of the battery; where, This indicates the percentage of remaining battery cycles out of the total number of cycles, and The value range is [0, 1].
2. The battery life prediction method according to claim 1, characterized in that, The battery life prediction model is trained through the following steps: The LSTM model is trained using the capacity increment feature of at least one battery as the main feature and the capacity drop feature as the auxiliary feature to generate the battery life prediction model.
3. The battery life prediction method according to claim 1 or 2, characterized in that, The capacity increment feature includes the peak value of the IC curve and the voltage corresponding to the peak value.
4. The battery life prediction method according to claim 3, characterized in that, Before extracting the capacity increment characteristics of the battery based on the IC curve of the battery, the method further includes: The IC curve is filtered using the SG filtering method.
5. The battery life prediction method according to claim 3, characterized in that, The LSTM model includes an attention mechanism layer, which is used to automatically adjust the peak value of the IC curve, the position of the peak value, and the weight of the capacity drop feature.
6. The battery life prediction method according to claim 5, characterized in that, The LSTM model includes a first LSTM neural network layer, a second LSTM neural network layer, a first dropout layer, a second dropout layer, an attention mechanism layer, and a fully connected layer; The first LSTM neural network layer, the first dropout layer, the second LSTM neural network layer, and the second dropout layer are connected in series in sequence.
7. A battery life prediction device, characterized in that, include: The feature extraction module is used to extract the capacity drop feature of the battery based on the battery's capacity loss curve, and to extract the capacity increment feature of the battery based on the battery's IC curve; wherein, the capacity loss curve includes a component in which the battery capacity loses linearly and a component in which the battery capacity loses non-linearly, and the capacity drop feature is extracted from the component in which the battery capacity loses non-linearly. The life prediction module is used to input the capacity increment feature and the capacity drop feature into a pre-trained battery life prediction model to obtain the battery life prediction result. The battery life prediction model is obtained by training an LSTM model with at least one battery capacity drop feature and capacity increment feature. The feature extraction module extracts the capacity plunge feature in the following way: Obtain the health status of the battery In the formula, The health status of the battery at cycle number t. The capacity of the battery at cycle number t. This refers to the initial capacity of the battery; The components in the capacity loss curve that exhibit linear capacity loss are obtained through one-dimensional linear fitting. To obtain the components in which the battery capacity exhibits a non-linear loss. = ; According to the components where the battery capacity exhibits a non-linear loss Through the normalization formula Extract the capacity drop characteristics of the battery; where, This indicates the percentage of remaining battery cycles out of the total number of cycles, and The value range is [0, 1].
8. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the battery life prediction method as described in any one of claims 1 to 6.
9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the battery life prediction method according to any one of claims 1 to 6.