Battery capacity turning point and residual life joint prediction method and system
By utilizing relaxation voltage data and a joint prediction model, the problem of insufficient prediction of nonlinear degradation characteristics of lithium-ion batteries was solved, enabling accurate prediction of battery capacity inflection points and remaining lifespan, thereby improving the accuracy and safety of the battery management system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TONGJI UNIV
- Filing Date
- 2024-02-29
- Publication Date
- 2026-05-29
Smart Images
Figure CN117872164B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of energy storage battery technology, and in particular to a method for jointly predicting battery capacity inflection point and remaining life. Background Technology
[0002] Lithium-ion batteries undergo irreversible aging and fatigue due to their time-varying and nonlinear characteristics. In particular, the nonlinear degradation that lithium batteries exhibit after long-term cycling manifests as accelerated capacity decay, posing a risk of failure to existing battery management systems in terms of lifespan estimation and prediction. This accelerated performance decline not only shortens battery lifespan but also compromises safety, becoming a technological bottleneck hindering their reliable application. The inflection point, as the starting point of nonlinear degradation, contains crucial temporal information, which is essential for better understanding and further addressing nonlinear battery degradation.
[0003] Current research and technologies place greater emphasis on predicting battery remaining lifespan, while research and technologies focusing on predicting inflection points are relatively few. For example, patent CN112949059B discloses a method for estimating the health status and predicting remaining lifespan of lithium batteries under time-varying discharge current, and patent application CN115236518A discloses a method for predicting the cycle life of lithium batteries, both used to predict battery remaining lifespan. Patent application CN115994441A discloses an online battery lifespan prediction method based on a big data cloud platform using mechanistic information. This method extracts mechanistic data from battery discharge cycle data and predicts the battery's remaining lifespan and capacity degradation inflection point through a regression model. Achieving accurate remaining lifespan estimation helps to make battery performance predictable. If the nonlinear degradation trend of the battery is clearer, more proactive measures can be taken to extend the battery's lifespan. Furthermore, if the inflection point occurs before the end of the lifespan, predicting the inflection point is also helpful in determining the ideal battery retirement time. Summary of the Invention
[0004] The purpose of this invention is to provide a method and system for jointly predicting battery capacity inflection point and remaining lifespan, enabling accurate prediction of battery capacity inflection point and remaining lifespan.
[0005] The objective of this invention can be achieved through the following technical solutions:
[0006] A method for jointly predicting battery capacity inflection point and remaining life includes the following steps:
[0007] Obtain the relaxation voltage data of the battery to be predicted after it is fully charged;
[0008] The relaxation voltage data is input into a pre-trained joint prediction model, which outputs the predicted battery capacity inflection point and remaining lifespan. The joint prediction model includes a capacity inflection point prediction sub-model and a remaining lifespan prediction sub-model.
[0009] The steps for predicting the battery capacity inflection point and remaining lifespan include:
[0010] The relaxation voltage data is input into the capacity inflection point prediction sub-model, and the predicted number of remaining battery inflection cycles is output to determine the battery capacity inflection point.
[0011] The predicted remaining battery cycle count and relaxation voltage data are combined and input into the remaining lifetime prediction sub-model to output the predicted remaining lifetime.
[0012] Furthermore, the steps for training the joint prediction model include:
[0013] Acquire the capacity decay curves of multiple batteries, calculate the remaining transition cycle number and remaining lifespan of each battery cycle, and collect the relaxation voltage data after each full charge in each battery cycle.
[0014] The relaxation voltage data after each full charge in each cycle of the battery is used as training sample one, and the corresponding remaining number of transition cycles is used as a label to train the capacity inflection point prediction sub-model.
[0015] The remaining transition cycle number of each cycle of the battery and the relaxation voltage data after each full charge in each cycle of the battery are used as training sample two, and the corresponding remaining lifetime is used as the label to train the remaining lifetime prediction sub-model.
[0016] The trained capacity inflection point prediction sub-model and remaining lifetime prediction sub-model are integrated into a joint prediction model to complete the training process of the joint prediction model.
[0017] Furthermore, the calculation expressions for the remaining transition cycle number and remaining lifespan of the battery in each cycle are as follows:
[0018] CTK i =C k -C i
[0019] RUL i =C e -C i
[0020] Among them, CTK i RUL i C represents the remaining transition cycle number and remaining lifespan of the battery in the i-th cycle, respectively. e C k and C i These represent the number of iterations at the end of the lifespan, the capacity inflection point, and the current i-th iteration, respectively.
[0021] Furthermore, the number of cycles C at the capacity inflection point k Identified by the Bacon-Watts model, the expression of which is:
[0022] Y = α0 + α1(xC) k )+α2(xC k )tanh((xC k ) / γ)+Z
[0023] In the formula, Y is the capacity decay curve from charge-discharge cycles to the end of the lifespan, Z is a normally distributed random variable centered at zero, representing the residual, and α0 is the value of x = C. k The intercept at point α1 and α2 adjust the slopes of the two intersecting lines, and γ is a constant that controls the abruptness of the transition.
[0024] Furthermore, the loss function of the joint prediction model is a hybrid loss function composed of the capacity inflection point prediction sub-model and the remaining lifetime prediction sub-model, and the expression of the hybrid loss function is:
[0025] L CTK-RUL =(1-λ)L CTK +L RUL
[0026] In the formula, L CTK-RUL =(1-λ)L CTK +L RUL The loss function is a mixture, where λ is the loss weight for capacity inflection point prediction, and L... CTK To predict losses at capacity inflection points, L RUL Predict losses for remaining lifetime.
[0027] This invention also provides a joint prediction system for battery capacity inflection point and remaining lifetime based on relaxation voltage, comprising:
[0028] Data acquisition module: used to acquire the relaxation voltage data of the battery to be predicted after it is fully charged;
[0029] Capacity inflection point prediction module: used to input the relaxation voltage data into the capacity inflection point prediction sub-model in the pre-trained joint prediction model, output the predicted number of remaining battery inflection cycles, and determine the battery capacity inflection point;
[0030] Remaining life prediction module: This module combines the predicted remaining battery cycle count and relaxation voltage data, inputs them into the remaining life prediction sub-model, and outputs the predicted remaining life.
[0031] Furthermore, the steps for training the joint prediction model include:
[0032] Acquire the capacity decay curves of multiple batteries, calculate the remaining transition cycle number and remaining lifespan of each battery cycle, and collect the relaxation voltage data after each full charge in each battery cycle.
[0033] The relaxation voltage data after each full charge in each cycle of the battery is used as training sample one, and the corresponding remaining number of transition cycles is used as a label to train the capacity inflection point prediction sub-model.
[0034] The remaining transition cycle number of each cycle of the battery and the relaxation voltage data after each full charge in each cycle of the battery are used as training sample two, and the corresponding remaining lifetime is used as the label to train the remaining lifetime prediction sub-model.
[0035] The trained capacity inflection point prediction sub-model and remaining lifetime prediction sub-model are integrated into a joint prediction model to complete the training process of the joint prediction model.
[0036] Furthermore, the calculation expressions for the remaining transition cycle number and remaining lifespan of the battery in each cycle are as follows:
[0037] CTK i =C k -C i
[0038] RUL i =C e -C i
[0039] Among them, CTK i RUL i C represents the remaining transition cycle number and remaining lifespan of the battery in the i-th cycle, respectively. e C k and C i These represent the number of iterations at the end of the lifespan, the capacity inflection point, and the current i-th iteration, respectively.
[0040] Furthermore, the number of cycles C at the capacity inflection point k Identified by the Bacon-Watts model, the expression of which is:
[0041] Y = α0 + α1(xC) k )+α2(xC k )tanh((xC k ) / γ)+Z
[0042] In the formula, Y is the capacity decay curve from charge-discharge cycles to the end of the lifespan, Z is a normally distributed random variable centered at zero, representing the residual, and α0 is the value of x = C. kThe intercept at point α1 and α2 adjust the slopes of the two intersecting lines, and γ is a constant that controls the abruptness of the transition.
[0043] Furthermore, the loss function of the joint prediction model is a hybrid loss function composed of the capacity inflection point prediction sub-model and the remaining lifetime prediction sub-model, and the expression of the hybrid loss function is:
[0044] L CTK-RUL =(1-λ)L CTK +L RUL
[0045] In the formula, L CTK-RUL =(1-λ)L CTK +L RUL The loss function is a mixture, where λ is the loss weight for capacity inflection point prediction, and L... CTK To predict losses at capacity inflection points, L RUL Predict losses for remaining lifetime.
[0046] Compared with the prior art, the present invention has the following beneficial effects:
[0047] (1) This invention takes into account the potential relationship between capacity inflection point and life end point, predicts the remaining number of battery inflection cycles based on relaxation voltage data and capacity inflection point prediction sub-model, and uses the remaining number of battery inflection cycles and relaxation voltage data as input to the remaining life prediction sub-model, making full use of the correlation between capacity inflection point and life to achieve accurate prediction of capacity inflection point and remaining life.
[0048] (2) The present invention uses relaxation voltage as sample data, which is simple to obtain and is not affected by the changing charging and discharging conditions. It is convenient to collect a large amount of data to support the training of the model and can realize frequent and rapid prediction of capacity inflection point and remaining lifetime.
[0049] (3) This invention helps to better monitor the degradation trend of batteries. Attached Figure Description
[0050] Figure 1 This is a schematic diagram of the method flow of the present invention;
[0051] Figure 2 This is a graph showing the capacity decay curve and the correspondence between capacity inflection point and lifespan end point of the battery dataset in this embodiment of the invention.
[0052] Figure 3 This is a graph showing the change in relaxation voltage of a single cell as the battery decays, according to an embodiment of the present invention. Detailed Implementation
[0053] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.
[0054] Example 1
[0055] This embodiment provides a method for jointly predicting battery capacity inflection point and remaining lifespan, such as... Figure 1 As shown, the method includes the following steps:
[0056] S1. Obtain the relaxation voltage data of the battery to be predicted after it is fully charged.
[0057] S2. Input the relaxation voltage data into the pre-trained joint prediction model and output the predicted battery capacity inflection point and remaining life.
[0058] The joint prediction model includes a capacity inflection point prediction sub-model and a remaining lifetime prediction sub-model. The steps for training the joint prediction model include:
[0059] (1) Obtain the capacity decay curves of multiple batteries, calculate the remaining transition cycle number and remaining life of each battery cycle, and collect the relaxation voltage data after each full charge in each battery cycle.
[0060] Remaining number of transition cycles CTK i and remaining life RUL i The calculation method is as follows:
[0061] CTK i =C k -C i
[0062] RUL i =C e -C i
[0063] Where C e C k and C i These represent the number of cycles at the end of the battery's lifespan, the capacity inflection point, and the current number of cycles, respectively. Clearly, over the entire lifespan of a given battery, CTK... i and RUL i It is a straight line with a slope of -1. For CTK i :
[0064] 1)C i During the linear phase before the capacity inflection point, CTK i It is a positive integer;
[0065] 2)C iAt the capacity inflection point, CTK i =0;
[0066] 3)C i When in the nonlinear phase following the capacity transition, CTK i It is a negative integer.
[0067] For RUL i :
[0068] 1) In C i Before reaching the end of its lifespan, RUL i It is a positive integer;
[0069] 2)C i When RUL reaches the end of its lifespan i It is 0.
[0070] Among them, the capacity inflection point C k Identified by the Bacon-Watts model:
[0071] Y = α0 + α1(xC) k )+α2(xC k )tanh((xC k ) / γ)+Z
[0072] Where Y is the capacity decay curve from charge-discharge cycles to the end of the lifetime, Z is a normally distributed random variable centered at zero, representing the residual; α0 is x = C k The intercept at point α1 and α2 adjust the slopes of the two intersecting lines; γ is a constant that controls the abruptness of the transition.
[0073] (2) The relaxation voltage data after each full charge in each cycle of the battery is used as training sample one, and the corresponding remaining number of transition cycles is used as a label to train the capacity transition point prediction sub-model.
[0074] The relaxation voltage data X for this step i for:
[0075] X i =[V i,1 V i,2 ,…,V i,m ]
[0076] Where V i,m This represents the voltage value at the m-th sampling point collected during the resting period after the battery is fully charged in the i-th cycle.
[0077] Sample data X of relaxation voltage for each battery in each cycle i and the corresponding remaining number of transition cycles CTK iThe labels form the training set, which is then input into the machine learning model to train the inflection point prediction sub-model.
[0078] (3) The combination of the remaining transition cycle number of each cycle of the battery and the relaxation voltage data after each full charge in each cycle of the battery is used as training sample two, and the corresponding remaining lifetime is used as a label to train the remaining lifetime prediction sub-model.
[0079] The remaining number of transition loops in this step CTK i and relaxation voltage X i The method for combining sample data is as follows:
[0080] U i =[CTK i ,X i ]
[0081] Combined sample data X of each battery cycle i and the corresponding remaining useful life (RUL) i The labels form the training set, which is then input into the machine learning model to train the inflection point prediction sub-model.
[0082] (4) Integrate the trained capacity inflection point prediction sub-model and remaining lifetime prediction sub-model into a joint prediction model to complete the training process of the joint prediction model.
[0083] Set a hybrid loss function L for the capacity inflection point prediction sub-model and the remaining lifetime prediction sub-model. CTK-RUL :
[0084] L CTK-RUL =(1-λ)L CTK +L RUL
[0085] Where λ is the loss weight for capacity inflection point prediction, and L is the loss for capacity inflection point prediction. CTK and remaining lifetime predicted loss L RUL Based on mean squared error, the prediction accuracy of the two outputs can be adjusted by changing the weights of the mixed loss, thus achieving model ensemble.
[0086] After the above training steps are completed, a trained joint prediction model is obtained. The specific steps for outputting the predicted battery capacity inflection point and remaining life in step S2 are as follows:
[0087] (1) Input the relaxation voltage data of the battery to be predicted after it is fully charged into the joint prediction model.
[0088] (2) The relaxation voltage sample data is first input into the capacity inflection point prediction sub-model, and the current remaining number of inflection cycles is output to determine the inflection point position.
[0089] (3) The predicted remaining transition cycle number and relaxation voltage data are combined into sample data and input into the remaining lifetime prediction sub-model to output the current remaining lifetime and determine the end position of the battery life.
[0090] To verify the effectiveness of the above method, this embodiment uses a ternary lithium-ion battery as the application background. The charge and discharge cutoff voltages of the ternary lithium-ion battery are 4.2V and 2.65V, respectively, and the nominal capacity is 3500mAh. In actual applications, it is not limited to this.
[0091] 1. Obtain the capacity decay curve of the battery, calculate the remaining transition cycle number and remaining life of each battery in each cycle, and collect the relaxation voltage data of each battery after each full charge.
[0092] In this embodiment, multiple battery cells are subjected to charge-discharge cycle tests until the battery life decays to 80% of its initial capacity, i.e., the end of its lifespan. After each full charge, the cells are left to rest for 10 minutes, and the relaxation voltage is recorded, with a sampling interval of 120 seconds. The discharge capacity of each battery in each cycle is recorded, such as... Figure 2 As shown, the capacity decay of each battery is non-linear, accelerating from a certain inflection point. It is important to note that the capacity inflection point calculated using the Bacon-Watts model is plotted against the number of cycles at the end of the battery's lifespan. Figure 2 In the upper right corner, a strong linear relationship can be seen between the capacity inflection point and the end of the lifetime.
[0093] 2. Use the relaxation voltage of each cycle of the training set battery as sample data, and use the corresponding remaining number of transition cycles as labels to train the capacity transition point prediction model.
[0094] The capacity decay data of 6 batteries were randomly selected from the battery dataset obtained from the experiment as the test set, and the data of the remaining batteries were used to form the training set to train the joint prediction model. Figure 3 This diagram illustrates the change in relaxation voltage of a single battery cell as it decays in this embodiment of the invention. It shows that the relaxation voltage gradually decreases with battery degradation. Based on this characteristic, the 10-minute relaxation voltage sequence of the training set batteries is used as sample data, and the corresponding remaining transition loop count is used as a label. This data is then input into a machine learning network with adaptive optimization and convergence capabilities. The machine learning network used in this embodiment is a Gate Recurrent Unit (GRU) network, but it is not limited to this in practical applications. The Adam optimizer is used to optimize the parameters during training, with an initial learning rate set to 10. -3 .
[0095] 3. The remaining transition cycle number and relaxation voltage of each cycle of the training set battery are combined as sample data, and the corresponding remaining lifetime is used as a label to train the remaining lifetime prediction model, which is then integrated with the capacity transition point prediction model into a joint prediction model.
[0096] In this embodiment, a GRU network is also used to train the remaining lifetime prediction model. When integrating the two prediction models, the prediction accuracy of the two outputs is adjusted by changing the weight of the mixed loss. Here, the loss weight λ for the capacity inflection point prediction is set to 0.5.
[0097] 4. Input the relaxation voltage data of the test cells into the joint prediction model to predict the capacity inflection point and remaining life of the cells.
[0098] In this embodiment, the mean absolute error (MAE) and root mean square error (RMSE) are used as evaluation metrics to assess the prediction accuracy of the test set. Ultimately, the overall MAE and RMSE for the capacity inflection point prediction of the six batteries were 20.2 cycles and 24.9 cycles, respectively, while the overall MAE and RMSE for the remaining lifetime prediction were 14.9 cycles and 19.4 cycles, respectively. These results demonstrate the feasibility and accuracy of this joint prediction method.
[0099] Example 2
[0100] This embodiment provides a joint prediction system for battery capacity inflection point and remaining life, the system comprising:
[0101] Data acquisition module: used to acquire the relaxation voltage data of the battery to be predicted after it is fully charged;
[0102] Capacity inflection point prediction module: used to input the relaxation voltage data into the capacity inflection point prediction sub-model in the pre-trained joint prediction model, output the predicted number of remaining battery inflection cycles, and determine the battery capacity inflection point;
[0103] Remaining life prediction module: This module combines the predicted remaining battery cycle count and relaxation voltage data, inputs them into the remaining life prediction sub-model, and outputs the predicted remaining life.
[0104] The rest are as in Example 1.
[0105] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. 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.
[0106] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.
[0107] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0108] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0110] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.
[0111] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A method for jointly predicting battery capacity inflection point and remaining life, characterized in that, Includes the following steps: Obtain the relaxation voltage data of the battery to be predicted after it is fully charged; The relaxation voltage data is input into a pre-trained joint prediction model, which outputs the predicted battery capacity inflection point and remaining lifespan. The joint prediction model includes a capacity inflection point prediction sub-model and a remaining lifespan prediction sub-model. The steps for predicting the battery capacity inflection point and remaining lifespan include: The relaxation voltage data is input into the capacity inflection point prediction sub-model, and the predicted number of remaining battery inflection cycles is output to determine the battery capacity inflection point. The predicted remaining battery cycle count and relaxation voltage data are combined and input into the remaining lifetime prediction sub-model to output the predicted remaining lifetime.
2. The method for jointly predicting battery capacity inflection point and remaining lifespan according to claim 1, characterized in that, The steps for training the joint prediction model include: Acquire the capacity decay curves of multiple batteries, calculate the remaining transition cycle number and remaining lifespan of each battery cycle, and collect the relaxation voltage data after each full charge in each battery cycle. The relaxation voltage data after each full charge in each cycle of the battery is used as training sample one, and the corresponding remaining number of transition cycles is used as a label to train the capacity inflection point prediction sub-model. The remaining transition cycle number of each cycle of the battery and the relaxation voltage data after each full charge in each cycle of the battery are used as training sample two, and the corresponding remaining lifetime is used as the label to train the remaining lifetime prediction sub-model. The trained capacity inflection point prediction sub-model and remaining lifetime prediction sub-model are integrated into a joint prediction model to complete the training process of the joint prediction model.
3. The method for jointly predicting battery capacity inflection point and remaining lifespan according to claim 2, characterized in that, The formulas for calculating the remaining transition cycle number and remaining lifespan of the battery for each cycle are as follows: in, CTK i , RUL i The battery number i The remaining number of transition loops and the remaining lifetime of the next cycle. C e , C k and C i These represent the cycle count at the end of life, the capacity inflection point, and the current cycle. i Number of iterations.
4. The method for jointly predicting battery capacity inflection point and remaining lifespan according to claim 3, characterized in that, The number of cycles at the capacity inflection point C k Identified by the Bacon-Watts model, the expression of which is: In the formula, Y This is the capacity decay curve from charge-discharge cycles to the end of the lifetime. Z It is a normally distributed random variable centered at zero, representing the residuals. α 0 is x = C k intercept at, α 1 and α 2. Adjust the slope of the two intersecting lines. γ A constant used to control the abruptness of transitions.
5. The method for jointly predicting battery capacity inflection point and remaining lifespan according to claim 1, characterized in that, The loss function of the joint prediction model is a hybrid loss function composed of the capacity inflection point prediction sub-model and the remaining lifetime prediction sub-model. The expression of the hybrid loss function is as follows: In the formula, For a mixed loss function, λ The loss weights for capacity inflection point prediction. L CTK Predicting losses for capacity inflection points, L RUL Predict losses for remaining lifetime.
6. A joint prediction system for battery capacity inflection point and remaining lifetime based on relaxation voltage, characterized in that, include: Data acquisition module: used to acquire the relaxation voltage data of the battery to be predicted after it is fully charged; Capacity inflection point prediction module: used to input the relaxation voltage data into the capacity inflection point prediction sub-model in the pre-trained joint prediction model, output the predicted number of remaining battery inflection cycles, and determine the battery capacity inflection point; Remaining life prediction module: This module combines the predicted remaining battery cycle count and relaxation voltage data, inputs them into the remaining life prediction sub-model in the joint prediction model, and outputs the predicted remaining life.
7. The battery capacity inflection point and remaining life joint prediction system according to claim 6, characterized in that, The steps for training the joint prediction model include: Acquire the capacity decay curves of multiple batteries, calculate the remaining transition cycle number and remaining lifespan of each battery cycle, and collect the relaxation voltage data after each full charge in each battery cycle. The relaxation voltage data after each full charge in each cycle of the battery is used as training sample one, and the corresponding remaining number of transition cycles is used as a label to train the capacity inflection point prediction sub-model. The remaining transition cycle number of each cycle of the battery and the relaxation voltage data after each full charge in each cycle of the battery are used as training sample two, and the corresponding remaining lifetime is used as the label to train the remaining lifetime prediction sub-model. The trained capacity inflection point prediction sub-model and remaining lifetime prediction sub-model are integrated into a joint prediction model to complete the training process of the joint prediction model.
8. The battery capacity inflection point and remaining life joint prediction system according to claim 7, characterized in that, The formulas for calculating the remaining transition cycle number and remaining lifespan of the battery for each cycle are as follows: in, CTK i , RUL i The battery number i The remaining number of transition loops and the remaining lifetime of the next cycle. C e , C k and C i These represent the cycle count at the end of life, the capacity inflection point, and the current cycle number, respectively. i Number of iterations.
9. The battery capacity inflection point and remaining life joint prediction system according to claim 8, characterized in that, The number of cycles at the capacity inflection point C k Identified by the Bacon-Watts model, the expression of which is: In the formula, Y This is the capacity decay curve from charge-discharge cycles to the end of the lifetime. Z It is a normally distributed random variable centered at zero, representing the residuals. α 0 is x = C k intercept at, α 1 and α 2. Adjust the slope of the two intersecting lines. γ A constant used to control the abruptness of transitions.
10. A joint prediction system for battery capacity inflection point and remaining life according to claim 6, characterized in that, The loss function of the joint prediction model is a hybrid loss function composed of the capacity inflection point prediction sub-model and the remaining lifetime prediction sub-model. The expression of the hybrid loss function is as follows: In the formula, For a mixed loss function, λ The loss weights for capacity inflection point prediction. L CTK Predicting losses for capacity inflection points, L RUL Predict losses for remaining lifetime.