Cascade utilization lithium ion battery capacity prediction method
By constructing a multi-parameter prediction model, using GITT titration experiments and historical dataset training, the problem that the Peukert equation cannot accurately predict the capacity of lithium-ion batteries under specific conditions is solved, and the effect of intelligent detection is achieved.
Patent Information
- Application Number
- CN202510147464.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-11
- Publication Date
- 2025-06-03
AI Technical Summary
The Peukert equation cannot accurately predict the capacity of lithium-ion batteries under specific conditions, and the degree of intelligence is difficult to meet engineering needs.
By collecting electrochemical performance data of lithium-ion batteries in different working environments, using GITT titration experiments to determine the chemical diffusion coefficient, construct a prediction model, and training the model based on the historical data set to achieve real-time capacity prediction.
The limitations of the prediction of the Peukert equation under specific conditions are effectively solved, the degree of intelligence of the model is improved, and the intelligent detection of the capacity of lithium-ion batteries is realized.
Smart Images
Figure CN120085184A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of new energy battery detection, and particularly to a method for predicting the capacity of a lithium-ion battery for secondary utilization. Background Art
[0002] With the rapid development of the new energy vehicle industry, power batteries are facing a retirement wave during the replacement process, and the secondary utilization technology has emerged as the times require. This technology can maximize the use of the entire life cycle of power batteries while alleviating the recycling pressure and environmental pollution problems. In recent years, with the implementation of the "carbon neutrality and carbon peak" strategy and the introduction of relevant policies, the production and sales of new energy vehicles in China have experienced explosive growth, making the lithium-ion batteries for secondary utilization an important research and application field.
[0003] Among them, the Peukert equation is a classic battery model used to predict the capacity of lithium-ion batteries at different discharge rates. Based on the changes in the capacity and internal resistance of the battery, this equation constructs an empirical model of battery aging and uses Peukert's law to predict the change in capacity with the standing time of the battery at different rates such as 0.1C, 0.5C, and 1.0C. The principle of the Peukert equation is to calculate the actual available capacity of the battery through the nominal capacity of the battery and the specified discharge rate.
[0004] Although the Peukert equation has certain reference value in battery capacity prediction, it has limitations. Because the Peukert equation can only be used to predict the battery capacity under constant temperature and constant current discharge and can only be applied to predictions under specific conditions. This means that in practical applications, the Peukert equation may not be able to accurately predict the battery performance under complex conditions such as variable temperature and variable current, limiting its application in a wider range of scenarios. Summary of the Invention
[0005] The embodiments of the present invention provide a method for predicting the capacity of a lithium-ion battery for secondary utilization, which at least solves the problem that the related technologies using the Peukert equation can only be applied to predictions under specific conditions and the degree of intelligence is difficult to meet the engineering requirements.
[0006] According to an embodiment of the present invention, there is provided a method for predicting the capacity of a lithium-ion battery for secondary utilization, including: collecting the electrochemical performance data of the lithium-ion battery for secondary utilization under different working environments, where the electrochemical performance data includes: internal resistance, voltage, current, temperature, and number of charge and discharge cycles; Determining the chemical diffusion coefficient of the lithium-ion battery based on the results of the GITT titration experiment; Construct a prediction model based on environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery, so as to capture the relationship among the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery based on the prediction model; Train the prediction model based on the historical data set and verify the prediction ability of the prediction model until the prediction model converges; wherein, the historical data set includes the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery; Use the converged prediction model to predict the capacity of the target lithium-ion battery based on real-time parameters, wherein the real-time parameters include the real-time electrochemical performance data, the real-time chemical diffusion coefficient, and the real-time environmental parameters of the target lithium-ion battery.
[0007] Through one embodiment of the present invention, due to the adoption of technical solutions such as comprehensively considering multiple parameters, applying GITT titration experiments, constructing a prediction model, training using a historical data set, and predicting using real-time parameters, the limitations of the Peukert equation in predicting under specific conditions are effectively solved, the intelligent level of the model is improved, and the effect of intelligently detecting the capacity of the lithium-ion battery is achieved. Description of the Drawings
[0008] Figure 1 is a flowchart of a method for predicting the capacity of a lithium-ion battery for cascade utilization according to an embodiment of the present invention; Figure 2 is a flowchart of a method for constructing a prediction model based on environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of a lithium-ion battery according to an embodiment of the present invention; Figure 3 is a flowchart of a method for fusing normalized data to obtain a comprehensive feature set according to an embodiment of the present invention; Figure 4 is a flowchart of a method for using the re-optimized parameter set as a comprehensive feature set according to an embodiment of the present invention; Figure 5 is a flowchart of a method for obtaining a re-optimized parameter set based on the non-linear least squares method according to an embodiment of the present invention; Figure 6 is a flowchart of a method for obtaining a comprehensive feature set based on a fused data set according to an embodiment of the present invention; Figure 7 is a flowchart of a method for converging a hybrid algorithm model module based on the non-linear least squares method according to an embodiment of the present invention; Figure 8 is a structural block diagram of a device for predicting the capacity of a lithium-ion battery for cascade utilization according to an embodiment of the present invention. Detailed implementation manners
[0009] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0010] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
[0011] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.
[0012] In an embodiment of the present invention, a method for predicting the capacity of a lithium-ion battery for cascade utilization is provided. Figure 1 It is a flowchart of the method for predicting the capacity of a lithium-ion battery for cascade utilization according to an embodiment of the present invention, as Figure 1 shown, and the process includes the following steps: Step S101, collect the electrochemical performance data of the lithium-ion battery for cascade utilization under different working environments, and the electrochemical performance data includes: internal resistance, voltage, current, temperature, number of charge and discharge cycles; In an exemplary implementation manner, this step not only considers the discharge rate of the battery, but also comprehensively considers multiple parameters such as internal resistance, voltage, current, temperature, number of charge and discharge cycles, and chemical diffusion coefficient. This way of considering multiple parameters enables the model to more comprehensively capture the performance changes of the battery under different working environments. Compared with the Peukert equation which mainly depends on the discharge current and the battery capacity, the present invention improves the applicability and accuracy of the model by considering more battery performance parameters, making it no longer limited to the prediction under specific conditions.
[0013] Step S102, determine the chemical diffusion coefficient of the lithium-ion battery based on the results of the GITT titration experiment; In an exemplary implementation manner, by determining the chemical diffusion coefficient of the lithium-ion battery based on the results of the GITT titration experiment, the present invention can more accurately grasp the lithium-ion migration situation inside the battery, thereby improving the accuracy of prediction. The application of the GITT titration experiment provides a more accurate chemical diffusion coefficient for the model, which is not available in the Peukert equation, thus improving the model's understanding of the dynamics inside the battery.
[0014] Step S103: Construct a prediction model based on environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery, so as to capture the relationships among the environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery based on the prediction model; In an exemplary embodiment, this step constructs a prediction model based on environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters to capture the relationships among these parameters, enabling the model to dynamically adapt to different working conditions and battery states. By constructing a comprehensive prediction model, the present invention can capture and learn the complex relationships among battery performance parameters, which cannot be achieved by the traditional Peukert equation.
[0015] Step S104: Train the prediction model based on the historical data set and verify the prediction ability of the prediction model until the prediction model converges; wherein, the historical data set includes environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery; In an exemplary embodiment, this step uses the historical data set to train the prediction model and verify the prediction ability of the model until the model converges. This method enables the model to learn and optimize based on a large amount of actual data, improving the reliability and accuracy of the prediction. Training and verifying the model using the historical data set enables the model to learn and optimize from the actual data, improving the generalization ability and prediction accuracy of the model.
[0016] Step S105: Use the converged prediction model to predict the capacity of the target lithium-ion battery based on real-time parameters, where the real-time parameters include real-time electrochemical performance data, real-time chemical diffusion coefficient, and real-time environmental parameters of the target lithium-ion battery.
[0017] In an exemplary embodiment, this step uses the converged prediction model to predict the capacity of the target lithium-ion battery based on real-time parameters, which includes real-time electrochemical performance data, real-time chemical diffusion coefficient, and real-time environmental parameters. This real-time prediction ability enables the model to adapt to the dynamic changes in battery performance. The introduction of real-time parameters enables the model to dynamically respond to changes in battery state, providing a more flexible and real-time prediction ability.
[0018] Through the above steps S101 to S105, the present invention effectively solves the limitations of the Peukert equation in predicting under specific conditions and improves the intelligence level of the model by comprehensively considering multiple parameters, applying GITT titration experiments, constructing a prediction model, training using a historical data set, and real-time parameter prediction, achieving the effect of intelligent detection of the capacity of lithium-ion batteries.
[0019] In one embodiment, Figure 2is a flowchart of a method for constructing a prediction model based on environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of a lithium-ion battery according to an embodiment of the present invention. As Figure 2 shown, the method further includes: Step S201, normalizing the environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery to obtain normalized data; In an exemplary embodiment, for example, normalizing the environmental parameters (such as temperature, humidity), electrochemical performance data (such as internal resistance, voltage, current, charge-discharge cycle times), chemical diffusion coefficient, and capacity parameters of the lithium-ion battery. Normalization can be achieved by methods such as min-max scaling or Z-score standardization, aiming to eliminate the dimensionality influence between different parameters, make the data on the same scale, and facilitate subsequent model training. Therefore, the normalized data can accelerate the convergence speed of the model, improve the stability of model training and the accuracy of prediction. Moreover, the normalized data provides preprocessed data for subsequent model training and real-time parameter prediction, enabling the model to more accurately predict battery performance under different working environments.
[0020] Step S202, fusing the normalized data to obtain a comprehensive feature set; wherein, the historical data set includes the comprehensive feature set; In an exemplary embodiment, for example, fusing the normalized data to form a comprehensive feature set. This can be achieved by simple data splicing or more complex feature engineering techniques such as principal component analysis (PCA), etc., to extract the most informative features. Therefore, the fused comprehensive feature set can more comprehensively represent the state of the battery and provide rich input features for the deep learning model. Moreover, the comprehensive feature set, as the basis for training and validating the prediction model, enables the model to capture complex patterns of battery performance changes.
[0021] Step S203, based on the deep learning model module, capturing the relationships among the environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery in the comprehensive feature set; In an exemplary embodiment, for example, based on the deep learning model module, capturing the relationships among the environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery in the comprehensive feature set. The models that can be used include convolutional neural network (CNN), recurrent neural network (RNN), long short-term memory network (LSTM), etc. Therefore, the deep learning model can learn the non-linear relationships and complex interaction patterns among the features, improving the accuracy of prediction. Moreover, the training results of the deep learning model can be directly used for real-time parameter prediction, providing fast and accurate prediction capabilities.
[0022] Step S204: Predict the capacity of the target lithium-ion battery based on the composite machine learning model module and real-time parameters; Among them, the prediction model includes a deep learning model module and a composite machine learning model module.
[0023] In an exemplary embodiment, for example, based on the composite machine learning model module, combining the output of the deep learning model and real-time parameters to predict the capacity of the target lithium-ion battery. The composite model can be an ensemble learning model, such as random forest, gradient boosting tree, etc., which can combine the prediction results of multiple models to improve the robustness of the prediction. Therefore, the composite model can integrate the advantages of multiple models to improve the accuracy and reliability of the prediction. Moreover, the prediction results of the composite model can be directly used for real-time monitoring and evaluation of the battery state, providing decision support for battery management.
[0024] Through the above steps S201 to S204, the present invention not only improves the intelligence level of lithium-ion battery capacity prediction, but also realizes real-time monitoring and evaluation of battery performance, forming a complete prediction and application process with steps S101 to S105.
[0025] In one embodiment, Figure 3 is a flowchart of a method for fusing normalized data to obtain a comprehensive feature set according to an embodiment of the present invention, as Figure 3 shown, fusing normalized data to obtain a comprehensive feature set includes: Step S301: Identify the normalized data based on the hybrid algorithm model module to obtain identification parameters; among them, the identification parameters include: positive and negative electrode solid-phase diffusion coefficients, insertion / extraction reaction rates, environmental parameters, electrochemical performance data, chemical diffusion coefficients, and capacity parameters of the lithium-ion battery; In an exemplary embodiment, the hybrid algorithm model module can combine multiple algorithms such as genetic algorithm (GA), particle swarm optimization (PSO), and nonlinear least squares method (NLS) to identify the normalized data. For example, the genetic algorithm is used for global optimization to find the optimal parameter combination; the particle swarm optimization algorithm is used for local search to refine the parameter estimation; the nonlinear least squares method is used for the initial estimation of parameters and the evaluation of optimized parameters. Therefore, through the hybrid algorithm model module, key parameters such as positive and negative electrode solid-phase diffusion coefficients and insertion / extraction reaction rates can be identified more accurately, which are crucial for understanding the electrochemical behavior of the battery. Moreover, through accurate parameter identification, the ability of the model to capture battery performance changes is improved, making the prediction model more accurate and reliable.
[0026] Step S302: Fuse the identification parameters to obtain a comprehensive feature set.
[0027] In an exemplary embodiment, in this step, the parameters identified in step S301 are fused with the normalized data to form a comprehensive feature set. This step can be achieved through data fusion techniques such as principal component analysis (PCA) or stacking model, which integrate data from different sources to improve the accuracy and robustness of prediction. Therefore, the fused comprehensive feature set contains more comprehensive information, can better represent the state of the battery, and provides richer input features for subsequent deep learning and composite machine learning models. Moreover, as the input of the deep learning model and the composite machine learning model, the comprehensive feature set enables the model to use more comprehensive data for training and prediction, improving the accuracy and real-time performance of prediction.
[0028] Through the above steps S301 to S302, the present invention not only improves the intelligence level of lithium-ion battery capacity prediction, but also realizes real-time monitoring and evaluation of battery performance, forming a complete prediction and application process with steps S101 to S105 and steps S201 to S204. This technical solution that comprehensively considers multiple parameters, applies a hybrid algorithm model module, constructs a prediction model, trains using a historical data set, and predicts real-time parameters effectively solves the limitations of the Peukert equation in prediction under specific conditions, improves the intelligence level of the model, and achieves the effect of intelligent detection of the capacity of lithium-ion batteries.
[0029] In one embodiment, Figure 4 is a flowchart of a method based on the re-optimized parameter set as the comprehensive feature set according to an embodiment of the present invention, as Figure 4 shown, the method further includes: Step S401, initializing the population of the genetic algorithm, the particle positions and velocities of the particle swarm algorithm, and the initial parameter estimation of the nonlinear least squares method based on the normalized data to obtain initial data; wherein, the initial data includes: the initialized population, the initialized particle positions and velocities, and the initialized initial parameter estimation; the hybrid algorithm model module includes: the genetic algorithm, the particle swarm algorithm, and the nonlinear least squares method. In an exemplary embodiment, the population of the genetic algorithm, the particle positions and velocities of the particle swarm algorithm, and the initial parameter estimation of the nonlinear least squares method are initialized using the normalized data. This step is the starting point of the hybrid algorithm model module, where the population represents a set of possible solutions, the particle positions and velocities guide the search direction, and the initial parameter estimation provides a starting point for the nonlinear least squares method. Therefore, through initialization, a data-driven starting point is provided for the subsequent optimization algorithms, which helps the algorithms converge quickly to better solutions. Moreover, the initialized parameters provide a basis for model training and prediction, enabling the model to learn the relationships between battery performance parameters from an optimized starting point.
[0030] Step S402: Optimize the initial parameter estimation based on the nonlinear least squares method to obtain the optimized parameter estimation. In an exemplary embodiment, the nonlinear least squares method is applied to optimize the initial parameter estimation to obtain a more accurate parameter estimation. This involves minimizing the sum of the squares of the residuals, thereby improving the accuracy of the model prediction. Therefore, the optimized parameter estimation can more accurately reflect the actual performance of the battery, improving the prediction accuracy of the model. Moreover, the optimized parameter estimation can be directly used in deep learning and composite machine learning models, improving the model's prediction ability for the battery state.
[0031] Step S403: Perform selection, crossover, and mutation operations on the optimized parameter estimation based on the genetic algorithm to search for the optimal parameters, obtain the initially optimized parameter set, and use the nonlinear least squares method to evaluate the fitness. In an exemplary embodiment, the genetic algorithm is used to perform selection, crossover, and mutation operations on the optimized parameter estimation to search for the optimal parameter set. This step simulates the process of natural selection and finds a better parameter combination through iterative evolution. Therefore, through the global search ability of the genetic algorithm, a better parameter set can be obtained, improving the fitness and prediction accuracy of the model. Moreover, the parameter set optimized by the genetic algorithm provides better parameters for the hybrid algorithm model module, enhancing the model's ability to capture changes in battery performance.
[0032] Step S404: Based on the particle swarm algorithm, make the particles update their positions and velocities in the solution space of the initially optimized parameter set according to the experience of individuals and the group to obtain the re-optimized parameter set, and use the nonlinear least squares method to evaluate the fitness. In an exemplary embodiment, the particle swarm algorithm is used to update the particle positions and velocities, and search for the optimal solution in the solution space according to the experience of individuals and the group. The particle swarm algorithm finds the global optimal solution by simulating the social behavior of bird flocks or fish schools. Therefore, the particle swarm algorithm can quickly converge and find a better parameter set, improving the search efficiency of the model. Moreover, the parameter set optimized by the particle swarm algorithm provides a basis for real-time parameter prediction, enabling the model to quickly respond to changes in the battery state.
[0033] Step S405: Use the re-optimized parameter set as the comprehensive feature set.
[0034] In an exemplary embodiment, the re-optimized parameter set is fused with the normalized data to form a comprehensive feature set. This step combines the results of the optimization algorithm with the original data, providing a more comprehensive feature representation for the model. Therefore, the comprehensive feature set contains the key information refined by the optimization algorithm and the details of the original data, providing rich input features for the model. Moreover, as the input of the deep learning and composite machine learning models, the comprehensive feature set improves the model's prediction ability for battery performance, achieving the effect of intelligent detection of the capacity of lithium-ion batteries.
[0035] Through the above steps S401 to S405, the present invention not only improves the intelligence level of lithium-ion battery capacity prediction, but also realizes the real-time monitoring and evaluation of battery performance, forming a complete prediction and application process with steps S101 to S105, steps S201 to S204, and steps S301 to S302. This technical solution that comprehensively considers multiple parameters, applies a hybrid algorithm model module, constructs a prediction model, trains using a historical data set, and predicts real-time parameters effectively solves the limitations of the Peukert equation in prediction under specific conditions and improves the intelligence level of the model.
[0036] In one embodiment, Figure 5 is a flowchart of a method for obtaining a re-optimized parameter set based on the non-linear least squares method according to an embodiment of the present invention, as Figure 5 shown, the method further includes: Step S501, feeding back the initially optimized parameter set and the re-optimized parameter set into the non-linear least squares method to obtain a refined parameter estimate until the hybrid algorithm model module converges, so that the genetic algorithm obtains the initially optimized parameter set based on the refined parameter estimate; In an exemplary embodiment, this step involves feeding back the initially optimized parameter set and the re-optimized parameter set into the non-linear least squares method for iterative optimization. The non-linear least squares method is a numerical optimization technique for solving non-linear model parameter estimation, which can find the best fitting parameters by minimizing the sum of the squares of the residuals. In this step, the algorithm continuously iteratively updates the parameter estimation until the hybrid algorithm model module converges. Therefore, by refining the parameter estimation, more accurate model parameters can be obtained, thereby improving the accuracy and reliability of model prediction. Moreover, in combination with steps S101 to S105, the refined parameter estimation can improve the model's ability to capture changes in the performance of lithium-ion batteries, making the prediction model more accurate and reliable. In combination with steps S201 to S204, the refined parameter estimation can be directly used in deep learning and composite machine learning models, improving the model's prediction ability for battery states. In combination with steps S301 to S302, the refined parameter estimation provides better parameters for the hybrid algorithm model module, enhancing the model's ability to capture changes in battery performance. In combination with steps S401 to S405, the refined parameter estimation provides a basis for real-time parameter prediction, enabling the model to quickly respond to changes in battery states.
[0037] Step S502: Obtain the re-optimized parameter set based on the particle swarm optimization algorithm using the initially optimized parameter set.
[0038] In an exemplary embodiment, in this step, the re-optimized parameter set is obtained based on the particle swarm optimization (PSO) algorithm using the initially optimized parameter set. The particle swarm optimization algorithm is an optimization algorithm that simulates the social behavior of bird flocks or fish schools to find the global optimal solution. In this step, each particle represents a set of parameters, and by updating the position and velocity of the particles, the algorithm searches for the optimal solution in the solution space. Therefore, the particle swarm optimization algorithm can quickly converge and find a better parameter set, improving the model's search efficiency and prediction accuracy. Moreover, in combination with steps S101 to S105, the parameter set optimized by the particle swarm optimization algorithm provides a basis for real-time parameter prediction, enabling the model to quickly respond to changes in battery states. In combination with steps S201 to S204, the parameter set optimized by the particle swarm optimization algorithm can be directly used in deep learning and composite machine learning models, improving the model's prediction ability for battery performance. In combination with steps S301 to S302, the parameter set optimized by the particle swarm optimization algorithm provides better parameters for the hybrid algorithm model module, enhancing the model's ability to capture changes in battery performance. In combination with steps S401 to S405, the parameter set optimized by the particle swarm optimization algorithm provides decision support for real-time monitoring and evaluating the state of the battery, achieving the effect of intelligent detection of the capacity of lithium-ion batteries.
[0039] Through the implementation of the above steps S501 to S502, the present invention not only improves the intelligence level of lithium-ion battery capacity prediction, but also realizes the real-time monitoring and evaluation of battery performance, forming a complete prediction and application process with steps S101 to S105, steps S201 to S204, steps S301 to S302, and steps S401 to S405. This technical solution that comprehensively considers multiple parameters, applies a hybrid algorithm model module, constructs a prediction model, trains using a historical data set, and predicts real-time parameters effectively solves the limitations of the Peukert equation in prediction under specific conditions and improves the intelligence level of the model.
[0040] In one embodiment, Figure 6 is a flowchart of a method for obtaining a comprehensive feature set based on a fusion data set according to an embodiment of the present invention, as Figure 6 shown, the method further includes: Step S601, fusing the parameter set optimized again with the normalized data to obtain a fusion data set; In an exemplary embodiment, in this step, the parameter set optimized again is fused with the normalized data to form a fusion data set. This process can be achieved through feature fusion technology, such as the feature-level fusion method, which combines data sets from different sources into a new data set, so that the model can understand the performance state of the battery from multiple perspectives. Therefore, the fusion data set contains the key information refined by the optimization algorithm and the details of the original data, providing a more comprehensive feature representation for the model. This helps improve the model's ability to capture changes in battery performance, making the prediction model more accurate and reliable. Moreover, combined with steps S101 to S105, the fusion data set provides a more comprehensive view of battery performance, enabling the model to more accurately predict battery performance under different working environments. Combined with steps S201 to S204, the fusion data set, as the input of the deep learning and composite machine learning models, improves the model's prediction ability for battery states. Combined with steps S301 to S302, the fusion data set provides better parameters for the hybrid algorithm model module, enhancing the model's ability to capture changes in battery performance. Combined with steps S401 to S405, the fusion data set provides a basis for real-time parameter prediction, enabling the model to quickly respond to changes in battery states. Combined with steps S501 to S502, the fusion data set combines the parameter sets optimized by the nonlinear least squares method and the particle swarm algorithm, further improving the prediction accuracy and robustness of the model.
[0041] Step S602, using the fusion data set as the comprehensive feature set.
[0042] In an exemplary embodiment, in this step, the fused dataset obtained in step S601 is used as the comprehensive feature set for subsequent model training and prediction. This comprehensive feature set can be regarded as a feature library containing rich information, which is used to improve the prediction performance of the model. Therefore, the establishment of the comprehensive feature set enables the model to use more comprehensive data for training and prediction, improving the accuracy and real-time performance of the prediction. Moreover, combined with steps S101 to S105, the comprehensive feature set provides a more accurate basis for battery performance prediction for the model, enabling the model to more accurately predict battery performance under different working environments. Combined with steps S201 to S204, the comprehensive feature set, as the input of the deep learning and composite machine learning models, improves the model's prediction ability for battery states. Combined with steps S301 to S302, the comprehensive feature set provides better parameters for the hybrid algorithm model module, enhancing the model's ability to capture changes in battery performance. Combined with steps S401 to S405, the comprehensive feature set provides decision-making support for real-time monitoring and evaluating the state of the battery, achieving the effect of intelligent detection of the capacity of lithium-ion batteries. Combined with steps S501 to S502, the comprehensive feature set combines the parameter sets optimized by the nonlinear least squares method and the particle swarm algorithm, further improving the prediction accuracy and robustness of the model.
[0043] Through the implementation of the above steps S601 to S602, the present invention not only improves the intelligence level of lithium-ion battery capacity prediction, but also realizes the real-time monitoring and evaluation of battery performance, forming a complete prediction and application process with steps S101 to S105, steps S201 to S204, steps S301 to S302, steps S401 to S405, and steps S501 to S502. This technical solution that comprehensively considers multiple parameters, applies a hybrid algorithm model module, constructs a prediction model, trains using historical datasets, and predicts real-time parameters effectively solves the limitations of the Peukert equation in prediction under specific conditions and improves the intelligence level of the model.
[0044] In one embodiment, Figure 7 is a flowchart of a method for converging a hybrid algorithm model module based on the nonlinear least squares method according to an embodiment of the present invention. As Figure 7 shown, the method further includes: Step S701, setting environmental parameters, electrochemical performance data, chemical diffusion coefficients, and capacity parameters of the lithium-ion battery as the model parameter vector; In an exemplary embodiment, in this step, environmental parameters (such as temperature, humidity), electrochemical performance data (such as internal resistance, voltage, current, number of charge-discharge cycles), chemical diffusion coefficient, and capacity parameters of the lithium-ion battery are set as the model parameter vector. These parameter vectors are the basis for model prediction and contain all the key variables affecting battery performance. Therefore, by defining the model parameter vector, all relevant factors can be systematically considered, providing comprehensive data support for subsequent model function definition and prediction. Moreover, combined with steps S101 to S105, these parameter vectors provide a basis for multi-parameter comprehensive consideration, enabling the model to capture battery performance changes more comprehensively. Combined with steps S201 to S204, the normalized parameter vectors can be directly used for the training and prediction of deep learning and composite machine learning models.
[0045] Step S702, define a model function based on the model parameter vector; In an exemplary embodiment, a model function is defined based on the model parameter vector, and this function describes the relationship between battery performance and the parameter vector. This may involve complex non-linear relationships and needs to be determined according to the physical and chemical characteristics of the battery. Therefore, the establishment of the model function provides a mathematical basis for predicting battery performance, enabling the prediction of battery capacity and other performance indicators based on changes in the parameter vector. Moreover, combined with steps S301 to S302, the model function can combine the results of the hybrid algorithm model module to improve the accuracy of parameter estimation. Combined with steps S401 to S405, the model function can utilize the optimized parameter set to improve the accuracy of prediction.
[0046] Step S703, calculate the residual between the model prediction value and the actual observed value based on the model function; In an exemplary embodiment, the residual between the model prediction value and the actual observed value is calculated based on the model function. The residual is a direct reflection of the model prediction accuracy, and the smaller the residual, the stronger the model's prediction ability. Therefore, the calculation of the residual provides feedback for model optimization, and the model parameters can be optimized by minimizing the residual. Moreover, combined with steps S501 to S502, the minimization of the residual can be achieved through non-linear least squares method and particle swarm algorithm, further improving the prediction accuracy of the model.
[0047] Step S704, determine the objective function based on the residual; In an exemplary embodiment, a target function is determined based on residuals. Generally, the target function is the sum of the squares of the residuals, which is a common form in the least squares method. Therefore, the determination of the target function provides a clear direction for the optimization of model parameters, that is, to find the optimal model parameters by minimizing the target function. Moreover, in combination with steps S601 to S602, the minimization of the target function can be achieved by fusing the data set, further improving the prediction performance of the model.
[0048] Step S705, update the model parameter vector based on a preset algorithm to minimize the target function.
[0049] In an exemplary embodiment, the model parameter vector is updated based on a preset algorithm (such as the Levenberg-Marquardt algorithm) to minimize the target function. This involves an iterative process of the algorithm, continuously adjusting the parameter vector until the optimal solution is found. Therefore, by updating the model parameter vector, the prediction accuracy and adaptability of the model can be significantly improved, enabling the model to better adapt to the changes in battery performance. Moreover, in combination with steps S101 to S105, the updated parameter vector can more accurately capture the changes in battery performance and improve the reliability of the prediction. In combination with steps S201 to S204, the updated parameter vector can be directly used in deep learning and composite machine learning models to improve the accuracy and real-time performance of the prediction.
[0050] Through the implementation of the above steps S701 to S705, the present invention not only improves the intelligence level of lithium-ion battery capacity prediction, but also realizes the real-time monitoring and evaluation of battery performance, forming a complete prediction and application process with steps S101 to S105, steps S201 to S204, steps S301 to S302, steps S401 to S405, steps S501 to S502, and steps S601 to S602. This technical solution that comprehensively considers multiple parameters, applies a hybrid algorithm model module, constructs a prediction model, trains using a historical data set, and predicts real-time parameters effectively solves the limitations of the Peukert equation in prediction under specific conditions and improves the intelligence level of the model.
[0051] In one embodiment, the model function is: ; where is the model parameter vector, , is the capacity parameter of the lithium-ion battery, is the internal resistance of the lithium-ion battery, is the voltage of the lithium-ion battery, is the current of the lithium-ion battery, is the operating temperature of the lithium-ion battery is the number of charge and discharge cycles of the lithium-ion battery, is the chemical diffusion coefficient of the lithium-ion battery.
[0052] In an exemplary embodiment, by incorporating these parameters into the model function, the embodiments of the present invention can provide a comprehensive battery performance prediction framework. The model function can not only capture the performance changes of the battery under different operating conditions, but also adapt to the effects of battery aging and performance degradation. This comprehensive prediction method based on multiple parameters can provide more accurate and reliable prediction results compared to the traditional Peukert equation, especially under complex conditions such as variable temperature and variable current, with significant advantages. In this way, the present invention realizes the intelligent detection of the capacity of lithium-ion batteries, providing strong support for the battery management system.
[0053] In one embodiment, the objective function is: ; where, is the residual, .
[0054] In an exemplary embodiment, in the embodiments of the present invention, the objective function adopted is to minimize the difference between the model prediction value and the actual observed value, that is, the sum of squared residuals. The role of this objective function is to measure the accuracy of the model prediction and guide the optimization process of the model parameters. By incorporating these parameters into the objective function, the embodiments of the present invention can provide a comprehensive battery performance prediction framework. The objective function can not only capture the performance changes of the battery under different operating conditions, but also adapt to the effects of battery aging and performance degradation. This comprehensive prediction method based on multiple parameters can provide more accurate and reliable prediction results compared to the traditional Peukert equation, especially under complex conditions such as variable temperature and variable current, with significant advantages. In this way, the present invention realizes the intelligent detection of the capacity of lithium-ion batteries, providing strong support for the battery management system.
[0055] In one embodiment, the functional expression of the preset algorithm is: ; where, , , is the residual, is the objective function, , are any two elements in the model parameter vector, is an element of the Jacobian matrix, representing the partial derivative of the residual with respect to the parameter , is an element of the Hessian matrix, representing the objective function with respect to the parameter and is the second-order partial derivative.
[0056] In an exemplary embodiment, through the function expression of the above preset algorithm, the LMA algorithm (Levenberg-Marquardt algorithm) can effectively handle the non-linear least squares problem and find the optimal model parameter vector. The dynamic adjustment mechanism of the LMA algorithm enables the algorithm to automatically adjust the damping factor according to the size of the residual and the condition number of the Jacobian matrix, thereby improving the search efficiency while ensuring the search stability. This algorithm is particularly suitable for dealing with the battery performance prediction problem with complex non-linear relationships and can provide more accurate and reliable prediction results.
[0057] In summary, by adopting the LMA algorithm as the preset algorithm, the present invention can effectively solve the limitations of the Peukert equation in prediction under specific conditions, improve the intelligence level of the model, and realize the intelligent detection of the capacity of lithium-ion batteries.
[0058] In one embodiment, the solid-phase diffusion coefficients of the positive and negative electrodes and the insertion / extraction reaction rates can be calculated from the chemical diffusion coefficient.
[0059] In an exemplary embodiment, for a reversible system controlled by the diffusion step, the chemical diffusion coefficient can be measured by cyclic voltammetry. The peak current and the diffusion coefficient are related by the Randles-Sevcik equation: ; where is the number of electrons participating in the reaction, is the electrode area immersed in the solution, is the diffusion coefficient of Li in the electrode, is the scan rate, is the change in the Li concentration before and after the reaction.
[0060] In an exemplary embodiment, the diffusion coefficient of the material can be obtained from the EIS data and calculated using the Warburg impedance. The relationship between the Warburg coefficient and the diffusion coefficient is: ; where is the gas constant, is the absolute temperature, is the number of electrons participating in the reaction, is the Faraday constant, is the electrode area.
[0061] In an exemplary embodiment, the diffusion coefficients of the positive and negative electrode materials are calculated by the PITT method, and the formula is: ; wherein, is the current value, is the diffusion coefficient of Li in the electrode, is the area of the active material, is the thickness of the active material, is the concentration of lithium ions, is the initial concentration of lithium ions.
[0062] In an exemplary embodiment, GITT measures the change of the electrode potential with time during a constant current process, and the formula for calculating the diffusion coefficient is: ; wherein, is the constant current, is the time, is the number of electrons participating in the reaction, is the Faraday constant, is the electrode area, is the potential change, is the change in the concentration of lithium ions.
[0063] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software adding the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.
[0064] In the embodiments of the present invention, a capacity prediction device for second-life utilization of lithium-ion batteries is further provided. This device is used to implement the above embodiments and preferred implementation methods, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.
[0065] Figure 8 is a structural block diagram of a capacity prediction device for the cascaded utilization of lithium-ion batteries according to an embodiment of the present invention. As Figure 8 shown, the device includes: A first acquisition module 81, configured to collect the electrochemical performance data of the cascaded utilization lithium-ion battery under different working environments, where the electrochemical performance data includes: internal resistance, voltage, current, temperature, and charge-discharge cycle times; A second acquisition module 82, configured to determine the chemical diffusion coefficient of the lithium-ion battery based on the results of the GITT titration experiment; A model construction module 83, configured to construct a prediction model based on the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery, so as to capture the relationship between the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery based on the prediction model; A deep learning model module 84, configured to train the prediction model based on a historical data set and verify the prediction ability of the prediction model until the prediction model converges; wherein, the historical data set includes the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery; A composite machine learning model module 85, configured to use the converged prediction model to predict the capacity of a target lithium-ion battery based on real-time parameters, where the real-time parameters include the real-time electrochemical performance data, real-time chemical diffusion coefficient, and real-time environmental parameters of the target lithium-ion battery; Wherein, the prediction model includes the deep learning model module and the composite machine learning model module.
[0066] In an implementation manner, the device is further configured to: perform normalization processing on the environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery to obtain normalized data; Fuse the normalized data to obtain a comprehensive feature set; wherein, the historical data set includes the comprehensive feature set; Capture the relationship between the environmental parameters, electrochemical performance data, chemical diffusion coefficient, and capacity parameters of the lithium-ion battery in the comprehensive feature set based on the deep learning model module; Predict the capacity of the target lithium-ion battery based on the real-time parameters based on the composite machine learning model module; Wherein, the prediction model includes the deep learning model module and the composite machine learning model module.
[0067] In an implementation manner, the device is further configured to: The normalized data is identified based on the hybrid algorithm model module to obtain identification parameters; wherein, the identification parameters include: the positive and negative solid-phase diffusion coefficients, the insertion / extraction reaction rates, environmental parameters, electrochemical performance data, chemical diffusion coefficients, and capacity parameters of the lithium-ion battery. The identification parameters are fused to obtain a comprehensive feature set.
[0068] In one embodiment, the device is further configured to: The population of the genetic algorithm, the particle positions and velocities of the particle swarm algorithm, and the initial parameter estimates of the nonlinear least squares method are initialized based on the normalized data to obtain initial data; wherein, the initial data includes: the initialized population, the initialized particle positions and velocities, and the initialized initial parameter estimates; the hybrid algorithm model module includes: the genetic algorithm, the particle swarm algorithm, and the nonlinear least squares method. The initial parameter estimates are optimized based on the nonlinear least squares method to obtain optimized parameter estimates. Selection, crossover, and mutation operations are performed on the optimized parameter estimates based on the genetic algorithm to search for the optimal parameters, so as to obtain a set of initially optimized parameters, and the fitness is evaluated using the nonlinear least squares method. Based on the particle swarm algorithm, the particles update their positions and velocities in the solution space of the set of initially optimized parameters according to the individual and group experience, so as to obtain a set of re-optimized parameters, and the fitness is evaluated using the nonlinear least squares method. The set of re-optimized parameters is used as the comprehensive feature set.
[0069] In one embodiment, the device is further configured to: The set of initially optimized parameters and the set of re-optimized parameters are fed back into the nonlinear least squares method to obtain refined parameter estimates until the hybrid algorithm model module converges, so that the genetic algorithm obtains the set of initially optimized parameters based on the refined parameter estimates. Based on the particle swarm algorithm, the set of re-optimized parameters is obtained using the set of initially optimized parameters.
[0070] In one embodiment, the device is further configured to: The set of re-optimized parameters is fused with the normalized data to obtain a fused data set. The fused data set is used as the comprehensive feature set.
[0071] In one embodiment, the device is further configured to: The environmental parameters, electrochemical performance data, chemical diffusion coefficients, and capacity parameters of the lithium-ion battery are set as the model parameter vector. A model function is defined based on the model parameter vector. Calculate the residual between the model prediction value and the actual observation value based on the model function; Determine the objective function based on the residual; Update the model parameter vector based on a preset algorithm to minimize the objective function.
[0072] In one embodiment, the model function is: ; where, is the model parameter vector, , is the capacity parameter of the lithium-ion battery, is the internal resistance of the lithium-ion battery, is the voltage of the lithium-ion battery, is the current of the lithium-ion battery, is the operating temperature of the lithium-ion battery, is the number of charge and discharge cycles of the lithium-ion battery, is the chemical diffusion coefficient of the lithium-ion battery.
[0073] In one embodiment, the model function is: The objective function is: ; where, is the residual, .
[0074] In one embodiment, the model function is The function expression of the preset algorithm is: ; where, , , is the residual, is the objective function, 、 are any two elements in the model parameter vector, is an element of the Jacobian matrix, representing the partial derivative of the residual with respect to the parameter , is an element of the Hessian matrix, representing the second-order partial derivative of the objective function with respect to the parameters and .
[0075] An embodiment of the present invention also provides a computer-readable storage medium, in which a computer program is stored. Wherein, the computer program is configured to execute the steps in any one of the above method embodiments when running.
[0076] In an exemplary embodiment, the above computer-readable storage medium may include, but is not limited to: various media such as USB flash drives, read-only memory (ROM for short), random access memory (RAM for short), external hard drives, magnetic disks, or optical discs that can store computer programs.
[0077] An embodiment of the present invention also provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.
[0078] In an exemplary embodiment, the above electronic device may further include a transmission device and input / output devices. Among them, the transmission device is connected to the above processor, and the input / output devices are connected to the above processor.
[0079] An embodiment of the present invention also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the steps of the methods described in various embodiments of the present invention are implemented.
[0080] The specific examples in this embodiment may refer to the examples described in the above embodiments and exemplary embodiments, and will not be repeated here.
[0081] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. They can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order than here, or they can be separately made into individual integrated circuit modules, or multiple modules or steps among them can be made into a single integrated circuit module to implement. In this way, the present invention is not limited to any specific combination of hardware and software.
[0082] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for predicting the capacity of a lithium-ion battery for cascade utilization, characterized in that: include: Collect electrochemical performance data of lithium-ion batteries used in cascade under different working environments, the electrochemical performance data including: internal resistance, voltage, current, temperature, and number of charge and discharge cycles; Determining the chemical diffusion coefficient of the lithium ion battery based on the GITT titration experiment results; Building a prediction model based on environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameter of the lithium-ion battery to capture the relationship among the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameter of the lithium-ion battery based on the prediction model; Training the prediction model based on a historical data set, and verifying the prediction ability of the prediction model until the prediction model converges; wherein the historical data set includes the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery; The converged prediction model is used to predict the capacity of a target lithium-ion battery based on real-time parameters, wherein the real-time parameters include real-time electrochemical performance data, real-time chemical diffusion coefficient, and real-time environmental parameters of the target lithium-ion battery.
2. The method according to claim 1, characterized in that Also includes: Normalizing the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameter of the lithium-ion battery to obtain normalized data; fusing the normalized data to obtain a comprehensive feature set; wherein the historical data set includes the comprehensive feature set; Capturing the relationship between the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery in the comprehensive feature set based on a deep learning model module; Predicting the capacity of the target lithium-ion battery based on the real-time parameters based on a composite machine learning model module; Among them, the prediction model includes the deep learning model module and the composite machine learning model module.
3. The method according to claim 2, characterized in that The normalized data is fused to obtain a comprehensive feature set, including: The normalized data is identified based on a hybrid algorithm model module to obtain identification parameters; wherein the identification parameters include: positive and negative electrode solid phase diffusion coefficients, insertion / deinsertion reaction rates, the environmental parameters, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameters of the lithium-ion battery; The identification parameters are fused to obtain the comprehensive feature set.
4. The method according to claim 3, characterized in that Also includes: Initializing the population of the genetic algorithm, the particle position and velocity of the particle swarm algorithm, and the initial parameter estimation of the nonlinear least squares method based on the normalized data to obtain initial data; wherein the initial data includes: the initialized population, the initialized particle position and velocity, and the initialized initial parameter estimation; the hybrid algorithm model module includes: the genetic algorithm, the particle swarm algorithm, and the nonlinear least squares method; Optimizing the initial parameter estimate based on the nonlinear least squares method to obtain an optimized parameter estimate; Performing selection, crossover, and mutation operations on the optimized parameter estimates based on a genetic algorithm to search for optimal parameters to obtain a parameter set after initial optimization, and evaluating fitness using the nonlinear least squares method; Based on the particle swarm algorithm, the particles are made to update their positions and velocities in the solution space of the initially optimized parameter set according to the experience of individuals and groups, so as to obtain a re-optimized parameter set, and the fitness is evaluated using the nonlinear least squares method; The re-optimized parameter set is used as the comprehensive feature set.
5. The method according to claim 4, characterized in that Also includes: Feeding back the initially optimized parameter set and the re-optimized parameter set to the nonlinear least squares method to obtain a refined parameter estimate, until the hybrid algorithm model module converges, so that the genetic algorithm obtains the initially optimized parameter set based on the refined parameter estimate; The re-optimized parameter set is obtained based on the particle swarm algorithm using the initially optimized parameter set.
6. The method according to claim 5, characterized in that Also includes: Fusion the re-optimized parameter set with the normalized data to obtain a fused data set; The fused data set is used as the comprehensive feature set.
7. The method according to claim 5, characterized in that Also includes: The environmental parameter, the electrochemical performance data, the chemical diffusion coefficient, and the capacity parameter of the lithium-ion battery are set as a model parameter vector; defining a model function based on the model parameter vector; Calculate the residual between the model prediction value and the actual observation value based on the model function; determining an objective function based on the residual; The model parameter vector is updated based on a preset algorithm to minimize the objective function.
8. The method according to claim 7, characterized in that The model function is: ； in, is the model parameter vector, , is the capacity parameter of lithium-ion battery, is the internal resistance of the lithium-ion battery, is the voltage of the lithium-ion battery, is the current of the lithium-ion battery, is the operating temperature of the lithium-ion battery, is the number of charge and discharge cycles of the lithium-ion battery, is the chemical diffusion coefficient of lithium-ion batteries.
9. The method according to claim 8, characterized in that The objective function is: ； in, is the residual, .
10. The method according to claim 7, characterized in that The function expression of the preset algorithm is: ； in, , , is the residual, is the objective function, , are any two elements in the model parameter vector, is the element of the Jacobian matrix, representing the residual Parameters The partial derivative of is the element of the Hessian matrix, representing the objective function Parameters and The second-order partial derivative of .