Battery life prediction method for estimating attenuation curve parameters based on LSTM (Long Short Term Memory)

Abstract

The invention relates to the technical field of batteries, in particular to a battery life prediction method for estimating attenuation curve parameters based on LSTM. Comprising the steps of collecting charging and discharging data of different batteries under each cycle, and establishing a battery pack database; on the basis of charging process data, screening charging fragments, and calculating tag capacity by using ampere-hour integral and an OCV-SOC correction method; according to the capacity attenuation curve track, fitting by using a parameterized model to obtain an aging coefficient set fitted by the capacity attenuation curve of the battery pack; extracting feature sets highly related to life prediction under early cycles of different batteries; taking the extracted feature set as the input of a model, taking the aging parameter set as the output of the model, and establishing a battery aging parameter estimation model based on LSTM; after the battery capacity aging parameters are obtained, the battery later cycle number is substituted into the parameterized model, and battery life prediction is obtained. According to the technical scheme, the battery life prediction precision under the condition of limited historical data can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of batteries, and particularly to a battery life prediction method for estimating decay curve parameters based on LSTM. Background Art

[0002] As an efficient energy storage device, lithium-ion batteries are widely used in fields such as electric vehicles and energy storage systems. As the battery usage time increases, its capacity gradually decays. Accurately predicting the remaining useful life (RUL) of the battery is of great significance for maintenance plan formulation, secondary utilization, and safety warning. Current battery life prediction technologies mainly include: Model-based methods, which describe the battery aging mechanism by establishing an electrochemical model or an empirical model. For example, an equivalent circuit model, an SEI film growth model, etc. are used to characterize the relationship between battery characteristics and capacity decay. However, this method relies on the accurate modeling of internal physical and chemical reactions of the battery, requires a large amount of experimental data to support parameter calibration, and has limited model generalization ability, making it difficult to adapt to the complexity of different battery types and operating conditions.

[0003] Data-driven methods. With the accumulation of battery monitoring data, machine learning algorithms (such as support vector regression SVR, Gaussian process regression GPR) and deep learning networks (such as convolutional neural network CNN, recurrent neural network RNN) have been introduced into the field of life prediction. Such methods extract features from historical charge and discharge data and construct a mapping relationship between capacity decay and cycle number. Although deep learning algorithms have the ability of automatic feature extraction and can directly process raw time series data, their performance highly depends on large-scale labeled data (such as the labeled capacity of a complete charge and discharge cycle). In practical applications, especially for real vehicle scenarios, the early cycle data of the battery is limited (such as data within the warranty period), and the battery failure cycle is long, making it difficult to obtain complete decay cycle data, resulting in insufficient model training and a significant decrease in prediction accuracy. Summary of the Invention

[0004] The purpose of the present invention is to propose a battery life prediction method for estimating decay curve parameters based on LSTM, which can improve the battery life prediction accuracy under the condition of limited historical data.

[0005] To achieve the above purpose, the present invention provides a battery life prediction method for estimating decay curve parameters based on LSTM, including: Collect charge and discharge data of different batteries for each cycle and establish a battery pack database; Based on the charging process data, screen the charging segments, and calculate the labeled capacity using the ampere-hour integration and OCV-SOC correction method; According to the capacity decay curve trajectory, use a parametric model for fitting to obtain a set of aging coefficients for fitting the capacity decay curve of the battery pack; Extract a feature set that is highly correlated with life prediction during the early cycles of different batteries; Use the extracted feature set as the input of the model and the aging parameter set as the output of the model, and establish a battery aging parameter estimation model based on LSTM; After obtaining the battery capacity aging parameters, substitute the number of later cycles of the battery into the parametric model to obtain the battery life prediction.

[0006] Beneficial effects of the basic solution: Traditional battery life prediction methods often require a large amount of historical charge and discharge data to train the model, which not only increases the difficulty of data collection but also limits the application of the model in real vehicle scenarios. In real vehicle scenarios, the battery pack can usually only obtain early operation data (such as data within the warranty period of electric vehicles), and it is often difficult to obtain battery data comprehensively and continuously. This method can directly use the charge and discharge segments recorded by the existing BMS to complete life prediction without waiting for the battery to completely fail. And this technical solution extracts an early cycle feature set that is highly correlated with life prediction and uses the LSTM model for aging parameter estimation, significantly reducing the demand for historical data. Even with only limited early data of the battery, the remaining life of the battery can be predicted relatively accurately.

[0007] This technical solution combines the parametric model with the data-driven method, uses the parametric model to describe the internal law of battery life attenuation, and adopts the LSTM network to establish a cross-cycle mapping relationship of "early features → aging parameters". By end-to-end learning of the non-linear correlation between the early performance degradation signal of the battery and the long-term attenuation trajectory parameters, it breaks through the limitation of traditional methods that require observing a complete attenuation cycle to model. The advantage of this technical solution is that it can flexibly select or adjust the parametric model according to different types of batteries and their attenuation characteristics, so as to more accurately capture the attenuation rate of the battery in different aging states. In addition, the LSTM model has a strong ability to process sequence data, can learn and remember the long-term dependence relationship in the battery charge and discharge data, and further improve the prediction accuracy. It not only enhances the model's understanding of the internal mechanism of battery life attenuation but also improves the generalization ability of the model in different battery types and aging states, making the prediction results more reliable and stable.

[0008] As an implementable and preferred solution, based on the charging process data, screen the charging segments, including the following: Use the interpolation method to fill in the missing data.

[0009] Analyze the battery cycle data, preprocess the data, and use the moving average filtering method to smooth the data. Let the window size be , for the data sequence , the result after moving average filtering is:

[0010] Among them, when exceeds the data range, the boundary value is taken; Screen out the charging segments that meet the requirements; based on the OCV-SOC table of the battery, correct the SOC based on the OCV to reduce the error in SOC estimation and obtain the accurate SOC.

[0011] As an implementable preferred solution, the labeled capacity of the battery is calculated using the ampere-hour integration method, and the calculation formula of the ampere-hour integration method is:

[0012] In the formula, and are the SOCs of the battery at time steps and respectively; is the current of the battery at time step , and is the cumulative charge obtained by integrating the current over time.

[0013] As an implementable preferred solution, extract the feature set that is highly correlated with life prediction under the early cycles of different batteries, including the following: Draw the capacity decay graph of the battery pack according to the calculated labeled capacity; Observe the shape of the battery capacity decay curve, analyze the internal law of battery life decay, and establish a parametric model ; Use different parametric models to fit the capacity decay curve of the battery pack to obtain the model parameter set under different parametric models; Conduct error analysis on the model parameter set and select the optimal fitting aging coefficient set.

[0014] As an implementable preferred solution, use the maximum absolute error (MaxAE), root mean square error (RMSE), and mean absolute error (MAE) to conduct error analysis on the model parameter set, and their expressions are:

[0015] Compare these error metrics under different parametric models. As an implementable preferred solution, extract the feature set that is highly correlated with life prediction under the early cycles of different batteries, including the following: Extract the feature data set based on the early cycle data of the battery pack; Use the Pearson correlation coefficient to conduct correlation analysis on the extracted features to obtain the feature data set that is highly correlated with battery capacity decay.

[0016] As an implementable preferred solution, an estimation model for battery aging parameters is established based on LSTM, which specifically includes the following content: An LSTM model is established, including three hidden layers and three fully connected layers. An activation function is set, and a test set and a training set are established for the extracted feature data and aging parameters, and the model is trained; the hyperparameters of the LSTM model are optimized based on the Bayesian optimization method.

[0017] As an implementable preferred solution, the LSTM model further includes a relu activation function and a linear activation function. The relu activation function is used for the hidden layer, and the linear activation function is used for the last fully connected layer.

[0018] As an implementable preferred solution, establishing a test set and a training set for the extracted feature data and aging parameters and training the model includes the following content: For the extracted feature data ( ) and aging parameters ( ), a test set and a training set are established, and the training set and the test set are reshaped into three-dimensional arrays with the shape of [number of samples, sequence length, number of features], and then the model is trained. The test data is input into the trained model for aging parameter estimation.

[0019] As an implementable preferred solution, for the hyperparameters in the LSTM model, based on the Bayesian optimization method, the hyperparameter range is set, the objective function is defined, and the optimal hyperparameter combination is found. Brief Description of the Drawings

[0020] Figure 1 It is the overall method flow chart of the present invention.

[0021] Figure 2 It is the logical schematic diagram of the present invention.

[0022] Figure 3 It is the logical schematic diagram of step S2 of the embodiment of the present invention.

[0023] Figure 4 It is the logical schematic diagram of step S3 of the embodiment of the present invention.

[0024] Figure 5 It is the logical schematic diagram of step S4 of the embodiment of the present invention.

[0025] Figure 6 It is the logical schematic diagram of step S5 of the embodiment of the present invention.

[0026] Figure 7 It is the schematic diagram of the electronic device of the embodiment of the present invention. Detailed Embodiment

[0027] To make the technical solutions and their advantages of this application clearer, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings. It can be understood that the specific embodiments described herein are only partial embodiments of the present invention, which are only used to explain this application and not to limit this application. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated, and they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the accompanying drawings of the following embodiments represent the same features or components, which can be applied to different embodiments.

[0028] In addition, unless otherwise defined, the technical terms or scientific terms used in the description of the present invention should have the ordinary meanings understood by those of ordinary skill in the art to which the present invention belongs.

[0029] The following further describes the present invention in detail with reference to the accompanying drawings: Reference numerals: electronic device 500, processor 501, communication interface 502, memory 503, bus 504.

[0030] Referring to Figure 1 and Figure 2 , the embodiments of the present disclosure provide a battery life prediction method for estimating decay curve parameters based on LSTM, including the following steps: Step S1, collect charge and discharge data for each cycle of different batteries, and establish a battery pack database. The charge and discharge data is usually recorded by a battery management system (BMS), including current data, voltage data, time data, temperature data, SOC data, etc. After the charge and discharge data is collected, it is sorted and stored in the database to construct a battery pack database.

[0031] Step S2, based on the charging process data, screen the charging segments, select the charging segments that meet the requirements, and calculate the labeled capacity using the ampere-hour integration and OCV-SOC correction method. Referring to Figure 3 , including: Step S21, during the data acquisition process, some data may be missing due to sensor failures or communication problems. For the missing current, voltage, temperature and other data, interpolation methods can be used for filling.

[0032] Preprocess the data, and use the moving average filtering method to smooth the data. Let the window size be , for the data sequence , the result after moving average filtering is:

[0033] where , when exceeds the data range, take the boundary value.

[0034] Not all charging segments are applicable to calculating the tag capacity. The screening criteria can be set as the charging current being stable within a certain range, the charging voltage being within the normal range, etc.

[0035] Step S22: According to the OCV - SOC table (open - circuit voltage - state of charge table), correct the SOC based on the OCV (open - circuit voltage). When the battery is in the open - circuit state (i.e., no charge - discharge operation), measure its open - circuit voltage, and then look up the corresponding SOC value from the OCV - SOC table.

[0036] Step S23: Calculate the tag capacity of the battery using the ampere - hour integration method. The calculation formula is as follows:

[0037] where, and are the SOCs of the battery pack at time steps and respectively, is the current of the battery pack at time step ; is the maximum available capacity of the battery pack. Since can be obtained by integrating the current recorded by the BMS over time, the key to obtaining the accurate battery capacity is to obtain accurate , look up the SOC through the OCV table, reduce the error in SOC estimation, and obtain the accurate SOC.

[0038] Step S3: According to the capacity decay curve trajectory, use a parametric model for fitting to obtain the set of aging coefficients for the capacity decay curve fitting of the battery pack. Refer to Figure 4 , including: Step S31: Based on the calculated tag capacity, with the number of cycles as the abscissa and the tag capacity as the ordinate, plot the capacity decay graph of the battery pack.

[0039] Step S32: Observe the shape of the battery capacity decay curve, analyze the internal law of battery life decay, and establish a parametric model. In this embodiment, the optional battery decay parametric models include but are not limited to the following:

[0040] Step S33: Use different parametric models to fit the capacity decay curve of the battery pack to obtain the set of model parameters under different parametric models.

[0041] Step S34, perform error analysis on the model parameter set using indicators such as Maximum Absolute Error (MaxAE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). Their expressions are as follows:

[0042] Compare these error indicators under different parameterized models, and select the set of aging coefficients corresponding to the model with the smallest error as the optimal fitting set of aging coefficients.

[0043] Step S4, based on the requirements of life prediction, extract a feature set that is highly correlated with life prediction during the early cycles of different batteries. Refer to Figure 5 , including: Step S41, based on the early cycle data of the battery pack, extract the feature data set. The features in the feature data set include but are not limited to: Charging capacity: The amount of electricity charged into the battery during each charging process, calculated by integrating the charging current over the charging time, with the unit of ampere-hour (Ah).

[0044] Charging time: The time taken from the start of charging to the end of charging, with the unit of minute (min) or hour (h).

[0045] Average charging temperature: The average temperature of the battery during charging, obtained by averaging the temperature data at each time point during charging, with the unit of degree Celsius (℃).

[0046] Rate of change of charging voltage: During the charging process, the change in charging voltage per unit time, reflecting the change trend of voltage during charging.

[0047] Step S42, perform correlation analysis on the extracted features to determine which features are highly correlated with battery capacity decay. Pearson correlation coefficient can be used for correlation analysis. The expression of Pearson correlation coefficient is:

[0048] According to the calculation results, select the features with high correlation to form a feature data set that is highly correlated with battery capacity decay.

[0049] Step S5: Use the extracted feature set as the input of the model and the set of aging parameters as the output of the model, and establish a battery aging parameter estimation model based on LSTM. Refer to Figure 6 , and the specific steps include: Step S51: Establish an LSTM model. In this embodiment, the LSTM estimation model consists of three hidden layers and three fully connected layers. Each layer consists of multiple units, and all units in each layer are interconnected. The number of units decreases sequentially. The number of units output by the last fully connected layer is the number of parameterized model coefficients.

[0050] The model activation functions include relu and linear activation functions. Specifically, the relu activation function is used in the hidden layer to solve the problem of gradient disappearance and increase the non-linear expression ability of the model. The linear activation function is used in the last fully connected layer for regression tasks to output the predicted aging parameter values.

[0051] For the extracted feature data ( ) and the aging parameters ( ), a test set and a training set are established, and the training set and the test set are reshaped into three-dimensional arrays with the shape of [number of samples, sequence length, number of features]. Then, the model is trained, and the test data is input into the trained model for aging parameter estimation.

[0052] Step S52: Optimize the hyperparameters of the LSTM model based on the Bayesian optimization method. For the hyperparameters such as the number of hidden layer units, the number of layers, the learning rate, the learning decay rate, and the batch size in the LSTM model, based on the Bayesian optimization method, set the hyperparameter range, define the objective function, and find the optimal hyperparameter combination to improve the model accuracy and obtain the optimal estimation result.

[0053] S6: After obtaining the battery capacity aging parameters, substitute the number of subsequent battery cycles into the parameterized model to obtain the final battery life prediction.

[0054] The embodiment of the present disclosure also provides a battery life prediction system based on LSTM for estimating decay curve parameters, which uses the above-mentioned battery life prediction method based on LSTM for estimating decay curve parameters.

[0055] Those of ordinary skill in the art can understand that all or part of the processes in the battery life prediction method for estimating the decay curve parameters based on LSTM can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of various embodiments of the battery life prediction method for estimating the decay curve parameters based on LSTM. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in this application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0056] An embodiment of this application also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps of the above-mentioned battery life prediction method for estimating decay curve parameters based on LSTM. In the embodiments of this application, the processor is the control center of the computer method, which can be the processor of a physical machine or the processor of a virtual machine.

[0057] Refer to Figure 7 , the electronic device 500 includes: at least one processor 501, at least one communication interface 502, at least one memory 503, and at least one bus 504. Among them, the bus 504 is used to realize the connection and communication between these components. The communication interface 502 is used to communicate with other node devices by signaling or data. The memory 503 stores machine-readable instructions executable by the processor 501. When the electronic device 500 runs, the processor 501 communicates with the memory 503 through the bus 504. When the machine-readable instructions are called by the processor 501, they execute the steps of the above-mentioned remaining life prediction method for estimating the battery capacity decay curve parameters based on LSTM.

[0058] The above content is only an embodiment of the present invention. Specific structures and common knowledge such as characteristics that are well-known in the art are not described in detail here. Those of ordinary skill in the art know all the common general technical knowledge in the technical field to which the invention pertains before the filing date or the priority date, can know all the prior art in this field, and have the ability to apply the conventional experimental means before this date. Those of ordinary skill in the art can, under the inspiration given in this application, combine their own abilities to improve and implement this solution. Some typical well-known structures or well-known methods should not become an obstacle for those of ordinary skill in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can still be made, and these should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicality of the patent. The protection scope required by this application should be based on the content of its claims, and the specific implementation manners and other records in the specification can be used to interpret the content of the claims.

Claims

1. A battery life prediction method for estimating the parameters of the attenuation curve based on LSTM, characterized in that: Including: Collect the charge and discharge data of different batteries for each cycle and establish a battery pack database; Based on the charging process data, screen the charging segments, and calculate the label capacity using the ampere-hour integration and OCV-SOC correction method; According to the capacity decay curve trajectory, use a parametric model for fitting to obtain the aging coefficient set of the capacity decay curve fitting of the battery pack; Extract the feature set that is highly correlated with life prediction for different batteries in the early cycles; Take the extracted feature set as the input of the model and the aging parameter set as the output of the model, and establish a battery aging parameter estimation model based on LSTM; After obtaining the battery capacity aging parameters, substitute the number of later cycles of the battery into the parametric model to obtain the battery life prediction.

2. The battery life prediction method for estimating decay curve parameters based on LSTM according to claim 1, wherein: Based on the charging process data, screen the charging segments, including the following: Use the interpolation method to fill in the missing data. Analyze the battery cycle data, preprocess the data, and use the moving average filtering method to smooth the data. Let the window size be , for the data sequence , the result after moving average filtering is: Among them, when exceeds the data range, the boundary value is taken; Screen out the charging segments that meet the requirements; based on the OCV-SOC table of the battery, correct the SOC based on the OCV to reduce the error in SOC estimation and obtain the accurate SOC.

3. The battery life prediction method for estimating attenuation curve parameters based on LSTM according to claim 2, characterized in that: Calculate the label capacity of the battery using the ampere-hour integration method. The calculation formula of the ampere-hour integration method is: Wherein, and are the SOCs of the battery at time steps and respectively; is the current of the battery at time step ; is the accumulated charge obtained by integrating the current with respect to time.

4. The battery life prediction method for estimating the decay curve parameters based on LSTM according to claim 1, characterized in that: Extract the feature set that is highly correlated with life prediction for different batteries in the early cycles, including the following: Draw the capacity decay diagram of the battery pack according to the calculated label capacity; Observe the shape of the battery capacity decay curve, analyze the internal law of battery life decay, and establish a parametric model ; Use different parametric models to fit the capacity decay curve of the battery pack to obtain the model parameter set under different parametric models; Conduct error analysis on the model parameter set and select the optimal fitting aging coefficient set.

5. The battery life prediction method for estimating decay curve parameters based on LSTM according to claim 1, characterized in that: Use the maximum absolute error (MaxAE), root mean square error (RMSE), and mean absolute error (MAE) to conduct error analysis on the model parameter set. Their expressions are: Compare these error metrics under different parametric models.

6. The battery life prediction method for estimating the attenuation curve parameters based on LSTM according to claim 1, characterized in that: Extract the feature set that is highly correlated with life prediction for different batteries in the early cycles, including the following: Based on the early cycle data of the battery pack, extract the feature data set; Use the Pearson correlation coefficient to conduct correlation analysis on the extracted features to obtain the feature data set with high correlation with battery capacity decay.

7. The battery life prediction method for estimating the attenuation curve parameters based on LSTM according to claim 1, characterized in that: Establish a battery aging parameter estimation model based on LSTM, specifically including the following: Establish an LSTM model, including three hidden layers and three fully connected layers, set the activation function, establish a test set and a training set for the extracted feature data and aging parameters, and train the model; optimize the hyperparameters of the LSTM model based on the Bayesian optimization method.

8. The battery life prediction method for estimating decay curve parameters based on LSTM according to claim 7, wherein: The LSTM model also includes a relu activation function and a linear activation function. The relu activation function is used for the hidden layer, and the linear activation function is used for the last fully connected layer.

9. The battery life prediction method for estimating decay curve parameters based on LSTM according to claim 7, wherein: Establish a test set and a training set for the extracted feature data and aging parameters, and train the model, including the following: Build a test set and a training set for the extracted feature data ( ) and the aging parameters ( ), reshape the training set and the test set into three-dimensional arrays with the shape of [number of samples, sequence length, number of features], then train the model, input the test data into the trained model, and perform aging parameter estimation.

10. The battery life prediction method for estimating decay curve parameters based on LSTM according to claim 7, characterized in that: For the hyperparameters in the LSTM model, based on the Bayesian optimization method, set the hyperparameter range, define the objective function, and find the optimal hyperparameter combination.

Images