Battery attenuation prediction method and device, equipment and medium
By predicting the future aging characteristic data of the battery and referring to the average probability density function of the battery, it accurately judges whether the battery is attenuated, and solves the problem that the battery cannot be accurately predicted in the prior art, and realizes accurate prediction of battery attenuation and applicable to multiple models. Effect.
Patent Information
- Application Number
- CN202411061883.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art cannot accurately predict battery attenuation and cannot effectively identify whether the battery will attenuate in the future.
By determining the target aging feature data of the battery to be tested and the target algorithm hyperparameters of the preset algorithm, the polynomial fitting algorithm is used to predict the future aging feature data of the battery, and combining the average probability density function of the reference battery, it is determined whether the battery is attenuated, including the already Attenuated, undecayed and dirty data.
It realizes accurate prediction of whether the battery is attenuated, has low computing complexity, is suitable for a variety of battery models, is highly generalized, and is suitable for environments with low computing power.
Smart Images

Figure CN118938017A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of battery anomaly detection, and in particular to a battery attenuation prediction method, device, equipment and medium. Background Art
[0002] Battery degradation usually refers to the phenomenon of a sudden decrease in battery capacity (often also called battery diving). This phenomenon may be caused by various factors such as battery aging, too many charge and discharge cycles, internal faults, etc. In order to identify battery degradation in related technologies, the following methods are usually used:
[0003] The first is the battery status monitoring indicator. Specifically, it continuously monitors the battery's capacity retention rate, internal resistance increase data, self-discharge rate, and one or more indicators among temperature, state of charge (SOC) and charge and discharge rate. If the indicators exceed a certain threshold, the battery is judged to have decayed. The function it realizes is more like decay identification rather than prediction.
[0004] The second is a model based on parameters affecting battery capacity. Specifically, by establishing a battery capacity loss model, an electrode capacity loss model (including a positive electrode capacity loss model and / or a negative electrode capacity loss model), and combining the battery's charging current, discharge current, temperature and other influencing factors, the changing trend of the battery capacity can be predicted. However, this method requires an in-depth understanding of the internal structure and working principle of the battery, and consideration of multiple influencing factors. Different types of batteries require separate construction of complex models, resulting in poor generalization of this method and limited practical application.
[0005] The third method is to disassemble the battery cell and observe the internal conditions. Specifically, by disassembling the battery cell and observing its internal structure, black spots on the diaphragm, electrolyte drying, etc., the possible causes of battery degradation can be inferred. This method is relatively direct, but requires professional operating skills and equipment. This method is only based on the results of degradation, and reversely infers the cause, rather than predicting the time point of possible degradation.
[0006] The above methods are unable to accurately predict battery degradation. Therefore, how to predict battery degradation by predicting the characteristic performance of the battery in the future is a technical problem that needs to be solved urgently. Summary of the invention
[0007] The present application provides a battery degradation prediction method, device, equipment and medium, which can solve the technical problem in the prior art that it is impossible to accurately predict whether the battery is degraded.
[0008] In a first aspect, an embodiment of the present application provides a battery degradation prediction method, the battery degradation prediction method comprising:
[0009] Determine target aging characteristic data of the battery to be tested and target algorithm hyperparameters of the preset algorithm;
[0010] Predicting first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter, so as to obtain first aging characteristic data of the battery to be tested;
[0011] The attenuation mark of the battery to be tested is determined according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation mark batteries in the preset multiple reference batteries, wherein the attenuation mark includes attenuated, unattenuated and dirty data.
[0012] In combination with the first aspect, in one implementation, determining target aging characteristic data of the battery to be tested includes:
[0013] Determining the cycle period and target aging characteristic data ratio of the battery to be tested;
[0014] The target aging characteristic data is determined by using the cycle period and the target aging characteristic data ratio.
[0015] In one embodiment, before determining the cycle period and target aging characteristic data ratio of the battery to be tested, the method further includes:
[0016] Respectively acquiring the second aging characteristic data of a plurality of reference batteries during the charge and discharge cycle process;
[0017] For any reference battery among the multiple reference batteries, using a parameter search algorithm, within a predefined range of the preset segmentation ratio of the aging characteristic data and a predefined range of the algorithm's preset hyperparameter values, respectively select a preset segmentation ratio of the aging characteristic data and a preset hyperparameter of the algorithm, and divide the second aging characteristic data into first historical aging characteristic data and second future aging characteristic data according to the predicted segmentation ratio of the aging characteristic data;
[0018] The third future aging characteristic data of the reference battery is predicted using the first historical aging characteristic data, the preset algorithm and the preset hyperparameters of the algorithm, and the preset segmentation ratio of the aging characteristic data and the preset hyperparameters of the algorithm that are closest to the third future aging characteristic data and the second future aging characteristic data are used as the aging characteristic data segmentation ratio and the algorithm hyperparameters of the corresponding reference battery, so as to obtain the aging characteristic data segmentation ratio and the algorithm hyperparameters of each reference battery.
[0019] In one implementation, determining the target aging characteristic data ratio includes:
[0020] Among multiple reference batteries, a reference battery whose single cell difference characteristic is most similar to the single cell difference characteristic of the battery to be tested is used as a target reference battery;
[0021] The aging characteristic data segmentation ratio and algorithm hyperparameters of the target reference battery are used as the target aging characteristic data ratio and target algorithm hyperparameters of the battery to be tested.
[0022] In one embodiment, the step of predicting first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter includes:
[0023] Using the target aging characteristic data and the target algorithm hyperparameters, the preset algorithm is trained to determine the algorithm parameters of the preset algorithm to obtain a target algorithm model of the preset algorithm;
[0024] The first future aging characteristic data is predicted using the target algorithm model to obtain the first future aging characteristic data.
[0025] In one embodiment, the determining the attenuation identification of the battery to be tested according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation identification batteries in a plurality of preset reference batteries includes:
[0026] The attenuation mark corresponding to the type with the smallest difference from the probability density function of the first aging characteristic data of the battery to be tested among the average probability density functions of the second aging characteristic data of the attenuated battery, the non-attenuated battery and the dirty data battery is used as the attenuation mark of the battery to be tested.
[0027] In one implementation, the preset algorithm is a polynomial fitting algorithm, and an algorithm hyperparameter of the polynomial fitting algorithm is the highest order of the polynomial.
[0028] In a second aspect, an embodiment of the present application provides a battery degradation prediction device, the battery degradation prediction device comprising:
[0029] A determination module, which is used to determine target aging characteristic data of the battery to be tested and target algorithm hyperparameters of a preset algorithm;
[0030] a prediction module, configured to predict first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter, so as to obtain first aging characteristic data of the battery to be tested;
[0031] An identification module is used to determine the attenuation identification of the battery to be tested according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation identification batteries in a plurality of preset reference batteries, wherein the attenuation identification includes attenuated, unattenuated and dirty data.
[0032] In a third aspect, an embodiment of the present application provides a battery degradation prediction device, which includes a processor, a memory, and a battery degradation prediction program stored in the memory and executable by the processor, wherein when the battery degradation prediction program is executed by the processor, the steps of the battery degradation prediction method as described in any one of the above items are implemented.
[0033] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a battery degradation prediction program is stored, wherein when the battery degradation prediction program is executed by a processor, the steps of the battery degradation prediction method as described in any one of the above items are implemented.
[0034] The present application provides a battery attenuation prediction method, device, equipment and medium, which determine the target aging characteristic data of a battery to be tested and the target algorithm hyperparameters of a preset algorithm; predict the first future aging characteristic data of the battery to be tested using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameters, so as to obtain the first aging characteristic data of the battery to be tested; determine the attenuation identification of the battery to be tested according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various types of attenuation identification batteries in a plurality of preset reference batteries, wherein the attenuation identification includes attenuated, unattenuated and dirty data. The method is used to accurately predict whether the battery to be tested is attenuated, and the method is essentially a statistical method with low computational complexity, so that the method can be stably executed in an environment with low computing power, and takes into account the differences between battery cells, so that the method is applicable to various types of batteries to be tested and has strong generalization. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 This is a flow chart of an embodiment of a battery degradation prediction method of the present application;
[0036] Figure 2 is a schematic diagram of third future aging characteristic data predicted according to the first aging characteristic data segmentation ratio and the highest order of the first polynomial;
[0037] Figure 3 is a schematic diagram of third future aging characteristic data predicted according to the second aging characteristic data segmentation ratio and the highest order of the first polynomial;
[0038] Figure 4 A schematic diagram of third future aging characteristic data predicted according to the first aging characteristic data segmentation ratio and the second polynomial highest order;
[0039] Figure 5 Schematic diagram of discharge capacity decay rate of undecayed battery;
[0040] Figure 6 It is a schematic diagram of the discharge capacity decay rate of a decayed battery;
[0041] Figure 7 Schematic diagram of discharge capacity attenuation rate of dirty data battery;
[0042] Figure 8 This is a schematic diagram of the functional modules of an embodiment of a battery degradation prediction device of the present application;
[0043] Fig. 9 This is a schematic diagram of the hardware structure of the battery degradation prediction device involved in the embodiment of the present application. DETAILED DESCRIPTION
[0044] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0045] First, some technical terms in the present application are explained to facilitate those skilled in the art to understand the present application.
[0046] Battery degradation prediction: Also known as battery diving prediction, battery "degradation" usually refers to a sharp decline in battery performance. For example, the capacity of lithium-ion batteries in electric vehicles (EVs) decreases rapidly after a period of use, resulting in a significant reduction in driving range. The purpose of predicting battery degradation is to understand the health of the battery in a timely manner so as to make reasonable maintenance and replacement decisions to avoid affecting vehicle performance and safety.
[0047] Parameter Search: Parameter Search is a common step in machine learning, deep learning, or any optimization problem. It aims to find a set of parameter values that optimize the model performance. Parameter search usually involves one or more hyperparameters.
[0048] Unsupervised clustering: Unsupervised clustering is a machine learning technique used to group data into multiple different clusters based on the similarity of the data. Unsupervised clustering does not require pre-labeling of category information, but classifies data based on its own characteristics. It can be used for tasks such as data exploration, dimensionality reduction, and preprocessing.
[0049] Kernel density estimation: kernel density estimation is a common nonparametric method in statistics, which is used to estimate the probability density function of a random variable. It estimates the probability density function by placing a kernel function (such as Gaussian kernel function and rectangular kernel function) on each data point and summing them. Kernel density estimation can provide an in-depth understanding of the distribution characteristics of the data, which is especially useful when dealing with unknown or complex distributions.
[0050] Distribution similarity measurement: Distribution similarity measurement is a commonly used method in machine learning, statistics, and data analysis. It is used to measure the similarity of the distribution of two or more data sets, which can help understand the relationship between data. Common measurement methods include JS divergence, KL divergence, Wasserstein distance, etc.
[0051] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.
[0052] In a first aspect, an embodiment of the present application provides a battery degradation prediction method.
[0053] In one embodiment, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the battery degradation prediction method of this application. Figure 1 As shown, the battery attenuation prediction method includes:
[0054] Step S101, determining target aging characteristic data of the battery to be tested and target algorithm hyperparameters of a preset algorithm.
[0055] Step S102: predicting first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter, so as to obtain first aging characteristic data of the battery to be tested.
[0056] Step S103, determining the attenuation mark of the battery to be tested according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation mark batteries in a plurality of preset reference batteries, wherein the attenuation mark includes attenuated, unattenuated and dirty data.
[0057] It is worth noting that the aging characteristic data can express the aging of the battery as the number of charge and discharge cycles or as time changes. The aging characteristic data can include any one or more parameters of the battery's charge or discharge capacity decay rate, internal resistance, power density, incremental capacity, differential voltage, sample entropy, and open circuit voltage.
[0058] The target aging characteristic data of the battery to be tested is the actual aging characteristic data of the battery to be tested used to predict the first future aging characteristic data. The first aging characteristic data of the battery to be tested is the target aging characteristic data of the battery to be tested and the first future aging characteristic data predicted by a preset algorithm. The first aging characteristic data of the battery to be tested expresses the complete aging characteristic data of the battery to be tested in the corresponding charge and discharge cycle. The second aging characteristic data of the reference battery is the real and complete aging characteristic data of the reference battery in the corresponding charge and discharge cycle. Then, by comparing the probability density function of the first aging characteristic data with the average probability density function of the second aging characteristic data of the attenuated battery, the non-attenuated battery and the dirty data battery, an accurate prediction of whether the battery to be tested is attenuated can be achieved. In addition, the battery attenuation prediction method provided in the present application is essentially a statistical-based method with low computational complexity, so that the method can be stably executed in an environment with low computing power.
[0059] The battery degradation prediction method of the present application is described in detail below with reference to specific embodiments.
[0060] In one embodiment, determining target aging characteristic data of the battery to be tested in step S101 includes: determining a cycle period and a ratio of target aging characteristic data of the battery to be tested; and determining the target aging characteristic data using the cycle period and the ratio of target aging characteristic data.
[0061] In a preferred embodiment, the target aging characteristic data ratio of the battery to be tested and the target algorithm hyperparameters of the preset algorithm can be determined according to the aging characteristic data of multiple reference batteries during the charge and discharge cycle. The embodiment of the present application does not limit the method for obtaining the aging characteristic data of the reference battery. For example, the reference battery can be subjected to an aging test to obtain the aging characteristic data of the reference battery, or the battery manufacturer can conduct a series of tests on the battery, extract qualified batteries from the test data as reference batteries, and extract the aging characteristic data therein, or the vehicle manufacturer can extract the aging characteristic data of the reference battery from the operation data of the actual vehicle.
[0062] Therefore, before determining the cycle period and target aging characteristic data ratio of the battery to be tested, the aging characteristic data of multiple reference batteries during the charge and discharge cycle can be acquired and processed to obtain the aging characteristic data segmentation ratio of each reference battery and the algorithm hyperparameters of the preset algorithm.
[0063] Specifically, the second aging characteristic data of multiple reference batteries during the charge and discharge cycle are respectively obtained; for any reference battery among the multiple reference batteries, a parameter search algorithm is used to select a preset segmentation ratio of the aging characteristic data and a preset hyperparameter of the algorithm within a predefined preset segmentation ratio range of the aging characteristic data and a preset hyperparameter value range of the algorithm, and the second aging characteristic data is divided into a first historical aging characteristic data and a second future aging characteristic data according to the predicted segmentation ratio of the aging characteristic data; the third future aging characteristic data of the reference battery is predicted using the first historical aging characteristic data, the preset algorithm and the preset hyperparameter of the algorithm, and the preset segmentation ratio of the aging characteristic data and the preset hyperparameter of the algorithm that are closest to the third future aging characteristic data and the second future aging characteristic data are used as the aging characteristic data segmentation ratio and the algorithm hyperparameter of the corresponding reference battery, so as to obtain the aging characteristic data segmentation ratio and the algorithm hyperparameter of each reference battery.
[0064] Among them, the preset algorithm in this embodiment can be a polynomial fitting algorithm, and the algorithm hyperparameter of the polynomial fitting algorithm is the highest order of the polynomial. In addition, the preset algorithm can also be a kernel function regression algorithm, and the algorithm hyperparameters corresponding to the kernel function regression algorithm are the kernel function type, kernel function hyperparameters (for example: the bandwidth of the Gaussian kernel, the highest order of the polynomial kernel), regularization parameters, etc., or the preset algorithm can also be a decision tree regression algorithm, and the algorithm hyperparameters of the decision tree regression algorithm are the maximum depth of the tree, the maximum number of leaf nodes of the tree, the segmentation point selection strategy, the segmentation quality evaluation standard, etc. In order to facilitate a better description of the scheme, this embodiment takes the preset algorithm as a polynomial fitting algorithm, and the algorithm hyperparameter of the polynomial fitting algorithm is the highest order of the polynomial as an example for explanation.
[0065] For example, an actual battery that has been subjected to multiple charge and discharge cycles can be selected as a reference battery. The number of charge and discharge cycles of the reference battery can be from several hundred to several thousand times. The aging characteristic data of each reference battery during the charge and discharge cycle is recorded to obtain the second aging characteristic data of each reference battery. The recorded second aging characteristic data of the reference battery can be regarded as a curve showing the change of the aging characteristic data with the number of charge and discharge cycles, for example Figure 2 As shown by the blue solid line in . The number of charge and discharge cycles increases with time, so the second aging characteristic data of the reference battery can also be regarded as a curve of the aging characteristic data changing with time, which expresses the aging of the battery over time.
[0066] Preferably, because other charging and discharging conditions may be superimposed during the charge and discharge cycle of the reference battery in the actual scenario. From the perspective of attenuation prediction, data other than the charge and discharge cycle are all abnormal data, and abnormal data may interfere with the subsequent attenuation prediction of the battery to be tested. Therefore, after obtaining the second aging characteristic data of the charge and discharge cycle of the reference battery, the acquired second aging characteristic data can be subjected to abnormal data point detection, abnormal data deletion, missing value filling, data filtering and other operations through the isolation forest algorithm, outlier detection LOF algorithm or K nearest neighbor classification KNN algorithm to eliminate interference.
[0067] Explanatory, such as Figures 2 to 4 As shown, the blue solid line in the figure is the second aging characteristic data of a reference battery during the charge and discharge cycle process. In this embodiment, the discharge capacity decay rate (or battery health state: SOH, State of Health) is used as the second aging characteristic data as an example, and the dotted line position is used as the segmentation point of the second aging characteristic data. The data before the segmentation point is the first historical aging characteristic data, and the data after the segmentation point is the second future aging characteristic data. The ratio between the number of cycles corresponding to the first historical aging characteristic data before the segmentation point and the number of cycles corresponding to the second future aging characteristic data after the segmentation point is the preset segmentation ratio of the aging characteristic data, recorded as seq.
[0068] exist Figures 2 to 4 In the figure, the red line part before the split point can be regarded as the first historical aging feature data. Based on the first historical aging feature data, the aging feature prediction is performed according to the polynomial fitting algorithm and the highest order of the polynomial, and the red line part after the split point can be obtained. The red line part after the split point is the predicted third future aging feature data. Here, the highest order of the polynomial is recorded as degree.
[0069] When seq=0.8 and degree=5, the predicted third future aging feature data and the second future aging feature data are as follows: Figure 2 As shown; when seq = 0.6, degree = 5, the predicted third future aging feature data and the second future aging feature data are as follows Figure 3 As shown; when seq = 0.6, degree = 4, the predicted third future aging feature data and the second future aging feature data are as follows Figure 4 As shown. Figure 3 and Figure 4 The predicted third future aging characteristic data deviates greatly from the second future aging characteristic data, that is, deviates greatly from the actual aging data of the reference battery.
[0070] It can be seen that for a battery that wants to predict its future aging characteristic data based on its existing aging characteristic data, its aging characteristic data preset segmentation ratio seq and the hyperparameter of the preset algorithm (polynomial highest order degree) are very important for the accuracy of the prediction. Therefore, this embodiment uses a parameter search algorithm to select the preset segmentation ratio of the aging characteristic data and the highest order of the polynomial in the predefined range of the preset segmentation ratio of the aging characteristic data and the range of the highest order of the polynomial, respectively, and then predict the third future aging characteristic data of the battery based on the first historical aging characteristic data corresponding to the selected preset segmentation ratio of the aging characteristic data, the selected highest order of the polynomial combined with the polynomial fitting algorithm, and use the preset segmentation ratio of the aging characteristic data and the highest order of the polynomial when the third future aging characteristic data is closest to the second future aging characteristic data as the aging characteristic data segmentation ratio and the highest order of the polynomial of the corresponding reference battery, thereby obtaining the aging characteristic data segmentation ratio and the highest order of the polynomial of each reference battery.
[0071] The parameter search algorithm used in this embodiment specifically includes a grid search algorithm, a random search algorithm or a Bayesian optimization algorithm.
[0072] It is worth noting that the single cell difference characteristics and cycle period of each reference battery can also be obtained. The single cell difference characteristics represent the differences in the calibration parameters between the reference batteries. The single cell difference characteristics include: rated voltage, rated capacity, calibrated capacity, cut-off voltage, chemical system and any one or more parameters in the production process. The cycle period of a battery refers to the number of charge and discharge cycles when the battery reaches a specific EOL (End of life). The typical value of the battery's EOL is 80% SOH or 70% SOH, for example Figure 2 The second aging characteristic data of the reference battery shown is aging characteristic data of 4000 charge and discharge cycles, so the cycle period of the reference battery is 4000 charge and discharge cycles.
[0073] Due to differences in battery models, cell inconsistency, cycle conditions, etc., the aging feature data segmentation ratio and algorithm hyperparameters of each reference battery are different. Therefore, after obtaining the aging feature data segmentation ratio and algorithm hyperparameters of each reference battery, the aging feature data segmentation ratio, algorithm hyperparameters, cycle period and cell difference characteristics of each reference battery can be saved accordingly, so that they can be directly extracted and used in the subsequent attenuation prediction of the battery to be tested, and they only need to be saved once. When the attenuation prediction of a new battery to be tested is required in the future, they can also be directly retrieved.
[0074] In one embodiment, after the aging characteristic data segmentation ratio, algorithm hyperparameters, cycle period and single cell difference characteristics of each reference battery are correspondingly saved, when determining the target aging characteristic data ratio, the reference battery whose single cell difference characteristics have the highest similarity with the single cell difference characteristics of the battery to be tested can be used as the target reference battery among multiple reference batteries; the aging characteristic data segmentation ratio and algorithm hyperparameters of the target reference battery are used as the target aging characteristic data ratio and target algorithm hyperparameters of the battery to be tested.
[0075] Exemplarily, there are two ways to find the target reference battery among the reference batteries: the first way is to use the K nearest neighbor (KNN) algorithm based on the monomer difference characteristics to match the one with the highest similarity between the monomer difference characteristics and the battery to be tested among all the reference batteries as the target reference battery; the second way is to cluster all the reference batteries based on the monomer difference characteristics, and obtain multiple cluster centers and the average aging feature data ratio and average algorithm hyperparameters corresponding to each cluster center. The average aging feature data ratio and average algorithm hyperparameters corresponding to the cluster center closest to the battery to be tested are used as the target aging feature data ratio and target algorithm hyperparameters of the battery to be tested.
[0076] The cycle period of the battery to be tested can be determined according to the cycle period of the target reference battery, that is, the cycle period of the target reference battery is used as the cycle period of the battery to be tested; or a preset empirical value is used as the cycle period of the battery to be tested.
[0077] Further, when the target aging characteristic data is determined by using the cycle period of the battery to be tested and the ratio of the target aging characteristic data, for example, if the cycle period of the battery to be tested is determined to be 2000 charge and discharge cycles and the ratio of the target aging characteristic data is 7:3, then the target aging characteristic data of the battery to be tested is aging characteristic data of 2000×70%=1400 charge and discharge cycles. In order to more accurately predict the attenuation of the battery to be tested, when the actual number of charge and discharge cycles of the battery to be tested is less than the number of charge and discharge cycles corresponding to the calculated target aging characteristic data, the attenuation prediction can be performed after the actual number of charge and discharge cycles of the battery to be tested reaches the number of charge and discharge cycles corresponding to the target aging characteristic data, that is, if the target aging characteristic data is aging characteristic data of 1400 charge and discharge cycles and the current battery to be tested has only aging characteristic data of 1300 charge and discharge cycles, then the attenuation prediction can be performed after the battery to be tested has only 1400 charge and discharge cycles.
[0078] Further, in step S102, predicting first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameters specifically includes the following steps:
[0079] The preset algorithm is trained using the target aging characteristic data and the target algorithm hyperparameters to determine the algorithm parameters of the preset algorithm to obtain the target algorithm model of the preset algorithm. Wherein, when the preset algorithm is a polynomial fitting algorithm, the preset algorithm parameters are the coefficients of each term in the polynomial. Then, the first future aging characteristic data is predicted using the target algorithm model to obtain the first future aging characteristic data. Specifically, the future time point corresponding to the first future aging characteristic data can be input into the target algorithm model to obtain the first future aging characteristic data. When the cycle period of the battery to be tested is 2000 times, and the target aging characteristic data is the aging characteristic data of 1400 charge and discharge cycles, the predicted first future aging characteristic data is the aging characteristic data of the charge and discharge cycles from 1401 to 2000. The target aging characteristic data plus the first future aging characteristic data form the complete aging characteristic data of the battery to be tested in the cycle period, that is, the first aging characteristic data.
[0080] As a preferred implementation, before performing step S103 or before performing attenuation prediction on the battery to be tested, the attenuation identification of each reference battery can be determined first. In this embodiment, the attenuation identification can be marked for each reference battery by observing the change of the second aging characteristic data of each reference battery during the charge and discharge cycle, and after the marking is completed, the average probability density function of the aging characteristic reference data of the attenuated battery, the non-attenuated battery and the dirty data battery is calculated respectively.
[0081] It is worth mentioning that the battery's charging or discharging capacity decay rate SOH = the battery's actual available capacity / rated capacity. The charging and discharging capacity decay rate is the parameter that can most intuitively reflect the battery's aging state. Therefore, the charging and discharging capacity decay rate can be directly selected as the aging characteristic data.
[0082] For example, the reference battery can be marked for attenuation by manual observation and judgment. Taking the discharge capacity attenuation rate as the second aging characteristic data as an example, the curve visualization result of the second aging characteristic data of the reference battery is as follows: Figures 5 to 7As shown, the horizontal axis is the number of charge and discharge cycles, and the vertical axis is the discharge capacity decay rate. It can be seen that the rate of change of the discharge capacity decay rate of the decaying battery is manifested as a significant change at a certain cycle point in the second half of the cycle period, and it will become significantly larger in the later stage of the cycle. Therefore, when manually observing the aging characteristic data, it is necessary to pay attention to whether the decay slope of the curve in the second half of the curve has increased significantly. If there is a significant increase, it means that the corresponding reference battery has decayed, and it is marked as a decayed battery (or a battery that has been water-dipped). Otherwise, it is further determined whether it is undecayed or dirty data. If the curve of the second aging characteristic data of the reference battery changes relatively slowly, the reference battery is marked as an undecayed battery (or a battery that has not been water-dipped). When the second aging characteristic reference data of some reference batteries oscillates more violently, it means that the historical charge and discharge cycle process of the reference battery is relatively complicated, and it is impossible to analyze whether the reference battery is decayed based on the second aging characteristic data, so it is marked as dirty data. Based on the above observations, the decay mark is marked for all reference batteries, and the water mark of each reference battery is one of the three categories of undecayed 0 (or undipped 0), decayed 1 (or dived 1), and dirty data 2.
[0083] When the discharge capacity decay rate curve in the second half shows fuzzy decay characteristics and it is difficult to classify it as non-decayed or decayed by manual observation (for example, the slope of the curve in the second half has changed, but the change is not too obvious), one or more parameters among internal resistance, incremental capacity, differential voltage, electrode temperature rise, sample entropy and open circuit voltage can be selected as auxiliary aging characteristic data of the reference battery, and the attenuation mark can be further determined based on the auxiliary characteristics.
[0084] Specifically, after marking the attenuation identification for each reference battery, the attenuated, unattenuated and dirty data identifications respectively contain the aging characteristic reference data of multiple reference batteries. The second aging characteristic data of all reference batteries in each type of attenuation identification are aligned according to the time point, and then the mean of the second aging characteristic data corresponding to each moment is obtained by weighted averaging, thereby obtaining the average second aging characteristic data of each type of attenuation identification, and the average probability density function of each type of attenuation identification is obtained based on the average second aging characteristic data. Exemplarily, the embodiments of the present application can use algorithms such as kernel density estimation KDE and Gaussian mixture model GMM to estimate the average probability density function.
[0085] The calculated average probability density functions of the second aging characteristic data of batteries with various attenuation identifications can be stored to form a feature library. When there is a need to predict the attenuation of the battery to be tested, there is no need to repeat the above steps. The average probability density functions of the second aging characteristic data of batteries with various attenuation identifications can be directly extracted from the feature library to perform subsequent attenuation judgments.
[0086] In step S103, the attenuation identification of the battery to be tested is determined based on the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation identification batteries in a plurality of preset reference batteries, specifically comprising the following steps: taking the attenuation identification corresponding to a type of the average probability density function of the second aging characteristic data of attenuated batteries, non-attenuated batteries and dirty data batteries that has the smallest difference with the probability density function of the first aging characteristic data of the battery to be tested as the attenuation identification of the battery to be tested.
[0087] Exemplarily, after obtaining the first aging characteristic data of the battery to be tested, the curve formed by the first aging characteristic data of the battery to be tested can also be regarded as a probability distribution corresponding to a certain random variable, and the probability density function of the first aging characteristic data of the battery to be tested can be estimated using kernel density estimation KDE. Then, the difference between the probability density function of the first aging characteristic data of the battery to be tested and the average probability density function of the second aging characteristic data of the attenuated battery, the unattenuated battery and the dirty data battery can be calculated by means of JS divergence, KL divergence or Wasserstein distance, and the corresponding attenuation mark with the smallest difference is selected as the attenuation mark of the battery to be tested. For example, if the difference between the probability density function of the first aging characteristic data of the battery to be tested and the average probability density function of the second aging characteristic data of the attenuated battery is the smallest, then the attenuation mark of the battery to be tested is attenuated.
[0088] The battery attenuation prediction method provided in this embodiment predicts the future aging characteristic data (i.e., the first future aging data) of the battery to be tested that has not yet been generated by the real aging characteristic data (i.e., the target aging characteristic data) of the battery to be tested during the charge and discharge cycle, thereby predicting the complete aging characteristic data (i.e., the first aging characteristic data) of the battery to be tested during the cycle period, and predicting whether the battery to be tested is attenuated based on the complete aging characteristic data of the battery to be tested. The beneficial effects are: on the one hand, by predicting the future aging characteristic data of the battery to be tested, the dimension of the aging characteristic data during the attenuation prediction of the battery to be tested is expanded, and the accuracy of attenuation determination can be significantly improved; on the other hand, the future aging characteristic data of the battery to be tested within a longer cycle is predicted, and the predicted aging characteristic data can be used to realize the battery performance change analysis other than attenuation determination. A large number of actual batteries are used as reference batteries, and a feature library is constructed based on the charge and discharge cycle data of the reference batteries, which can take into account the differences in battery monomers, models, and cycle conditions, and has strong generalization and versatility. In addition, the method of the present application is essentially a statistical method, which has low computational complexity, is more friendly to embedded execution environments with low computing power, and can be stably executed in environments with low computing power.
[0089] In a second aspect, an embodiment of the present application also provides a battery degradation prediction device.
[0090] In one embodiment, referring to Figure 8 , Figure 8 This is a functional module diagram of an embodiment of a battery degradation prediction device of the present application. Figure 8 As shown, the battery degradation prediction device includes:
[0091] A determination module, which is used to determine target aging characteristic data of the battery to be tested and target algorithm hyperparameters of a preset algorithm;
[0092] a prediction module, configured to predict first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter, so as to obtain first aging characteristic data of the battery to be tested;
[0093] An identification module is used to determine the attenuation identification of the battery to be tested according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation identification batteries in a plurality of preset reference batteries, wherein the attenuation identification includes attenuated, unattenuated and dirty data.
[0094] Furthermore, in one embodiment, the determining module is further configured to:
[0095] Determining the cycle period and target aging characteristic data ratio of the battery to be tested;
[0096] The target aging characteristic data is determined by using the cycle period and the target aging characteristic data ratio.
[0097] Furthermore, in one embodiment, the device further includes a parameter search module, and the parameter search module is used to:
[0098] Respectively acquiring the second aging characteristic data of a plurality of reference batteries during the charge and discharge cycle process;
[0099] For any reference battery among the multiple reference batteries, using a parameter search algorithm, within a predefined range of the preset segmentation ratio of the aging characteristic data and a predefined range of the algorithm's preset hyperparameter values, respectively select a preset segmentation ratio of the aging characteristic data and a preset hyperparameter of the algorithm, and divide the second aging characteristic data into first historical aging characteristic data and second future aging characteristic data according to the predicted segmentation ratio of the aging characteristic data;
[0100] The third future aging characteristic data of the reference battery is predicted using the first historical aging characteristic data, the preset algorithm and the preset hyperparameters of the algorithm, and the preset segmentation ratio of the aging characteristic data and the preset hyperparameters of the algorithm that are closest to the third future aging characteristic data and the second future aging characteristic data are used as the aging characteristic data segmentation ratio and the algorithm hyperparameters of the corresponding reference battery, so as to obtain the aging characteristic data segmentation ratio and the algorithm hyperparameters of each reference battery.
[0101] Furthermore, in one embodiment, the determining module is further configured to:
[0102] Among multiple reference batteries, a reference battery whose single cell difference characteristic is most similar to the single cell difference characteristic of the battery to be tested is used as a target reference battery;
[0103] The aging characteristic data segmentation ratio and algorithm hyperparameters of the target reference battery are used as the target aging characteristic data ratio and target algorithm hyperparameters of the battery to be tested.
[0104] Furthermore, in one embodiment, the prediction module is also used for:
[0105] Using the target aging characteristic data and the target algorithm hyperparameters, the preset algorithm is trained to determine the algorithm parameters of the preset algorithm to obtain a target algorithm model of the preset algorithm;
[0106] The first future aging characteristic data is predicted using the target algorithm model and the target aging characteristic data to obtain the first future aging characteristic data.
[0107] Furthermore, in one embodiment, the identification module is also used for:
[0108] The attenuation mark corresponding to the type with the smallest difference from the probability density function of the first aging characteristic data of the battery to be tested among the average probability density functions of the second aging characteristic data of the attenuated battery, the non-attenuated battery and the dirty data battery is used as the attenuation mark of the battery to be tested.
[0109] Furthermore, in one embodiment, the preset algorithm is a polynomial fitting algorithm, and the algorithm hyperparameter of the polynomial fitting algorithm is the highest order of the polynomial.
[0110] Among them, the functional implementation of each module in the above-mentioned battery degradation prediction device corresponds to the various steps in the above-mentioned battery degradation prediction method embodiment, and its functions and implementation processes will not be repeated here one by one.
[0111] In a third aspect, an embodiment of the present application provides a battery degradation prediction device, which may be a personal computer (PC), a laptop computer, a server, or other device with data processing capabilities.
[0112] Reference Fig. 9 , Fig. 9 Schematic diagram of the hardware structure of the battery degradation prediction device involved in the embodiment of the present application. In the embodiment of the present application, the battery degradation prediction device may include a processor, a memory, a communication interface and a communication bus.
[0113] The communication bus may be of any type and is used to interconnect the processor, the memory, and the communication interface.
[0114] The communication interface includes input / output (I / O) interface, physical interface and logical interface, etc., which are used to realize the interconnection of devices inside the battery decay prediction device, and the interface used to realize the interconnection of the battery decay prediction device with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, a fiber optic interface, an ATM interface, etc.; the user device can be a display, a keyboard, etc.
[0115] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.
[0116] The processor may be a general-purpose processor, which may call a battery decay prediction program stored in a memory and execute the battery decay prediction method provided in the embodiment of the present application. For example, the general-purpose processor may be a central processing unit (CPU). The method executed when the battery decay prediction program is called may refer to the various embodiments of the battery decay prediction method of the present application, which will not be described in detail here.
[0117] Those skilled in the art will understand that Fig. 9 The hardware structure shown in the figure does not constitute a limitation on the present application, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.
[0118] In a fourth aspect, an embodiment of the present application also provides a computer-readable storage medium.
[0119] A battery degradation prediction program is stored on a computer-readable storage medium of the present application, wherein when the battery degradation prediction program is executed by a processor, the steps of the battery degradation prediction method as described above are implemented.
[0120] Among them, the method implemented when the battery degradation prediction program is executed can refer to the various embodiments of the battery degradation prediction method of the present application, and will not be repeated here.
[0121] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.
[0122] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit "first", "second" and "third" to different types.
[0123] In the description of the embodiments of the present application, "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a specific way.
[0124] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; the “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.
[0125] In some processes described in the embodiments of the present application, multiple operations or steps that appear in a specific order are included, but it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or in parallel, and the sequence number of the operation is only used to distinguish the different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.
[0126] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD) as described above, and includes a number of instructions for a terminal device to execute the methods described in each embodiment of the present application.
[0127] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.
Claims
1. A battery degradation prediction method, characterized in that: The battery attenuation prediction method comprises: Determine target aging characteristic data of the battery to be tested and target algorithm hyperparameters of the preset algorithm; Predicting first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter, so as to obtain first aging characteristic data of the battery to be tested; The attenuation mark of the battery to be tested is determined according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation mark batteries in the preset multiple reference batteries, wherein the attenuation mark includes attenuated, unattenuated and dirty data.
2. The battery degradation prediction method according to claim 1, characterized in that: Determine the target aging characteristic data of the battery to be tested, including: Determining the cycle period and target aging characteristic data ratio of the battery to be tested; The target aging characteristic data is determined by using the cycle period and the target aging characteristic data ratio.
3. The battery degradation prediction method according to claim 2, characterized in that: Before determining the cycle period and target aging characteristic data ratio of the battery to be tested, the method further includes: Respectively acquiring the second aging characteristic data of a plurality of reference batteries during the charge and discharge cycle process; For any reference battery among the multiple reference batteries, using a parameter search algorithm, within a predefined range of a preset segmentation ratio of the aging characteristic data and a predefined range of a preset hyperparameter value of the algorithm, respectively select a preset segmentation ratio of the aging characteristic data and a preset hyperparameter of the algorithm, and divide the second aging characteristic data into first historical aging characteristic data and second future aging characteristic data according to the preset segmentation ratio of the aging characteristic data; The third future aging characteristic data of the reference battery is predicted using the first historical aging characteristic data, the preset algorithm and the preset hyperparameters of the algorithm, and the preset segmentation ratio of the aging characteristic data and the preset hyperparameters of the algorithm that are closest to the third future aging characteristic data and the second future aging characteristic data are used as the aging characteristic data segmentation ratio and the algorithm hyperparameters of the corresponding reference battery, so as to obtain the aging characteristic data segmentation ratio and the algorithm hyperparameters of each reference battery.
4. The battery attenuation prediction method according to any one of claims 2 or 3, characterized in that: Determining the target aging characteristic data ratio includes: Among multiple reference batteries, a reference battery whose single cell difference characteristic is most similar to the single cell difference characteristic of the battery to be tested is used as a target reference battery; The aging characteristic data segmentation ratio and algorithm hyperparameters of the target reference battery are used as the target aging characteristic data ratio and target algorithm hyperparameters of the battery to be tested.
5. The battery degradation prediction method according to claim 1, characterized in that: The step of predicting first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter includes: Using the target aging characteristic data and the target algorithm hyperparameters, the preset algorithm is trained to determine the algorithm parameters of the preset algorithm to obtain a target algorithm model of the preset algorithm; The first future aging characteristic data is predicted using the target algorithm model to obtain the first future aging characteristic data.
6. The battery degradation prediction method according to claim 1, characterized in that: The step of determining the attenuation identification of the battery to be tested according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of batteries with various attenuation identifications in a plurality of preset reference batteries comprises: The attenuation mark corresponding to the type with the smallest difference from the probability density function of the first aging characteristic data of the battery to be tested among the average probability density functions of the second aging characteristic data of the attenuated battery, the non-attenuated battery and the dirty data battery is used as the attenuation mark of the battery to be tested.
7. The battery degradation prediction method according to claim 1, characterized in that: The preset algorithm is a polynomial fitting algorithm, and the algorithm hyperparameter of the polynomial fitting algorithm is the highest order of the polynomial.
8. A battery degradation prediction device, characterized in that: The battery degradation prediction device comprises: A determination module, which is used to determine target aging characteristic data of the battery to be tested and target algorithm hyperparameters of a preset algorithm; a prediction module, configured to predict first future aging characteristic data of the battery to be tested by using the preset algorithm according to the target aging characteristic data of the battery to be tested and the target algorithm hyperparameter, so as to obtain first aging characteristic data of the battery to be tested; An identification module is used to determine the attenuation identification of the battery to be tested according to the probability density function of the first aging characteristic data and the average probability density function of the second aging characteristic data of various attenuation identification batteries in a plurality of preset reference batteries, wherein the attenuation identification includes attenuated, unattenuated and dirty data.
9. A battery degradation prediction device, characterized in that: The battery degradation prediction device includes a processor, a memory, and a battery degradation prediction program stored in the memory and executable by the processor, wherein when the battery degradation prediction program is executed by the processor, the steps of the battery degradation prediction method as described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a battery degradation prediction program, wherein when the battery degradation prediction program is executed by a processor, the steps of the battery degradation prediction method according to any one of claims 1 to 7 are implemented.
