Data processing method and device for battery attenuation
By constructing a battery attenuation prediction model based on the Wiener process based on the nonlinear integral term, the problem of scarcity of battery sample data is solved, and the high accuracy of battery life prediction and the richness of data sets are achieved.
Patent Information
- Application Number
- CN202411429792.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-14
- Publication Date
- 2025-02-28
AI Technical Summary
The prior art is difficult to effectively deal with the problem of scarcity of battery sample data in battery life prediction, resulting in inefficient data set amplification method.
The battery attenuation prediction model based on the Wiener process based on the nonlinear integral term is used to fit the current attenuation data of the battery, estimate the model parameters, reconstruct the data set, and evaluate and update it through preset evaluation indicators.
It improves the accuracy of battery life prediction, increases the richness of the data set, can accurately capture the nonlinear characteristics of battery data, and maintains the time series characteristics and physical consistency of the data.
Smart Images
Figure CN119535216A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of battery management technology, and in particular to a method and device for processing battery attenuation data. Background Art
[0002] Lithium-ion batteries are widely used in various energy storage systems, and the safety and efficiency of their operation depend on accurate prediction of their remaining useful life. RUL prediction models are usually used for remaining useful life prediction. However, in many cases, data set augmentation is required to improve model performance by increasing the diversity and number of data samples.
[0003] The existing data set augmentation methods mainly include: data augmentation based on interpolation, collecting more data, and adjusting model hyperparameters to optimize model structure. However, these methods have the following shortcomings: Data augmentation methods based on interpolation usually assume that the data is linear, which is not applicable to the nonlinear characteristics of batteries. Collecting more data will cost more time and testing costs, and is not applicable to situations where the amount of battery data is extremely limited. Adjusting model hyperparameters is not applicable to situations where the amount of battery data is small, and the implementation process is relatively complicated. Summary of the invention
[0004] The embodiments of the present application provide a battery attenuation data processing method and device, which are suitable for battery life prediction when battery sample data is scarce, increase the richness of the data set, and improve the accuracy of battery life prediction.
[0005] In order to solve the above technical problems, the embodiments of the present application disclose the following technical solutions:
[0006] In a first aspect, a method for processing battery attenuation data is provided, comprising: inputting current attenuation data of the battery into a pre-constructed battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, thereby obtaining a parameter estimation model of the battery attenuation prediction model;
[0007] estimating each parameter of the battery degradation prediction model according to the parameter estimation model to obtain estimated values of each parameter of the battery degradation prediction model;
[0008] Reconstructing the current attenuation data of the battery according to the estimated values of the parameters of the battery attenuation prediction model and the battery attenuation prediction model to obtain a reconstructed data set of the current attenuation data of the battery;
[0009] The reconstructed data set is evaluated and updated according to preset evaluation indicators.
[0010] In combination with the first aspect, the nonlinear integral term in the battery degradation prediction model based on the Wiener process with nonlinear integral term includes one or more of an exponential function, a logarithmic function and a power function.
[0011] In combination with the first aspect, the parameter estimation model of the battery degradation prediction model is a log-likelihood equation of the battery degradation prediction model;
[0012] The estimating each parameter of the battery degradation prediction model according to the parameter estimation model includes:
[0013] The log-likelihood equation of the battery degradation prediction model is simplified and its partial derivatives are solved in turn to obtain estimated values of various parameters of the battery degradation prediction model.
[0014] In combination with the first aspect, the parameters of the battery degradation prediction model include a first part of parameters and a second part of parameters;
[0015] The log-likelihood equation of the battery degradation prediction model is simplified and partial derivatives are solved in sequence to obtain estimated values of various parameters of the battery degradation prediction model, including:
[0016] Simplifying the log-likelihood equation of the battery degradation prediction model to obtain a simplified log-likelihood equation;
[0017] Solving the partial derivative of the simplified log-likelihood equation to obtain estimated values of the first part of parameters of the battery degradation prediction model;
[0018] Substitute the estimated values of the first part of the parameters of the battery degradation prediction model into the simplified log-likelihood equation to obtain the estimated values of the second part of the parameters of the battery degradation prediction model.
[0019] In combination with the first aspect, reconstructing the current attenuation data of the battery according to the estimated values of the parameters of the battery attenuation prediction model and the battery attenuation prediction model includes:
[0020] Taking the estimated values of the parameters of the battery degradation prediction model as the prior distribution, and calculating the posterior distribution of the parameters of the battery degradation prediction model based on the Bayesian theory;
[0021] The update parameters of each parameter of the battery attenuation prediction model are determined according to the posterior distribution of each parameter, and the update parameters of each parameter are input into the battery attenuation prediction model to reconstruct the current attenuation data of the battery.
[0022] In combination with the first aspect, the preset evaluation index includes at least one or more of a correlation coefficient evaluation index, a decay rate evaluation index and a Euclidean distance evaluation index.
[0023] In combination with the first aspect, the preset evaluation index includes a correlation coefficient evaluation index, a decay rate evaluation index and a Euclidean distance evaluation index;
[0024] The step of evaluating and updating the reconstructed data set according to a preset evaluation index includes:
[0025] The reconstructed data set is evaluated and updated in turn according to the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index.
[0026] In combination with the first aspect, the reconstructed data set includes a plurality of groups of reconstructed curves;
[0027] The step of evaluating and updating the reconstructed data set in accordance with the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index in sequence includes:
[0028] Acquire the real attenuation data of the battery, calculate the correlation coefficient between each group of reconstructed curves and the real attenuation data, screen out the reconstructed curves with a correlation coefficient greater than a preset correlation coefficient, and record them as the first data set;
[0029] Calculating the true attenuation rate of each group of reconstructed curves in the first data set, screening out the reconstructed curves whose true attenuation rates meet the preset attenuation rate conditions, and recording them as the second data set;
[0030] Calculating the Euclidean distance of each group of reconstructed curves in the second data set, selecting the reconstructed curve with the smallest Euclidean distance, and recording it as the third data set;
[0031] And the various parameters of the battery degradation prediction model are updated through the various parameters of the reconstructed curve corresponding to the third data set.
[0032] In combination with the first aspect, after evaluating and updating the reconstructed data set according to the preset evaluation index, the method further includes:
[0033] The reconstructed data set evaluated according to the preset evaluation index and updated is returned and repeatedly input into the pre-built battery degradation prediction model based on the nonlinear integral term Wiener process for fitting until the reconstructed data set becomes stable.
[0034] In a second aspect, a battery attenuation data processing device is provided, comprising: a fitting module, configured to input the current attenuation data of the battery into a pre-constructed battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, and obtain a parameter estimation model of the battery attenuation prediction model;
[0035] A parameter estimation module, used to estimate each parameter of the battery degradation prediction model according to the parameter estimation model to obtain an estimated value of each parameter of the battery degradation prediction model;
[0036] A reconstruction module, used to reconstruct the current attenuation data of the battery according to the estimated values of each parameter of the battery attenuation prediction model and the battery attenuation prediction model to obtain a reconstructed data set of the current attenuation data of the battery;
[0037] The evaluation and update module is used to evaluate and update the reconstructed data set according to preset evaluation indicators.
[0038] One of the above technical solutions has the following advantages or beneficial effects:
[0039] Compared with the prior art, a data processing method for battery attenuation of the present application includes: inputting the current attenuation data of the battery into a pre-constructed battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, and obtaining a parameter estimation model of the battery attenuation prediction model; estimating each parameter of the battery attenuation prediction model according to the parameter estimation model, and obtaining the estimated values of each parameter of the battery attenuation prediction model; reconstructing the current attenuation data of the battery according to the estimated values of each parameter of the battery attenuation prediction model and the battery attenuation prediction model, and obtaining a reconstructed data set of the current attenuation data of the battery; and evaluating and updating the reconstructed data set according to a preset evaluation index. The data processing method for battery attenuation provided by the present application can reconstruct the attenuation data of the battery by constructing a battery attenuation prediction model based on a nonlinear integral term Wiener process, increase the richness of the battery attenuation data set, accurately capture the nonlinear characteristics of the battery data, and maintain the time series characteristics and physical consistency of the battery attenuation data, and provide more real and diverse training data for battery life prediction, which is suitable for situations where battery data samples are scarce. In addition, the reconstructed data set is evaluated and updated by a preset evaluation index, further improving the accuracy of battery attenuation data reconstruction and improving the accuracy of battery life prediction.
[0040] The battery attenuation data processing device of the present application can reconstruct the battery attenuation data, increase the richness of the battery attenuation data set, accurately capture the nonlinear characteristics of the battery data, and maintain the time series characteristics and physical consistency of the battery attenuation data, providing more realistic and diverse training data for battery life prediction, which is suitable for situations where battery data samples are scarce. In addition, the reconstructed data set is evaluated and updated through preset evaluation indicators, further improving the accuracy of battery attenuation data reconstruction and the accuracy of battery life prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.
[0042] Figure 1 is a flow chart of a battery attenuation data processing method provided in an embodiment of the present application;
[0043] Figure 2 is a flow chart of another battery attenuation data processing method provided in an embodiment of the present application;
[0044] Figure 3 is a flow chart of another battery attenuation data processing method provided in an embodiment of the present application;
[0045] Figure 4 It is a principle structural block diagram of a battery attenuation data processing device provided in an embodiment of the present application.
[0046] Reference numerals:
[0047] 10-fitting module; 20-parameter estimation module; 30-reconstruction module; 40-evaluation and update module; 100-battery attenuation data processing device. DETAILED DESCRIPTION
[0048] The technical solutions in the embodiments of the present application will be described clearly and completely 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, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.
[0049] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first" and "second" are used only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more features. In the description of the present application, "multiple" means two or more, and at least one means one, two or more, unless otherwise clearly and specifically defined.
[0050] Figure 1 This is a flowchart of a battery attenuation data processing method provided in an embodiment of the present application. This method is suitable for processing battery attenuation data in battery life prediction, especially in the scenario of battery remaining useful life (RUL) prediction where battery sample data is scarce, to improve the diversity, richness and accuracy of battery data. This method can be executed by a battery attenuation data processing device, which can be implemented by software and / or hardware, and the device can be configured in the management system or processor of the battery management platform. Please refer to Figure 1 The battery attenuation data processing method comprises the following steps:
[0051] Step 110: input the current attenuation data of the battery into a pre-built battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, and obtain a parameter estimation model of the battery attenuation prediction model.
[0052] The current attenuation data of the battery is the data on battery attenuation obtained during the operation of the battery, for example, battery capacity-cycle number data. The current attenuation data of the battery is the original data of the battery attenuation prediction model.
[0053] Among them, the mathematical formula of the Weiner Process model (WP) is as follows:
[0054]
[0055] Among them, X(t) is the battery decay rate at time t, X(0) is the initial battery decay rate, and the integral term is a nonlinear function used to fit the battery attenuation trend, λ is the variable parameter of the nonlinear function, B(t) is used to describe the random fluctuations on the battery capacity attenuation curve, σ B Represents the diffusion coefficient of the battery and is also used to describe the degree of random fluctuations.
[0056] Exemplarily, in the technical solution of this embodiment, a battery attenuation prediction model is constructed based on a Wiener process with a nonlinear integral term, and the current attenuation data of the battery is input into the prediction model, and fitting is performed to obtain a parameter estimation model of the battery attenuation prediction model. This can accurately capture the nonlinear characteristics of the battery attenuation data, and can maintain the time series characteristics and physical consistency of the battery attenuation data, thereby improving the accuracy of subsequent prediction model parameter estimation, as well as the stability, continuity and accuracy of the reconstruction of the battery attenuation data.
[0057] In some embodiments, the number of nonlinear integral terms in the battery attenuation prediction model based on the Wiener process of nonlinear integral terms may include a single term, two terms or even more terms, which can be set according to actual conditions and is not specifically limited here.
[0058] In some embodiments, the nonlinear integral term in the battery degradation prediction model based on the Wiener process of the nonlinear integral term includes one or more of an exponential function, a logarithmic function, and a power function.
[0059] The types and quantities of the nonlinear integral terms included can be set according to actual conditions and are not specifically limited here.
[0060] Exemplarily, in the scheme of the embodiment of the present application, the battery degradation prediction model based on the Wiener process of nonlinear integral term includes two exponential functions, that is, the battery degradation prediction model of the double exponential Wiener model is taken as an example for explanation (the same below, no further details are given).
[0061] Among them, the calculation formula of the battery degradation prediction model of the double exponential Wiener model is:
[0062] X(t)=X(0)+a(exp(bt)-1)+c(exp(dt)-1)+σ B B(t)
[0063] Among them, a and b are the parameters of the first exponential term, and c and d are the parameters of the second exponential term.
[0064] Specifically, the current attenuation data of the battery is fitted by the calculation formula of the battery attenuation prediction model of the above-mentioned double exponential Wiener model, in which the battery attenuation trend is fitted by the double exponential term. Among them, the prior distribution of each parameter is derived from the real data, and the posterior distribution probability of each parameter after the parameter update iteration can be calculated by the calculation formula of the battery attenuation prediction model of the double exponential Wiener model. Therefore, by fitting the current attenuation data of the battery with the battery attenuation prediction model of the double exponential Wiener model, the nonlinear characteristics of the battery attenuation data can be accurately captured, and the time series characteristics and physical consistency of the battery attenuation data can be maintained, so that the parameter estimation model of the fitted battery attenuation prediction model has high prediction accuracy, thereby improving the accuracy of the subsequent prediction model parameter estimation and the accuracy of the reconstruction of the battery attenuation data. In addition, the battery attenuation prediction model of the double exponential Wiener model is simple and easy to implement, with adjustable parameters and strong applicability, especially suitable for the case of scarce samples in battery life prediction.
[0065] Before the current attenuation data of the battery is input into the battery attenuation prediction model based on the nonlinear integral term Wiener process for fitting, the current attenuation data of the battery (i.e., the original data) is smoothed using a filtering method. The filtering method may be a Savitzky–Golay filtering method.
[0066] Step 120: Estimate each parameter of the battery degradation prediction model according to the parameter estimation model to obtain estimated values of each parameter of the battery degradation prediction model.
[0067] Specifically, first, the current attenuation data of the battery is input into the battery attenuation prediction model based on the Wiener process of the nonlinear integral term for fitting, and the parameter estimation model of the battery attenuation prediction model is obtained. Then, the various parameters of the battery attenuation prediction model are estimated according to the parameter estimation model to obtain the estimated values of the various parameters of the battery attenuation prediction model, so that the battery attenuation data can be reconstructed according to the estimated values of the various parameters of the prediction model and the prediction model. Therefore, by fitting the current attenuation data of the battery using the battery attenuation prediction model based on the Wiener process of the nonlinear integral term, the nonlinear characteristics of the battery attenuation data can be accurately captured, and the time series characteristics and physical consistency of the battery attenuation data can be maintained, so that the prediction accuracy of the parameter estimation model of the battery attenuation prediction model obtained by fitting is high, thereby improving the accuracy of the parameter estimation of the battery attenuation prediction model, and further improving the accuracy and reliability of the subsequent reconstruction of the battery attenuation data.
[0068] Step 130: reconstruct the current attenuation data of the battery according to the estimated values of the parameters of the battery attenuation prediction model and the battery attenuation prediction model to obtain a reconstructed data set of the current attenuation data of the battery.
[0069] Specifically, first, the current attenuation data of the battery is input into the battery attenuation prediction model based on the Wiener process of the nonlinear integral term for fitting, and the parameter estimation model of the battery attenuation prediction model is obtained. Then, the various parameters of the battery attenuation prediction model are estimated according to the parameter estimation model to obtain the estimated values of the various parameters of the battery attenuation prediction model. Secondly, the current attenuation data of the battery is reconstructed according to the estimated values of the various parameters of the battery attenuation prediction model and the battery attenuation prediction model, so as to achieve the reconstruction of the battery attenuation data. Therefore, by fitting the current attenuation data of the battery using the battery attenuation prediction model based on the Wiener process of the nonlinear integral term, the nonlinear characteristics of the battery attenuation data can be accurately captured, and the time series characteristics and physical consistency of the battery attenuation data can be maintained, so that the prediction accuracy of the parameter estimation model of the battery attenuation prediction model obtained by fitting is high, thereby improving the accuracy of the parameter estimation of the battery attenuation prediction model, and further improving the accuracy and reliability of the reconstruction of the battery attenuation data.
[0070] Step 140: Evaluate and update the reconstructed data set according to preset evaluation indicators.
[0071] Specifically, after reconstructing the current attenuation data of the battery to obtain a reconstructed data set of the battery attenuation data, in order to improve the accuracy of the battery attenuation data reconstruction, the reconstructed data set is evaluated and updated according to preset evaluation indicators, thereby further improving the accuracy and reliability of the battery attenuation data reconstruction, and further improving the accuracy of the battery life prediction.
[0072] In the technical solution of this embodiment, the working principle of the battery attenuation data processing method is as follows: Figure 1First, the current attenuation data of the battery is input into the pre-built battery attenuation prediction model based on the nonlinear integral term of the Wiener process for fitting, and the parameter estimation model of the battery attenuation prediction model is obtained. Then, the various parameters of the battery attenuation prediction model are estimated according to the parameter estimation model to obtain the estimated values of the various parameters of the battery attenuation prediction model. Secondly, the current attenuation data of the battery is reconstructed according to the estimated values of the various parameters of the battery attenuation prediction model and the battery attenuation prediction model to obtain a reconstructed data set of the current attenuation data of the battery. Finally, the reconstructed data set is evaluated and updated according to the preset evaluation indicators. It can be seen that by constructing a battery attenuation prediction model using the Wiener process of nonlinear integral terms, and inputting the current attenuation data of the battery into the prediction model for fitting, the estimated values of each parameter of the prediction model can be obtained, and the attenuation data of the battery can be reconstructed according to the estimated values of each parameter and the battery attenuation prediction model based on the Wiener process of nonlinear integral terms, which can accurately capture the nonlinear characteristics of the battery attenuation data, and maintain the time series characteristics and physical consistency of the battery attenuation data during the data reconstruction process, thereby increasing the richness, stability, continuity and accuracy of the battery attenuation data reconstruction, and improving the prediction accuracy of the battery attenuation data in the case of scarce data samples, providing more real and diverse training data for battery attenuation prediction. Moreover, the method is simple and easy to implement, with adjustable parameters and strong applicability, especially suitable for the case of scarce samples in battery life prediction. In addition, after reconstructing the current attenuation data of the battery, the reconstructed data set is evaluated and updated by preset evaluation indicators, thereby further improving the accuracy and reliability of the battery attenuation data reconstruction, and thus improving the accuracy of the battery life prediction.
[0073] Figure 2 is a flow chart of another method for processing battery attenuation data provided in an embodiment of the present application. Based on the above embodiment, optionally, refer to Figure 2 , the method comprises the following steps:
[0074] Step 210: input the current attenuation data of the battery into a pre-built battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, and obtain a parameter estimation model of the battery attenuation prediction model.
[0075] Step 220: Simplify the log-likelihood equation of the battery degradation prediction model and solve the partial derivatives in turn to obtain the estimated values of various parameters of the battery degradation prediction model.
[0076] Among them, the parameter estimation model of the battery attenuation prediction model is the log-likelihood equation of the battery attenuation prediction model. Among them, the log-likelihood equation of the battery attenuation prediction model is used to evaluate the degree of fit of the observed data under a given model, such as the battery attenuation prediction model of a given double exponential Wiener process. Among them, for battery attenuation data containing multiple time series, the log-likelihood equation of the battery attenuation prediction model is as follows:
[0077]
[0078] The log-likelihood equation of the battery degradation prediction model includes three terms. The first term is the constant term of the log probability density function of the normal distribution, which reflects the impact of the number of observed data on the likelihood function. The second term represents the logarithmic determinant of the covariance matrix, which measures the impact of the covariance structure of the data on the likelihood function. The third term is the main part of the log probability density function of the normal distribution, which measures the difference between the observed data and the prediction of the battery degradation prediction model. Specifically, ΔX is the observed vector data, aΔT b and cT d It is the product of the battery degradation prediction model parameters and the time step, which represents the capacity change predicted by the battery degradation prediction model. is a covariance matrix, which is used to measure the correlation between data points. For example, through a set of battery current attenuation data, for example, a set of battery historical capacity-cycle number data, multiple capacity differences between two points can be obtained, and multiple predicted capacity differences between two points can also be obtained, thereby obtaining multiple sets of likelihood equations. Then, by taking the minimum value of the sum of these multiple sets of equations, the initial estimated values of each parameter of the battery attenuation prediction model can be obtained.
[0079] Where, ΔT b =[exp(bt2)-exp(bt1), exp(bt3)-exp(bt2),...exp(bt m )-exp(bt m-1 )], ΔT d =[exp(dt2)-exp(dt1), exp(dt3)-exp(dt2),...exp(dt m )-exp(dt m-1 )], ΔX=[x2-x1, x2-x1,...x m -x m-1 ].in,
[0080] As a specific implementation, optionally, the various parameters of the battery attenuation prediction model include first-part parameters and second-part parameters; the log-likelihood equation of the battery attenuation prediction model is simplified and partial derivatives are solved in turn to obtain estimated values of the various parameters of the battery attenuation prediction model, including: simplifying the log-likelihood equation of the battery attenuation prediction model to obtain a simplified log-likelihood equation; solving partial derivatives of the simplified log-likelihood equation to obtain estimated values of the first-part parameters of the battery attenuation prediction model; substituting the estimated values of the first-part parameters of the battery attenuation prediction model into the simplified log-likelihood equation to obtain estimated values of the second-part parameters of the battery attenuation prediction model.
[0081] Exemplarily, the first part of parameters includes parameters a, c and σ 2 B , the second part of parameters includes b and d.
[0082] Among them, for the log-likelihood equation of the battery degradation prediction model, Since I is an m-1-order unit matrix, the log-likelihood equation of the battery degradation prediction model is simplified, and the simplified log-likelihood equation can be obtained as follows:
[0083] Then we get:
[0084]
[0085] Then, solving the partial derivatives of the simplified log-likelihood equation, we can get the first part of the parameters a, c and σ 2 B The maximum likelihood estimate of . Among them, the corresponding partial derivatives of the simplified log-likelihood equation are solved as follows:
[0086]
[0087] From this, we can get the maximum likelihood estimate:
[0088]
[0089] The first part of the parameters a, c and σ 2 B Substituting the maximum likelihood estimate of into the simplified log-likelihood equation, we get the following equation:
[0090]
[0091] Therefore, the estimated values of the parameters b and d according to the second part of the above equation can be obtained as follows:
[0092]
[0093] in, and are the estimated values of parameters b and d, respectively.
[0094] It can be understood that the present application estimates the various parameters of the battery degradation prediction model of the double exponential Wiener process through the log-likelihood equation, and finds the optimal parameter value by maximizing the log-likelihood function, thereby improving the fitting effect of the battery degradation prediction model of the double exponential Wiener process, and further helping to improve the accuracy and reliability of battery degradation data reconstruction.
[0095] Step 230: Taking the estimated values of the parameters of the battery degradation prediction model as the prior distribution, based on the Bayesian theory, calculate the posterior distribution of the parameters of the battery degradation prediction model.
[0096] Among them, the parameters b, d and σ B After the initial maximum likelihood estimation, it will be fixed, and subsequent updates will focus on parameters a and c.
[0097] In order to reflect the random differences between batteries, a t With c t are considered as random variables and follow a normal distribution. and Among them, u a The initial value is set to The initial value is set to u c The initial value is set to The initial value is set to Where k is a constant. Assume that the target feature in the real data set is X 0，n =[x 0，1 , x 0，2 ,...x 0，n ], where the time corresponding to the feature is t 0，1 , t 0，2 ,...t 0，n , the exponential function term is as well as
[0098] Then according to Bayesian theory, update the posterior distribution N(u a，n , σ 2 a，n ) and N(u c，n , σ 2 c，n ),there will be:
[0099]
[0100] Right now
[0101]
[0102] Where ΔX 0，n =[X 0，2 -X 0，1 , X 0，3 -X 0，2 , ..., X 0，n -X 0，n-1 ] is X 0，n The first-order difference of . Where I is the identity matrix of order n-1. t With c t Equal, the posterior distributions corresponding to the two equations can be updated as follows:
[0103]
[0104] Among them, a t With c t It follows a normal distribution.
[0105] Step 240, determine the update parameters of each parameter according to the posterior distribution of each parameter of the battery attenuation prediction model, and input the update parameters of each parameter into the battery attenuation prediction model to reconstruct the current attenuation data of the battery to obtain a reconstructed data set of the current attenuation data of the battery.
[0106] Exemplarily, according to the above step 230, the posterior distribution of each parameter of the battery degradation prediction model can be obtained. Since the parameters b, d and σ B After the initial maximum likelihood estimation, it will be fixed, and the posterior distribution of parameters a and c can be obtained. For example, a t With c t Take the first M maximum expected values in the corresponding normal distribution as the true value and as the update parameter, substitute them into the battery attenuation prediction model of the double exponential Wiener process, reconstruct the current attenuation data of the battery, and obtain the reconstructed data set of the battery attenuation prediction model. t With c t The first M maximum expected values in the corresponding normal distribution are taken as true values respectively, and the obtained reconstructed data set includes M groups of reconstructed curves.
[0107] Step 250: evaluate and update the reconstructed data set according to the correlation coefficient evaluation index, the decay rate evaluation index and the Euclidean distance evaluation index in sequence.
[0108] The preset evaluation index at least includes one or more of a correlation coefficient evaluation index, a decay rate evaluation index and a Euclidean distance evaluation index. The preset evaluation index specifically includes different types of evaluation indexes and different execution orders, and the corresponding parameter evaluation and update accuracy are different. The preset evaluation index includes the three indicators of correlation coefficient evaluation index, decay rate evaluation index and Euclidean distance evaluation index at the same time, and the evaluation and update method according to the correlation coefficient evaluation index, decay rate evaluation index and Euclidean distance evaluation index in sequence is the best.
[0109] Exemplarily, in the technical solution of this embodiment, the preset evaluation indicators include a correlation coefficient evaluation indicator, a decay rate evaluation indicator and a Euclidean distance evaluation indicator, and are first evaluated and updated according to the correlation coefficient evaluation indicator, then evaluated and updated according to the decay rate evaluation indicator, and finally evaluated and updated according to the Euclidean distance evaluation indicator.
[0110] In the technical solution of this embodiment, the working principle of the battery attenuation data processing method is as follows: Figure 2First, the current attenuation data of the battery is input into the pre-built battery attenuation prediction model based on the Wiener process of the nonlinear integral term for fitting, and the parameter estimation model of the battery attenuation prediction model is obtained. Then, the log-likelihood equation of the battery attenuation prediction model is simplified and the partial derivatives are solved in turn to obtain the estimated values of the various parameters of the battery attenuation prediction model. Secondly, the estimated values of the various parameters of the battery attenuation prediction model are used as the prior distribution, and the posterior distribution of the various parameters of the battery attenuation prediction model is calculated based on the Bayesian theory. According to the posterior distribution of the various parameters of the battery attenuation prediction model, the update parameters of each parameter are determined, and the updated parameters of each parameter are input into the battery attenuation prediction model to reconstruct the current attenuation data of the battery and obtain the reconstructed data set of the current attenuation data of the battery. Finally, the reconstructed data set is evaluated and updated according to the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index. It can be seen that by using the double exponential Wiener process to construct a battery attenuation prediction model, inputting the current attenuation data of the battery into the prediction model for fitting, and using maximum likelihood estimation to obtain the prior distribution of each parameter of the prediction model, the fitting effect of the battery attenuation prediction model of the double exponential Wiener process can be improved, thereby improving the accuracy of parameter estimation. Then, the posterior distribution of each parameter of the prediction model is updated through Bayesian theory to reconstruct the attenuation data of the battery, which can accurately capture the nonlinear characteristics of the battery attenuation data. In the process of data reconstruction, the time series characteristics and physical consistency of the battery attenuation data can be maintained, thereby increasing the richness, stability, continuity and accuracy of the battery attenuation data reconstruction, and improving the prediction accuracy of the battery attenuation data in the case of scarce data samples, providing more real and diverse training data for battery attenuation prediction. Moreover, this method is simple and easy to implement, with adjustable parameters and strong applicability, especially suitable for the case of scarce samples in battery life prediction. In addition, after reconstructing the current attenuation data of the battery, the reconstructed data set is evaluated and updated in the order of the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index, which can further improve the accuracy and reliability of the battery attenuation data reconstruction, thereby improving the accuracy of the battery life prediction.
[0111] Figure 3 is a flow chart of another method for processing battery attenuation data provided in an embodiment of the present application. Based on the above embodiment, optionally, refer to Figure 3 , the method comprises the following steps:
[0112] Step 310: input the current attenuation data of the battery into a pre-built battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, and obtain a parameter estimation model of the battery attenuation prediction model.
[0113] Step 320: Simplify the log-likelihood equation of the battery degradation prediction model and solve the partial derivatives in turn to obtain the estimated values of the parameters of the battery degradation prediction model.
[0114] Step 330: Taking the estimated values of the parameters of the battery degradation prediction model as the prior distribution, based on the Bayesian theory, calculate the posterior distribution of the parameters of the battery degradation prediction model.
[0115] Step 340: determine the update parameters of each parameter according to the posterior distribution of each parameter of the battery attenuation prediction model, and input the update parameters of each parameter into the battery attenuation prediction model to reconstruct the current attenuation data of the battery to obtain a reconstructed data set of the current attenuation data of the battery.
[0116] Step 350: Acquire the real attenuation data of the battery, calculate the correlation coefficient between each group of reconstructed curves and the real attenuation data, select the reconstructed curves with a correlation coefficient greater than a preset correlation coefficient, and record them as the first data set.
[0117] Exemplarily, assuming that the reconstructed data set of the battery attenuation prediction model obtained in the above step 340 includes M groups of reconstructed curves, the reconstructed curves of the M groups of battery attenuation prediction models are first evaluated and updated according to the correlation coefficient evaluation index, and some (for example, P group, P is less than or equal to M group) reconstructed curves that meet the correlation coefficient evaluation index are screened out from the M groups of reconstructed curves. Then, the P group of reconstructed curves are evaluated and updated according to the attenuation rate evaluation index, and some (for example, Q group, Q is less than or equal to P) reconstructed curves that meet the attenuation rate evaluation index are screened out from the P group of reconstructed curves. Finally, the Q group of reconstructed curves are evaluated and updated according to the Euclidean distance evaluation index, and the only reconstructed curve is screened out from the Q group as the optimal reconstruction curve for this iteration. Specifically, the implementation process of evaluating and updating according to each evaluation index in turn is as follows:
[0118] Since the parameters of the battery attenuation prediction model of the double exponential Wiener process are randomly sampled from the related normal distribution, the generated battery capacity attenuation curve may have a variety of shapes, so it is necessary to evaluate whether the reconstructed curve is reasonable. Among them, the rationality of the reconstructed curve is judged based on the correlation coefficient between the true value and the reconstructed value. Exemplarily, based on M groups of reconstructed curves and the real attenuation data of the battery, the correlation coefficients between the M groups of reconstructed curves and the real attenuation data of the battery are calculated according to the mathematical formula of the correlation coefficient between variables, and all the reconstructed curves in the M groups with correlation coefficients greater than the preset correlation coefficient are screened out and recorded as the first data set. It is assumed that the first data set has P groups (P is less than or equal to M) of reconstructed curves.
[0119] Among them, the mathematical formula of the correlation coefficient between variables is as follows:
[0120]
[0121] Among them, ρ XY is the correlation coefficient, X and Y are two variables, cov(X, Y) is the covariance of the two variables, σ X is the standard deviation of variable X, σ Y is the standard deviation of variable Y.
[0122] If the correlation coefficient between the two groups of variables is greater than the preset correlation coefficient, it is considered that the two groups of variables have a strong correlation. When the correlation coefficient between the generated reconstructed curve and the original data is greater than the preset correlation coefficient, the reconstructed curve is adopted and the next round of evaluation is performed according to step 360.
[0123] For example, the preset correlation coefficient is 0.8. In addition, the preset correlation coefficient may also be other values, which may be set according to actual conditions and are not specifically limited here.
[0124] Step 360: Calculate the true attenuation rate of each group of reconstructed curves in the first data set, select the reconstructed curves whose true attenuation rate meets the preset attenuation rate condition, and record them as the second data set.
[0125] Specifically, the evaluation is performed according to the above step 350. Since the strong correlation only indicates that the shape of the reconstructed curve is similar to the true value, it cannot guarantee that the reconstructed curve has a similar battery life, for example, two parallel battery capacity decay curves. Therefore, the real decay rate is introduced as the basis for the evaluation and screening of the second round of reconstructed curves. The mathematical expression of the real decay rate DR is as follows:
[0126]
[0127] Among them, x initial represents the initial capacity value, x targetcycle Indicates the capacity value of the target number of cycles, cycle target Indicates the number of turns that the capacity curve goes through.
[0128] Specifically, the first data set obtained in the above step 350, i.e., the P group of reconstructed curves, is calculated according to the formula of the real attenuation rate to obtain the real attenuation rate of the P group of reconstructed curves, and the reconstructed curves that meet the conditions are screened out from the P group of reconstructed curves according to the preset attenuation rate conditions. Assuming that there are Q groups of reconstructed curves, they are recorded as the second data set. Among them, the preset attenuation rate condition is: take the first 50% (it can also be set to other proportional values, which can be set according to actual conditions) of the reconstructed curves whose real attenuation rate values DR are closest to the real value and adopt them, and perform the next round of evaluation according to step 370.
[0129] Step 370: Calculate the Euclidean distance of each group of reconstructed curves in the second data set, select the reconstructed curve with the smallest Euclidean distance, and record it as the third data set.
[0130] Due to the random fluctuations in the battery attenuation process, similar attenuation rates DR cannot guarantee that the reconstructed curve has a similar attenuation trend as the true value. Therefore, the Euclidean distance evaluation index is introduced as a new evaluation condition. The mathematical expression of the Euclidean distance is as follows:
[0131]
[0132] Among them, cycle target Indicates the number of turns the capacity curve has gone through, represents the i-th round of reconstructed data, x i Represents the true value of the i-th circle.
[0133] Exemplarily, the minimum Euclidean distance is used as a constraint condition to restrict the parameters of the prediction model and the reconstruction curve, that is, the reconstruction curve with the minimum Euclidean distance is selected from the Q group as the third data set, that is, the optimal reconstruction data set of this iteration.
[0134] Therefore, by constraining each group of reconstructed data sets of the current attenuation data of the battery obtained in step 340 according to the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index, the accuracy of updating the parameters of the battery attenuation prediction model can be improved, thereby improving the accuracy and reliability of the battery attenuation data reconstruction.
[0135] Step 380: Update various parameters of the battery degradation prediction model using various parameters of the reconstructed curve corresponding to the third data set.
[0136] Specifically, the reconstructed curve selected in step 370 is used to update the posterior distribution of each parameter of the battery attenuation prediction model of the double exponential Wiener process. The updated parameter posterior distribution is regarded as the prior distribution of each parameter of the battery attenuation prediction model of the double exponential Wiener process, and is substituted into the battery attenuation prediction model of the double exponential Wiener process for a new round of parameter update.
[0137] In the technical solution of this embodiment, the working principle of the battery attenuation data processing method is as follows: Figure 3First, the current attenuation data of the battery is input into the pre-built battery attenuation prediction model based on the Wiener process of the nonlinear integral term for fitting, and the parameter estimation model of the battery attenuation prediction model is obtained. Then, the log-likelihood equation of the battery attenuation prediction model is simplified and the partial derivatives are solved in turn to obtain the estimated values of the various parameters of the battery attenuation prediction model. Secondly, the estimated values of the various parameters of the battery attenuation prediction model are used as the prior distribution, and the posterior distribution of the various parameters of the battery attenuation prediction model is calculated based on the Bayesian theory. According to the posterior distribution of the various parameters of the battery attenuation prediction model, the update parameters of each parameter are determined, and the updated parameters of each parameter are input into the battery attenuation prediction model to reconstruct the current attenuation data of the battery and obtain the reconstructed data set of the current attenuation data of the battery. The real attenuation data of the battery is obtained, and the correlation coefficient between each group of reconstructed curves and the real attenuation data is calculated, and the reconstructed curves with a correlation coefficient greater than the preset correlation coefficient are screened out and recorded as the first data set. The real attenuation rate of each group of reconstructed curves in the first data set is calculated, and the reconstructed curves whose real attenuation rate meets the preset attenuation rate condition are screened out and recorded as the second data set. The Euclidean distance of each group of reconstructed curves in the second data set is calculated, and the reconstructed curve with the smallest Euclidean distance is selected and recorded as the third data set. Finally, the parameters of the battery attenuation prediction model are updated by the parameters of the reconstructed curve corresponding to the third data set. It can be seen that by constructing a battery attenuation prediction model using a double exponential Wiener process, inputting the current attenuation data of the battery into the prediction model for fitting, and using maximum likelihood estimation to obtain the prior distribution of each parameter of the prediction model, the fitting effect of the battery attenuation prediction model of the double exponential Wiener process can be improved, thereby improving the accuracy of parameter estimation. Then, the posterior distribution of each parameter of the prediction model is updated through Bayesian theory to reconstruct the attenuation data of the battery, which can accurately capture the nonlinear characteristics of the battery attenuation data, and can maintain the time series characteristics and physical consistency of the battery attenuation data during the data reconstruction process, thereby increasing the richness, stability, continuity and accuracy of the battery attenuation data reconstruction, and improving the prediction accuracy of the battery attenuation data in the case of scarce data samples, providing more real and diverse training data for battery attenuation prediction. The method is simple and easy to implement, with adjustable parameters and strong applicability, especially for the case of scarce samples in battery life prediction. In addition, after reconstructing the current attenuation data of the battery, the reconstruction curves corresponding to each group of reconstructed data sets are evaluated and updated in the order of the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index, and the optimal reconstruction curve is screened out, thereby further improving the accuracy and reliability of the battery attenuation data reconstruction, and thus improving the accuracy of the battery life prediction.
[0138] On the basis of the above embodiment, optionally, the optimal reconstruction curve of this iteration selected in the above step 370 is used to update the posterior distribution of each parameter of the battery attenuation prediction model of the double exponential Wiener process. And the updated parameter posterior distribution is regarded as the prior distribution of each parameter of the battery attenuation prediction model of the double exponential Wiener process, and is substituted into the battery attenuation prediction model of the double exponential Wiener process for a new round of parameter update. Repeat the above steps 310-380 until and The optimal reconstruction curve finally screened out from the corresponding number of iterations when the value remains almost unchanged is used as the optimal reconstruction data set to achieve data amplification for battery degradation prediction. Therefore, through repeated iterations, the accuracy and reliability of battery degradation data reconstruction can be improved, thereby improving the accuracy of battery life prediction.
[0139] Figure 4 This is a principle structure block diagram of a battery attenuation data processing device provided in an embodiment of the present application. This embodiment of the present application also provides a battery attenuation data processing device, please refer to Figure 4 The device 100 includes: a fitting module 10, which is used to input the current attenuation data of the battery into a pre-constructed battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, so as to obtain a parameter estimation model of the battery attenuation prediction model; a parameter estimation module 20, which is used to estimate each parameter of the battery attenuation prediction model according to the parameter estimation model, so as to obtain estimated values of each parameter of the battery attenuation prediction model; a reconstruction module 30, which is used to reconstruct the current attenuation data of the battery according to the estimated values of each parameter of the battery attenuation prediction model and the battery attenuation prediction model, so as to obtain a reconstructed data set of the current attenuation data of the battery; and an evaluation and update module 40, which is used to evaluate and update the reconstructed data set according to preset evaluation indicators.
[0140] The technical solution of the embodiment of the present application provides a data processing device for battery attenuation, which includes: a fitting module for inputting the current attenuation data of the battery into a pre-built battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, and obtaining a parameter estimation model of the battery attenuation prediction model; a parameter estimation module for estimating the various parameters of the battery attenuation prediction model according to the parameter estimation model, and obtaining the estimated values of the various parameters of the battery attenuation prediction model; a reconstruction module for reconstructing the current attenuation data of the battery according to the estimated values of the various parameters of the battery attenuation prediction model and the battery attenuation prediction model, and obtaining a reconstructed data set of the current attenuation data of the battery; an evaluation and update module for evaluating and updating the reconstructed data set according to a preset evaluation index. Through the device, the attenuation data of the battery can be reconstructed, the richness of the battery attenuation data set can be increased, the nonlinear characteristics of the battery data can be accurately captured, and the time series characteristics and physical consistency of the battery attenuation data can be maintained, so as to provide more real and diversified training data for battery life prediction, which is suitable for situations where battery data samples are scarce. In addition, the reconstructed data set is evaluated and updated by a preset evaluation index, so as to further improve the accuracy of the battery attenuation data reconstruction and the accuracy of the battery life prediction.
[0141] In some embodiments, the nonlinear integral term in the battery degradation prediction model based on the Wiener process of the nonlinear integral term includes one or more of an exponential function, a logarithmic function, and a power function.
[0142] In some embodiments, the parameter estimation model of the battery degradation prediction model is a log-likelihood equation of the battery degradation prediction model.
[0143] In some embodiments, the parameter estimation module 20 further includes: a parameter estimation unit, which is used to sequentially simplify the log-likelihood equation of the battery degradation prediction model and solve partial derivatives to obtain estimated values of various parameters of the battery degradation prediction model.
[0144] In some embodiments, the various parameters of the battery attenuation prediction model include first-part parameters and second-part parameters; parameter estimation is also used for: simplifying the log-likelihood equation of the battery attenuation prediction model to obtain a simplified log-likelihood equation; solving partial derivatives of the simplified log-likelihood equation to obtain estimated values of the first-part parameters of the battery attenuation prediction model; substituting the estimated values of the first-part parameters of the battery attenuation prediction model into the simplified log-likelihood equation to obtain estimated values of the second-part parameters of the battery attenuation prediction model.
[0145] In some embodiments, the reconstruction module 30 further includes: a posterior distribution calculation unit, which is used to calculate the posterior distribution of each parameter of the battery degradation prediction model based on the Bayesian theory by taking the estimated values of each parameter of the battery degradation prediction model as the prior distribution.
[0146] The reconstruction unit is used to determine the update parameters of each parameter according to the posterior distribution of each parameter of the battery attenuation prediction model, and input the update parameters of each parameter into the battery attenuation prediction model to reconstruct the current attenuation data of the battery.
[0147] In some embodiments, the preset evaluation index includes at least one or more of a correlation coefficient evaluation index, a decay rate evaluation index, and a Euclidean distance evaluation index.
[0148] In some embodiments, the preset evaluation index includes a correlation coefficient evaluation index, a decay rate evaluation index and a Euclidean distance evaluation index. The evaluation and update module 40 also includes: an evaluation and update unit, which is used to evaluate and update the reconstructed data set according to the correlation coefficient evaluation index, the decay rate evaluation index and the Euclidean distance evaluation index in sequence.
[0149] In some embodiments, the evaluation and update unit is also used to: obtain the actual attenuation data of the battery, calculate the correlation coefficient between each group of reconstructed curves and the actual attenuation data, screen out the reconstructed curves with a correlation coefficient greater than a preset correlation coefficient, and record them as the first data set; calculate the actual attenuation rate of each group of reconstructed curves in the first data set, screen out the reconstructed curves whose actual attenuation rate meets the preset attenuation rate condition, and record them as the second data set; calculate the Euclidean distance of each group of reconstructed curves in the second data set, screen out the reconstructed curve with the smallest Euclidean distance, and record it as the third data set; and update the various parameters of the battery attenuation prediction model through the various parameters of the reconstructed curve corresponding to the third data set.
[0150] In some embodiments, the battery attenuation data processing device also includes a return and repeat execution module, which is used to return and repeatedly execute the operation of inputting the updated reconstructed data set evaluated according to preset evaluation indicators into a pre-built battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting until the reconstructed data set becomes stable.
[0151] The above is a detailed introduction to a battery attenuation data processing method and device provided in an embodiment of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the technical solution and core idea of the present application. Ordinary technicians in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for processing battery attenuation data, characterized in that: include: Inputting the current attenuation data of the battery into a pre-built battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, thereby obtaining a parameter estimation model of the battery attenuation prediction model; estimating each parameter of the battery degradation prediction model according to the parameter estimation model to obtain estimated values of each parameter of the battery degradation prediction model; Reconstructing the current attenuation data of the battery according to the estimated values of the parameters of the battery attenuation prediction model and the battery attenuation prediction model to obtain a reconstructed data set of the current attenuation data of the battery; The reconstructed data set is evaluated and updated according to preset evaluation indicators.
2. The battery attenuation data processing method according to claim 1, characterized in that: The nonlinear integral term in the battery degradation prediction model based on the Wiener process with nonlinear integral term includes one or more of an exponential function, a logarithmic function and a power function.
3. The battery attenuation data processing method according to claim 1, characterized in that: The parameter estimation model of the battery degradation prediction model is a log-likelihood equation of the battery degradation prediction model; The estimating each parameter of the battery degradation prediction model according to the parameter estimation model includes: The log-likelihood equation of the battery degradation prediction model is simplified and its partial derivatives are solved in turn to obtain estimated values of various parameters of the battery degradation prediction model.
4. The battery attenuation data processing method according to claim 3, characterized in that: The parameters of the battery degradation prediction model include first part parameters and second part parameters; The log-likelihood equation of the battery degradation prediction model is simplified and partial derivatives are solved in sequence to obtain estimated values of various parameters of the battery degradation prediction model, including: Simplifying the log-likelihood equation of the battery degradation prediction model to obtain a simplified log-likelihood equation; Solving the partial derivative of the simplified log-likelihood equation to obtain estimated values of the first part of parameters of the battery degradation prediction model; Substitute the estimated values of the first part of the parameters of the battery degradation prediction model into the simplified log-likelihood equation to obtain the estimated values of the second part of the parameters of the battery degradation prediction model.
5. The battery attenuation data processing method according to claim 1, characterized in that: The reconstructing the current attenuation data of the battery according to the estimated values of the parameters of the battery attenuation prediction model and the battery attenuation prediction model includes: Taking the estimated values of the parameters of the battery degradation prediction model as the prior distribution, and calculating the posterior distribution of the parameters of the battery degradation prediction model based on the Bayesian theory; The update parameters of each parameter of the battery attenuation prediction model are determined according to the posterior distribution of each parameter, and the update parameters of each parameter are input into the battery attenuation prediction model to reconstruct the current attenuation data of the battery.
6. The battery attenuation data processing method according to claim 1, characterized in that: The preset evaluation index includes at least one or more of a correlation coefficient evaluation index, a decay rate evaluation index and a Euclidean distance evaluation index.
7. The method for processing battery attenuation data according to claim 6, characterized in that: The preset evaluation indicators include a correlation coefficient evaluation indicator, a decay rate evaluation indicator and a Euclidean distance evaluation indicator; The step of evaluating and updating the reconstructed data set according to a preset evaluation index includes: The reconstructed data set is evaluated and updated in turn according to the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index.
8. The method for processing battery attenuation data according to claim 7, characterized in that: The reconstructed data set includes a plurality of groups of reconstructed curves; The step of evaluating and updating the reconstructed data set in accordance with the correlation coefficient evaluation index, the attenuation rate evaluation index and the Euclidean distance evaluation index in sequence includes: Acquire the real attenuation data of the battery, calculate the correlation coefficient between each group of reconstructed curves and the real attenuation data, screen out the reconstructed curves with a correlation coefficient greater than a preset correlation coefficient, and record them as the first data set; Calculating the true attenuation rate of each group of reconstructed curves in the first data set, screening out the reconstructed curves whose true attenuation rates meet the preset attenuation rate conditions, and recording them as the second data set; Calculating the Euclidean distance of each group of reconstructed curves in the second data set, selecting the reconstructed curve with the smallest Euclidean distance, and recording it as the third data set; And the various parameters of the battery degradation prediction model are updated through the various parameters of the reconstructed curve corresponding to the third data set.
9. The method for processing battery attenuation data according to claim 8, characterized in that: After evaluating and updating the reconstructed data set according to the preset evaluation index, the method further includes: The reconstructed data set evaluated according to the preset evaluation index and updated is returned and repeatedly input into the pre-built battery degradation prediction model based on the nonlinear integral term Wiener process for fitting until the reconstructed data set becomes stable.
10. A battery attenuation data processing device, characterized in that: include: A fitting module, used for inputting the current attenuation data of the battery into a pre-built battery attenuation prediction model based on a nonlinear integral term Wiener process for fitting, so as to obtain a parameter estimation model of the battery attenuation prediction model; A parameter estimation module, used to estimate each parameter of the battery degradation prediction model according to the parameter estimation model to obtain an estimated value of each parameter of the battery degradation prediction model; A reconstruction module, used to reconstruct the current attenuation data of the battery according to the estimated values of each parameter of the battery attenuation prediction model and the battery attenuation prediction model to obtain a reconstructed data set of the current attenuation data of the battery; The evaluation and update module is used to evaluate and update the reconstructed data set according to preset evaluation indicators.