Method and system for predicting residual life of battery based on survival analysis
Through the battery residual life prediction method based on survival analysis, an accelerated failure time model is constructed, which solves the problem that traditional methods are difficult to deal with truncated data, and achieves more accurate battery life prediction, which is suitable for various types of batteries.
Patent Information
- Application Number
- CN202411080553.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-12-06
AI Technical Summary
Traditional battery life prediction methods are difficult to effectively process truncated data, resulting in bias and uncertainty in prediction.
The remaining life prediction method of the battery is adopted based on survival analysis. By acquiring and preprocessing the historical data of the battery, an accelerated failure time model is constructed, and the model parameters are optimized using methods such as maximum likelihood estimation and Weibull distribution to predict the remaining life of the battery.
It improves the processing capability of truncated data, enhances the accuracy of life prediction, provides an important basis for equipment maintenance, and is suitable for various types of batteries.
Smart Images

Figure CN119087262A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of energy storage batteries, and in particular relates to a method and system for predicting remaining battery life based on survival analysis. Background Art
[0002] In the field of battery applications, accurate prediction of battery life is crucial. The prediction of battery life not only affects the performance and reliability of products such as electric vehicles and mobile devices, but is also directly related to the user experience and production costs. However, the life of the battery is affected by many factors, including the number of charge and discharge cycles, operating temperature, charging rate, depth of charge and discharge, etc. In order to accurately predict the remaining life of the battery, researchers and engineers have been looking for various methods. Traditional battery life prediction methods are mainly based on the analysis of battery historical data and the establishment of physical models. By collecting and analyzing battery parameters such as the number of charge and discharge cycles, voltage, temperature, etc., statistical analysis methods or physical models are used for prediction. However, this method has certain limitations, especially when faced with truncated data.
[0003] In actual applications, due to the different use environments and working conditions of batteries, some batteries may become invalid or scrapped early during the study period, while other batteries continue to be used. This results in the collected battery data being truncated, that is, the observed data only reflects the battery conditions during the observation period, and the information of batteries that are scrapped in the future is not included in the data. Traditional analysis methods often cannot handle this kind of truncated data well, resulting in certain deviations and uncertainties in the prediction of battery life. Summary of the invention
[0004] The purpose of the present invention is to provide a method and system for predicting the remaining battery life based on survival analysis to solve the problems raised in the background technology.
[0005] The present invention achieves the above-mentioned purpose through the following technical solutions:
[0006] In a first aspect, the present invention proposes a method for predicting remaining battery life based on survival analysis, the method comprising:
[0007] Acquire temperature values, current values, soc values and time information in the sample battery charge and discharge history data to form an indicator data set, and divide the preprocessed indicator data set into a training set and a test set;
[0008] Constructing an accelerated failure time model, constructing an accelerated failure time model, training the accelerated failure time model through the training set to fit the model, optimizing parameters of the accelerated failure time model using the maximum likelihood estimation method and the test set, and determining a final accelerated failure time model;
[0009] The indicator data set under the current state of the target battery is collected and a target verification set is determined, the target verification set is input into the final accelerated failure time model, and the remaining life of the target battery is determined according to the model output.
[0010] Furthermore, the indicator data set includes the following indicators of the sample batteries: low-temperature charging time, high-temperature charging time, high-temperature discharging time, low-temperature discharging time, high-current charging time, high-current discharging time, duration of discharge depth soc<15%, duration of discharge depth soc<10%, duration of discharge depth soc<5%, and the current calendar life of the battery.
[0011] Furthermore, the preprocessing includes: processing missing values, detecting and processing outliers, data conversion and standardization, removing duplicate data, and unifying data formats for the indicator data set.
[0012] Furthermore, the construction of the accelerated failure time model includes the following steps:
[0013] (1) The event is defined as battery scrapping, the indicator data set is used as the risk set, and each indicator in the indicator data set is used as a covariate;
[0014] (2) Constructing an accelerated failure time model based on the above events, the risk set and the covariates, wherein the survival time is the remaining battery life, and is implemented based on the following formula:
[0015] log(T)=β 0 +β 1 *X 1 +β 2 *X 2 +…+β p *X p +
[0016] Wherein, T is the survival time, X 1 ,X 2 ,…,X p is the covariate, β 0 ,β 1 ,…,β p is the accelerated failure time model parameter, and is the error term.
[0017] Furthermore, constructing the accelerated failure time model also includes: using any one of lognormal distribution, exponential distribution or Weibull distribution as a distribution hypothesis of the error term.
[0018] Furthermore, the method of optimizing the parameters of the accelerated failure time model by using the residual square minimization algorithm and the test set to determine the final accelerated failure time model includes the following steps:
[0019] (1) Constructing the likelihood function: Let the survival time T obey the Weibull distribution, its density function is f(t), the survival function is s(t), and the risk function is h(t), then the likelihood function is as follows:
[0020]
[0021] Among them, θ is the likelihood function model parameter, t i is the failure time of the ith observation, δ i is an indicator variable, when δ i =1 indicates a failure event occurs, when δ i =0 means truncation; γ is the shape parameter, λ is the scale parameter, in the AFT model, λ=exp(β 0 +β'X), where β is the coefficient of the covariate and X is the covariate vector;
[0022] (2) Optimize the likelihood function: maximize the log-likelihood function logL(θ):
[0023]
[0024] (3) Solving parameters: Use numerical optimization methods to solve the parameter value where the derivative is zero. The parameter value β is 0 ,β 1 ,...,β p That is the accelerated failure time model parameter.
[0025] In a second aspect, the present invention proposes a battery remaining life prediction system based on survival analysis, the system comprising:
[0026] A data acquisition module is used to obtain temperature values, current values, soc values and time information in the charge and discharge history data of the sample battery to form an indicator data set, and divide the preprocessed indicator data set into a training set and a test set;
[0027] A model determination module is used to construct an accelerated failure time model, train the accelerated failure time model through the training set to fit the model, optimize the parameters of the accelerated failure time model using the maximum likelihood estimation method and the test set, and determine the final accelerated failure time model;
[0028] The life prediction module is used to collect the indicator data set under the current state of the target battery and determine the target verification set, input the target verification set into the final accelerated failure time model, and determine the remaining life of the target battery according to the model output.
[0029] The beneficial effects of the present invention are:
[0030] (1) The prediction method of the present invention can better handle truncated data, that is, the data observed during the study period only reflects the battery status during the observation period, while the information of batteries that are scrapped in the future and not observed is not included in the data, thereby improving the accuracy of the estimation.
[0031] (2) The present invention can accurately estimate the remaining life of the battery by combining the battery history data and the probability of the remaining life event, providing an important basis for equipment maintenance.
[0032] (3) The prediction method in the present invention is simple and easy to implement, applicable to various types of batteries, and has broad application prospects in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 The figure is a flow chart of a method for predicting remaining battery life based on survival analysis according to an embodiment of the present invention. DETAILED DESCRIPTION
[0034] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.
[0035] Example 1
[0036] like Figure 1 As shown, this embodiment proposes a method for predicting the remaining battery life based on survival analysis, and the method includes the following steps:
[0037] S1, obtaining temperature values, current values, soc values and time information in the charge and discharge history data of the sample battery to form an index data set, and dividing the preprocessed index data set into a training set and a test set;
[0038] S2, constructing an accelerated failure time model, training the accelerated failure time model through the training set to fit the model, optimizing the parameters of the accelerated failure time model using the maximum likelihood estimation method and the test set, and determining the final accelerated failure time model;
[0039] S3. Collect the indicator data set of the target battery in the current state and determine the target verification set, input the target verification set into the final accelerated failure time model, and determine the remaining life of the target battery according to the model output.
[0040] Further preferably, the indicator data set includes the following indicators of the sample batteries: low temperature charging time, high temperature charging time, high temperature discharging time, low temperature discharging time, high current charging time, high current discharging time, duration of discharge depth soc<15%, duration of discharge depth soc<10%, duration of discharge depth soc<5%, and the current calendar life of the battery.
[0041] Further preferably, the preprocessing includes: processing missing values, detecting and processing outliers, data conversion and standardization, removing duplicate data, and unifying data formats for the indicator data set. The purpose of preprocessing is to provide a reliable data basis for subsequent survival analysis modeling.
[0042] In this embodiment, a prediction model of the remaining life of the battery is constructed by using a survival analysis method, and the remaining life of the battery is more accurately predicted based on the truncated data of the battery. Further preferably, constructing an accelerated failure time model (AFT) includes the following steps:
[0043] (1) Define events and risk sets; define the event as battery scrapping, use the indicator data set as the risk set, and use each indicator in the indicator data set as a covariate;
[0044] It is understandable that in the scenario of battery life prediction of the present invention, the event is defined as battery scrapping (that is, the battery can no longer be used). At the same time, a risk set needs to be defined, which includes all factors that may cause the event to occur, which refers to the indicator data set obtained in S1.
[0045] (2) Constructing an accelerated failure time model: Constructing an accelerated failure time model based on the above events, risk sets and covariates, where the survival time is the remaining life of the battery and is implemented based on the following formula:
[0046] log(T)=β 0 +β 1 *X 1 +β 2 *X 2 +…+β p *X p +
[0047] Wherein, T is the survival time, X 1 ,X 2 ,…,X p is the covariate, β 0 ,β 1 ,…,β p is the accelerated failure time model parameter, and is the error term.
[0048] In this embodiment, the accelerated failure time model is adopted to assume that the above covariates directly affect the survival time by shortening or extending the life span, that is, to explain the influence of the above different risk factors on the survival time (remaining battery life).
[0049] More specifically, the accelerated failure time model in the prediction scenario of the present invention is expressed as:
[0050] log(T)=β 0 +β 1 *LT_charge+β 2 *HT_charge+β 3 *HT_discharge+β
[0051] 4 *LT_discharge+β 5 *HC_charge+β 6 *HC_discharge+β 7 *SOC15+β 8 *SOC10+β 9 *SOC5+β 10 *Calendar_life+;
[0052] in:
[0053] LT_charge is the low temperature charging time.
[0054] HT_charge is the high temperature charging time.
[0055] HT_discharge is the high temperature discharge duration.
[0056] LT_discharge is the low temperature discharge duration.
[0057] HC_charge is the high current charging time.
[0058] HC_discharge is the high current discharge duration.
[0059] SOC15 is the duration of discharge depth SOC<15.
[0060] SOC10 is the duration of discharge depth SOC<10.
[0061] SOC5 is the duration of discharge depth SOC<5.
[0062] Calendar_life is the current calendar life of the battery.
[0063] Further preferably, constructing the accelerated failure time model further comprises: using any one of lognormal distribution, exponential distribution or Weibull distribution as the distribution assumption of the error term.
[0064] Preferably, the accelerated failure time model is optimized using the minimization of residual squares algorithm and the test set, and the model parameter β is estimated using the maximum likelihood estimation method (MLE). 0 ,β 1 ,...,β p , determine the final accelerated failure time model, including the following steps:
[0065] The maximum likelihood estimation method and the test set are used to optimize the parameters of the accelerated failure time model to determine the final accelerated failure time model, including the following steps:
[0066] Solving for the coefficient β in the accelerated failure time (AFT) model 0 ,β 1 ,...,β p This can be achieved by the maximum likelihood estimation (MLE) method. Solution steps:
[0067] (1) Constructing the likelihood function
[0068] The likelihood function is constructed based on the failure time data. Let the failure time T obey the Weibull distribution, its density function is f(t), the survival function is s(t), and the risk function is h(t). Then the likelihood function can be expressed as:
[0069]
[0070] Among them, θ is the likelihood function model parameter, t i is the failure time of the ith observation, δ i is an indicator variable, when δ i =1 indicates a failure event occurs, when δ i =0 means truncation.
[0071]
[0072] Where γ is the shape parameter and λ is the scale parameter. In the AFT model, λ=exp(β 0 +β'X), where β is the coefficient of the covariate and X is the covariate vector.
[0073] (2) Optimizing the likelihood function
[0074] In order to find the parameter θ that maximizes the likelihood function L(θ), we need to maximize the log-likelihood function logL(θ):
[0075]
[0076] (3) Solving parameters
[0077] Use numerical optimization methods (such as the Newton-Raphson method) to solve the parameter value where the derivative is zero. The parameter value β is 0 ,β 1 ,...,β p is the accelerated failure time model parameter.
[0078] According to the embodiment of the present invention, since the battery data collected in the prior art is truncated, the present invention uses the accelerated failure time model to construct a prediction model. After further optimizing the parameters through the above algorithm, the indicator data set of the current state of the battery is input into the model for prediction to obtain the survival time log(T), and then the actual survival time T, i.e. the remaining life of the battery, is obtained through exponential operation. The data observed during the current study period is used to reflect the remaining life of the battery during the current study period in real time.
[0079] Based on the same inventive concept, this embodiment also proposes a battery remaining life prediction system based on survival analysis, the system is used to implement the above prediction method, and the system includes:
[0080] A data acquisition module is used to obtain temperature values, current values, soc values and time information in the charge and discharge history data of the sample battery to form an indicator data set, and divide the preprocessed indicator data set into a training set and a test set;
[0081] A model determination module is used to construct an accelerated failure time model, train the accelerated failure time model through the training set to fit the model, optimize the parameters of the accelerated failure time model using the maximum likelihood estimation method and the test set, and determine the final accelerated failure time model;
[0082] The life prediction module is used to collect the indicator data set of the target battery in the current state and determine the target verification set, input the target verification set into the final accelerated failure time model, and determine the remaining life of the target battery based on the model output.
[0083] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0084] In addition, each functional module in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0085] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0086] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for predicting remaining battery life based on survival analysis, characterized in that: The method comprises: Acquire temperature values, current values, soc values and time information in the sample battery charge and discharge history data to form an indicator data set, and divide the preprocessed indicator data set into a training set and a test set; Constructing an accelerated failure time model, training the accelerated failure time model through the training set to fit the model, optimizing the parameters of the accelerated failure time model using the maximum likelihood estimation method and the test set, and determining a final accelerated failure time model; The indicator data set under the current state of the target battery is collected and a target verification set is determined, the target verification set is input into the final accelerated failure time model, and the remaining life of the target battery is determined according to the model output.
2. The method for predicting remaining battery life based on survival analysis according to claim 1, characterized in that: The indicator data set includes the following indicators of the sample battery: low-temperature charging time, high-temperature charging time, high-temperature discharging time, low-temperature discharging time, high-current charging time, high-current discharging time, duration of discharge depth soc<15%, duration of discharge depth soc<10%, duration of discharge depth soc<5%, and the current calendar life of the battery.
3. The method for predicting remaining battery life based on survival analysis according to claim 2, characterized in that: The preprocessing includes: processing missing values, detecting and processing outliers, data conversion and standardization, removing duplicate data, and unifying data formats for the indicator data set.
4. The method for predicting remaining battery life based on survival analysis according to claim 2, characterized in that: The construction of the accelerated failure time model comprises the following steps: (1) The event is defined as battery scrapping, the indicator data set is used as the risk set, and each indicator in the indicator data set is used as a covariate; (2) Constructing an accelerated failure time model based on the above events, the risk set and the covariates, wherein the survival time is the remaining battery life, and is implemented based on the following formula: log(T)=β0+β1*X1+β2*X2+…+β p *X p + Where T is the survival time, X1, X2, ..., X p are the covariates, β0,β1,…,β p is the accelerated failure time model parameter, and is the error term.
5. The method for predicting remaining battery life based on survival analysis according to claim 4, characterized in that: Constructing the accelerated failure time model also includes: using any one of lognormal distribution, exponential distribution or Weibull distribution as a distribution hypothesis of the error term.
6. The method for predicting remaining battery life based on survival analysis according to claim 4, characterized in that: The method of using the maximum likelihood estimation method and the test set to optimize the parameters of the accelerated failure time model to determine the final accelerated failure time model includes the following steps: (1) Constructing the likelihood function: Let the survival time T obey the Weibull distribution, its density function is f(t), the survival function is s(t), and the risk function is h(t), then the likelihood function is as follows: Among them, θ is the likelihood function model parameter, t i is the failure time of the ith observation, δ i is an indicator variable, when δ i =1 indicates a failure event occurs. i =0 means truncation; γ is the shape parameter, λ is the scale parameter, and in the AFT model, λ = exp(β0+β'X), where β is the coefficient of the covariate and X is the covariate vector; (2) Optimize the likelihood function: maximize the log-likelihood function logL(θ): (3) Solving parameters: Use numerical optimization methods to solve the parameter values where the derivative is zero. The parameter values β0, β1, ..., β p That is the accelerated failure time model parameter.
7. A battery remaining life prediction system based on survival analysis, characterized in that: The system comprises: A data acquisition module is used to obtain temperature values, current values, soc values and time information in the charge and discharge history data of the sample battery to form an indicator data set, and divide the preprocessed indicator data set into a training set and a test set; A model determination module, constructing an accelerated failure time model, training the accelerated failure time model through the training set to fit the model, optimizing the parameters of the accelerated failure time model using the maximum likelihood estimation method and the test set, and determining a final accelerated failure time model; The life prediction module is used to collect the indicator data set under the current state of the target battery and determine the target verification set, input the target verification set into the final accelerated failure time model, and determine the remaining life of the target battery according to the model output.
Citation Information
Cited By
Battery spare part prediction method and device, computer equipment and storage medium
CN119622314A
Battery Spare Part Prediction Method, Device, Computer Equipment and Storage Medium
CN119622314B
