Battery residual life prediction and model construction method, system and equipment, and medium

Abstract

The invention provides a battery residual life prediction and model construction method, system and device, and a medium. The construction method comprises the following steps: acquiring a battery aging data set corresponding to a plurality of sample batteries; obtaining a target cycle number interval based on the battery aging data set; preprocessing the battery aging data set based on the target cycle index interval to obtain training sample data; and inputting the training sample data into a preset model to train and obtain a battery remaining life prediction model. According to the method, only the target cycle index interval needs to be determined, and the training sample data is constructed according to the charging and discharging data corresponding to the target cycle index interval, so that the sample data is greatly simplified, the model training is simpler, and meanwhile, the actual residual life of the battery can be accurately predicted; and the confidence coefficient and the confidence interval corresponding to the actual residual life are predicted, so that the probability and the distribution condition of the actual residual life are predicted.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of batteries, and particularly relates to a battery remaining life prediction and model construction method, system, device and medium. BACKGROUND

[0002] Battery Remaining Useful Life (RUL) is a key indicator for safe operation of electric vehicles and energy storage power stations, and is used to represent the remaining number of charge and discharge cycles of the battery. The End of Life (EOL) of a power battery is marked by the attenuation of the battery capacity to a certain value of the rated capacity. There are various research programs for estimating the remaining life of the battery. The most commonly used semi-empirical model is used to fit the battery capacity attenuation curve, including the use of exponential model, polynomial model, etc. However, relying entirely on empirical models will cause large errors, so a certain correction will be made according to the real observed value, for example, a temperature correction exponential model is added.

[0003] In addition, another solution is to extract battery aging parameters and establish a battery electrochemical or equivalent circuit model for analysis, to determine the battery health state and remaining life through the aging trend of the key parameters of the battery, including the growth of the solid electrolyte interface, lithium deposition, active material loss, impedance increase, etc. This solution has the advantage of interpretability, which can determine the attenuation mechanism of the battery. At present, various heuristic algorithms have been researched for battery feature extraction. However, this solution relies on a relatively accurate battery model and parameter extraction algorithm. At present, there are studies that can identify the parameters of one battery in one hour, and can reduce the RMSE (root mean square error) to 14mv (millivolt). However, when there are a large number of batteries, the calculation is large and resource-consuming.

[0004] With the continuous accumulation of real battery operation data and simulation data, various data-driven based schemes begin to be studied. For example, using the autoregressive integrated moving average (ARIMA) model to predict the state of health (SOH), a kind of life prediction model based on double unscented particle filtering (UPF) and least squares support vector machines (LSSVM) algorithm. EMD (Empirical Mode Decomposition) and grey relational analysis (GRA) are used to extract the time series characteristics of the data, and they are input into the deep RNN (recursive neural network), and the RUL of the battery is modeled. The early charge-discharge data before the battery attenuation is used to estimate the RUL of the battery.

[0005] The above various schemes need a large amount of sample data to construct network model when predicting the RUL of the ion battery, or need to spend a lot of time and resources to build an electrochemical model. SUMMARY

[0006] The technical problem to be solved by the present disclosure is to overcome the defects in the prior art that a large amount of sample data is needed to construct a network model when predicting the RUL of the battery, or a large amount of time and resources are needed to build an electrochemical model, and to provide a battery remaining life prediction and model construction method, system, device and medium.

[0007] The present disclosure solves the above technical problems by the following technical solutions:

[0008] In a first aspect, a battery remaining life prediction model construction method is provided, and the construction method comprises:

[0009] Obtaining a battery aging data set corresponding to a plurality of sample batteries;

[0010] The battery aging data set comprises a corresponding sample remaining life and sample charge-discharge data of a plurality of sample batteries from the beginning of the life to the end of the life;

[0011] Obtaining a target cycle number interval based on the battery aging data set;

[0012] Preprocessing the battery aging data set based on the target cycle number interval to obtain training sample data;

[0013] inputting the training sample data into a preset model to obtain the battery remaining life prediction model through training;

[0014] The battery remaining life prediction model is configured to predict a corresponding actual remaining life according to input target battery capacity characteristic data of an actual battery, and a confidence degree and a confidence interval corresponding to the actual remaining life.

[0015] Preferably, the step of obtaining the target cycle interval based on the battery aging data set comprises:

[0016] Based on the sample charge-discharge data corresponding to the sample battery, first battery capacity data corresponding to a first preset battery cycle number and second battery capacity data corresponding to a second preset battery cycle number of the sample battery are obtained.

[0017] The second preset battery cycle number is greater than the first preset battery cycle number.

[0018] The first target sample battery remaining life corresponding to the second battery capacity data of the sample battery is obtained.

[0019] The target cycle interval is calculated based on the first battery capacity data, the second battery capacity data and the first target sample battery remaining life.

[0020] Preferably, the step of calculating the target cycle interval based on the first battery capacity data, the second battery capacity data and the first target sample battery remaining life comprises:

[0021] The capacity difference between the second battery capacity data and the first battery capacity data is calculated.

[0022] The variance of the capacity difference is logarithmically transformed to obtain sample voltage data.

[0023] The target cycle interval is calculated based on the sample voltage data and the first target sample battery remaining life by using a Pearson correlation coefficient method.

[0024] Preferably, the step of preprocessing the battery aging data set based on the target cycle interval to obtain training sample data comprises:

[0025] A preset voltage interval, a third preset battery cycle number and a fourth preset battery cycle number corresponding to the sample battery are obtained.

[0026] The fourth preset battery cycle number is obtained based on a sum of the third preset battery cycle number and the target cycle interval.

[0027] divide the preset voltage interval based on a preset voltage interval, and obtain third battery capacity data and fourth battery capacity data corresponding to the divided preset voltage interval;

[0028] The third battery capacity data is capacity data of the sample battery corresponding to the third preset battery cycle number, and the fourth battery capacity data is capacity data of the sample battery corresponding to the fourth preset battery cycle number.

[0029] Based on the third battery capacity data and the fourth battery capacity data, a plurality of preset dimension sample battery capacity feature data of the sample battery is obtained.

[0030] Based on the sample battery capacity feature data and the second target sample battery remaining life corresponding to the fourth battery capacity data of the sample battery, the training sample data is obtained.

[0031] Preferably, the preset model includes an Ngboost (Nature Gradient Boosting, based on natural gradient boosting) model.

[0032] And / or, the battery capacity corresponding to the process of the plurality of sample batteries from the beginning of the life to the end of the life decays from the rated capacity to 80% of the rated capacity.

[0033] And / or, the battery aging data set includes the remaining life of the sample battery under the constant current charging and variable frequency discharging working condition.

[0034] Preferably, the preset voltage interval is within the full charge and full discharge voltage interval of the sample battery.

[0035] And / or, the target cycle number interval includes a target charging cycle number interval or a target discharging cycle number interval.

[0036] The second aspect also provides a battery remaining life prediction method, the method comprising:

[0037] Obtaining initial charging and discharging data corresponding to an actual battery;

[0038] Pretreating the initial charging and discharging data to obtain target battery capacity feature data;

[0039] Inputting the target battery capacity feature data into a battery remaining life prediction model to output an actual remaining life corresponding to the actual battery, and a confidence and a confidence interval corresponding to the actual remaining life;

[0040] The battery remaining life prediction model is obtained based on the above-mentioned battery remaining life prediction model construction method.

[0041] The third aspect also provides a system for constructing a battery residual life prediction model, comprising:

[0042] a data set acquisition module configured to acquire battery aging data sets corresponding to a plurality of sample batteries;

[0043] The battery aging data sets comprise corresponding sample residual life and sample charge-discharge data of the plurality of sample batteries from the beginning of the life to the end of the life;

[0044] a cycle interval acquisition module configured to acquire a target cycle interval based on the battery aging data sets;

[0045] a sample data acquisition module configured to preprocess the battery aging data sets based on the target cycle interval to acquire training sample data;

[0046] a model training module configured to input the training sample data into a preset model to train the battery residual life prediction model;

[0047] The battery residual life prediction model is configured to predict a corresponding actual residual life, a confidence degree and a confidence interval corresponding to the actual residual life according to input target battery capacity feature data of an actual battery.

[0048] Preferably, the cycle interval acquisition module comprises:

[0049] a first capacity data acquisition unit configured to acquire first battery capacity data corresponding to a first preset battery cycle number and second battery capacity data corresponding to a second preset battery cycle number of the sample batteries based on the sample charge-discharge data of the sample batteries;

[0050] The second preset battery cycle number is greater than the first preset battery cycle number.

[0051] a residual life acquisition unit configured to acquire a first target sample battery residual life corresponding to the second battery capacity data of the sample batteries;

[0052] a cycle interval acquisition unit configured to calculate the target cycle interval based on the first battery capacity data, the second battery capacity data and the first target sample battery residual life.

[0053] Preferably, the cycle interval acquisition unit is configured to calculate a capacity difference value of the second battery capacity data and the first battery capacity data.

[0054] The capacity difference value is logarithmized to obtain sample voltage data.

[0055] The target cycle interval is calculated based on the sample voltage data and the first target sample battery remaining life by using a Pearson correlation coefficient method.

[0056] Preferably, the sample data acquisition module comprises:

[0057] A preset voltage interval acquisition unit is configured to acquire a preset voltage interval corresponding to the sample battery, a third preset battery cycle number, and a fourth preset battery cycle number.

[0058] The fourth preset battery cycle number is obtained based on a sum of the third preset battery cycle number and the target cycle interval.

[0059] A second capacity data acquisition unit is configured to divide the preset voltage interval based on a preset voltage interval, and acquire third battery capacity data and fourth battery capacity data corresponding to the divided preset voltage interval.

[0060] The third battery capacity data is capacity data of the sample battery corresponding to the third preset battery cycle number, and the fourth battery capacity data is capacity data of the sample battery corresponding to the fourth preset battery cycle number.

[0061] A feature data acquisition unit is configured to obtain sample battery capacity feature data of a plurality of preset dimensions of the sample battery based on the third battery capacity data and the fourth battery capacity data.

[0062] A training sample data acquisition unit is configured to obtain the training sample data based on the sample battery capacity feature data and a second target sample battery remaining life corresponding to the fourth battery capacity data of the sample battery.

[0063] Preferably, the preset model comprises an Ngboost model.

[0064] And / or, the battery capacity corresponding to the process from the beginning of the sample battery life to the end of the sample battery life decays from the rated capacity to 80% of the rated capacity.

[0065] And / or, the battery aging data set comprises the remaining life of the sample battery under the constant current charging and variable frequency discharging working condition.

[0066] Preferably, the preset voltage interval is within a full charge and full discharge voltage interval of the sample battery.

[0067] And / or, the target cycle interval comprises a target charging cycle interval or a target discharging cycle interval.

[0068] The fourth aspect further provides a battery residual life prediction system, and the system comprises:

[0069] An initial data acquisition module is configured to acquire initial charge-discharge data corresponding to an actual battery.

[0070] A target data acquisition module is configured to preprocess the initial charge-discharge data to obtain target battery capacity characteristic data.

[0071] A prediction module is configured to input the target battery capacity characteristic data into a battery residual life prediction model to output an actual residual life corresponding to the actual battery and a confidence degree and a confidence interval corresponding to the actual residual life.

[0072] The battery residual life prediction model is obtained based on the above-mentioned battery residual life prediction model construction system.

[0073] The fifth aspect further provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, the battery residual life prediction model construction method and / or the battery residual life prediction method are implemented.

[0074] The sixth aspect further provides a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, the battery residual life prediction model construction method and / or the battery residual life prediction method are implemented.

[0075] On the basis of common sense in the art, the above-mentioned preferred conditions can be combined arbitrarily, i.e., to obtain each preferred example of the present disclosure.

[0076] The positive progress effect of the present disclosure is that:

[0077] The battery residual life prediction and model construction method, system, device, and medium of the present disclosure obtain a target cycle number interval through a battery aging data set, preprocess the battery aging data set based on the target cycle number interval to obtain training sample data, input the training sample data into a preset model to train a battery residual life prediction model, predict an actual residual life corresponding to an actual battery and a confidence degree and a confidence interval corresponding to the actual residual life through the battery residual life prediction model. Only the target cycle number interval needs to be determined, and the training sample data is constructed according to the charge-discharge data corresponding to the target cycle number interval, which greatly simplifies the sample data, makes the entire model training simpler, and can accurately predict the actual residual life of the battery and the confidence degree and the confidence interval corresponding to the actual residual life, thereby realizing the prediction of the probability and distribution of the actual residual life. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1 A first flowchart of a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0079] Figure 2 A second flowchart of a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0080] Figure 3 A third flowchart of a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0081] Figure 4 A diagram of a correspondence between sample voltage data and a first target sample battery remaining life in a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0082] Figure 5 A fourth flowchart of a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0083] Figure 6 A first diagram of a degree of importance of each voltage in a preset voltage interval in a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0084] Figure 7 A second diagram of a degree of importance of each voltage in a preset voltage interval in a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0085] Figure 8 A diagram of a prediction result in a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0086] Figure 9 A diagram of a prediction result error in a method for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0087] Figure 10 A flowchart of a method for predicting a battery remaining life according to an embodiment of the present disclosure;

[0088] Figure 11 A structural diagram of a system for constructing a battery remaining life prediction model according to an embodiment of the present disclosure;

[0089] Figure 12 A structural diagram of a system for predicting a battery remaining life according to an embodiment of the present disclosure;

[0090] Figure 13 A structural diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0091] The present disclosure is further illustrated by the following examples without thereby limiting the present disclosure to the examples.

[0092] Example 1

[0093] The present embodiment provides a method for constructing a battery remaining life prediction model, as shown in the following. Figure 1 The method comprises the following steps:

[0094] S1, obtaining a battery aging data set corresponding to a plurality of sample batteries.

[0095] The battery aging data set comprises a sample remaining life and sample charge-discharge data corresponding to the sample battery from the beginning of the life to the end of the life.

[0096] S2, obtaining a target cycle number interval based on the battery aging data set.

[0097] S3, preprocessing the battery aging data set based on the target cycle number interval to obtain training sample data.

[0098] S4, inputting the training sample data into a preset model to obtain a battery remaining life prediction model.

[0099] The battery remaining life prediction model is used to predict the actual remaining life corresponding to the target battery capacity feature data of the input actual battery, and the confidence and confidence interval corresponding to the actual remaining life.

[0100] The confidence in the present embodiment is the probability corresponding to the actual remaining life, and the confidence interval is the distribution interval corresponding to the actual remaining life.

[0101] The battery remaining life is usually represented by the battery remaining charge-discharge cycle number, and the sample charge-discharge data generally corresponds to the charge-discharge curve data, which includes the charge-discharge cycle number of the sample battery from the beginning of the life to the end of the life, and the charge-discharge data (such as voltage, current, time, battery capacity, etc.) corresponding to each cycle number.

[0102] The target cycle number interval corresponds to the interval between two different charge cycle numbers or the interval between two different discharge cycle numbers; the training sample data is constructed according to the charge data of the two charge cycles or the discharge data of the two discharge cycles corresponding to the target cycle number interval.

[0103] Specifically, the process of the sample battery from the beginning of the life to the end of the life can correspond to the attenuation of the battery capacity from the rated capacity to 80% of the rated capacity.

[0104] The existing network model-based RUL prediction needs a large amount of battery aging data to construct sample data, and the sample data usually needs to cover data during the period from the beginning of the battery life to the end of the battery life, the data amount is huge, and the model training process is relatively complex.

[0105] The construction method of the battery remaining life prediction model of the embodiment only needs to determine the target cycle number interval, and constructs the training sample data according to the charge-discharge data corresponding to the target cycle number interval, greatly simplifies the sample data, makes the whole model training more simple, and can accurately predict the actual remaining life of the battery, and predict the confidence and confidence interval corresponding to the actual remaining life, realize the prediction of the probability and distribution of the actual remaining life.

[0106] In an optional embodiment, as shown in Figure 2 The step S2 includes:

[0107] S21, based on the sample charge-discharge data corresponding to the sample battery, the first battery capacity data corresponding to the first preset battery cycle number and the second battery capacity data corresponding to the second preset battery cycle number of the sample battery are obtained.

[0108] The second preset battery cycle number is greater than the first preset battery cycle number.

[0109] S22, the first target sample battery remaining life corresponding to the second battery capacity data of the sample battery is obtained.

[0110] S23, the target cycle number interval is calculated based on the first battery capacity data, the second battery capacity data and the first target sample battery remaining life.

[0111] Each battery cycle number has corresponding sample charge-discharge data, and the sample charge-discharge data includes battery capacity. The battery has a corresponding battery remaining life (i.e. RUL) at a certain battery capacity.

[0112] For example, the first preset battery cycle number of a sample battery is i, the first battery capacity data is Qi, the battery remaining life corresponding to the first battery capacity data Qi is Ri', the second preset battery cycle number is i+n, wherein n is the interval between the second preset battery cycle number and the first preset battery cycle number, and is a positive integer, the second battery capacity data is Qi+n, and the battery remaining life corresponding to the second battery capacity data Qi+n is Ri.

[0113] The selection of i is relatively flexible and can be selected according to the actual usage conditions of the sample battery. The value of i should be within the number of charge and discharge cycles of the sample battery from the beginning to the end of its life. The selection of n is also relatively flexible, but it should be ensured that i+n is within the number of charge and discharge cycles of the sample battery from the beginning to the end of its life.

[0114] The first target sample battery remaining life of the sample battery is Ri, and the target cycle number interval N can be calculated according to the first battery capacity data Qi, the second battery capacity data Qi+n and the first target sample battery remaining life Ri.

[0115] In an optional embodiment, if Figure 3 As shown, the above step S23 includes:

[0116] S231: Calculate a capacity difference between the second battery capacity data and the first battery capacity data.

[0117] S232. Take the logarithm of the variance of the capacity difference to obtain sample voltage data.

[0118] S233 , using a Pearson correlation coefficient method, based on the sample voltage data and the remaining life of the first target sample battery, calculate a target cycle interval.

[0119] Battery capacity decay is a slow process. The target cycle interval is the optimal cycle interval. The target cycle interval needs to ensure that there is enough cycle data to effectively predict RUL, and this cycle interval should be shortened as much as possible.

[0120] For example, if the first preset battery cycle number of a sample battery is i, the first battery capacity data is Qi, and the second battery capacity data is Qi+n, then the sample voltage data Li=log(Var(Q i+n -Q i ); the remaining life of the first target sample battery is Ri; the sample voltage data and the remaining life of the first target sample battery are used as variables to calculate their Pearson correlation coefficient r. Since i and n can be flexibly selected, for a given i, as n increases, the correlation r also increases. When |r|>0.7, the obtained n is considered to be the optimal cycle interval point, that is, the target cycle interval N corresponding to the entire sample battery is obtained. The correlation formula is as follows:

[0121]

[0122] in, Represents the mean value of the sample voltage data corresponding to the entire sample battery, Indicates the mean value of the remaining life of the sample batteries corresponding to the entire sample battery.

[0123] For example, if the number of charge-discharge cycles of the sample battery during the period from the beginning of the life to the end of the life is 1-2000 cycles, the value of i can be 1-10, the value of n can be 60-100, and the finally calculated target cycle interval N can be 90.

[0124] In a specific example, i is 10, r calculated is -0.879, 0.7<|r|≤0.9 represents strong correlation, and N is 90, that is, the first preset battery cycle number is the 10th cycle, and the second preset battery cycle number is the 100th cycle, indicating that the charging or discharging data between the 10th and 100th cycles is taken.

[0125] Figure 4 A schematic diagram of the corresponding relationship between the sample voltage data and the first target sample battery remaining life, wherein the abscissa represents the sample voltage data, the ordinate represents the first target sample battery remaining life, different sample voltage data has corresponding first target sample battery remaining life, and in Figure 4 , Pearson correlation coefficient r=-0.877, indicating that the target cycle interval and the battery remaining life have strong correlation, reflecting the rationality of the target cycle interval, and the subsequent battery remaining life can be accurately predicted.

[0126] The construction method of the battery remaining life prediction model of the embodiment can accurately calculate the target cycle interval through the Pearson correlation coefficient method, and then construct the training sample data, greatly simplify the sample data, make the whole model training more simple, and accurately predict the actual remaining life of the battery, and predict the confidence and confidence interval corresponding to the actual remaining life, realize the prediction of the probability and distribution of the actual remaining life.

[0127] In an optional embodiment, as shown in Figure 5 , the above step S3 comprises:

[0128] S31, obtain the preset voltage interval, the third preset battery cycle number and the fourth preset battery cycle number corresponding to the sample battery.

[0129] The fourth preset battery cycle number is obtained based on the sum of the third preset battery cycle number and the target cycle interval.

[0130] S32, divide the preset voltage interval based on the preset voltage interval, and obtain the third battery capacity data and the fourth battery capacity data corresponding to the divided preset voltage interval.

[0131] The third battery capacity data is capacity data of the sample battery corresponding to a third preset battery cycle number, and the fourth battery capacity data is capacity data of the sample battery corresponding to a fourth preset battery cycle number.

[0132] S33, based on the third battery capacity data and the fourth battery capacity data, obtaining sample battery capacity feature data of a plurality of preset dimensions of the sample battery.

[0133] S34, based on the sample battery capacity feature data and the second target sample battery remaining life corresponding to the fourth battery capacity data of the sample battery, obtaining training sample data.

[0134] In a specific example, the preset voltage interval is 2.5V-3.3V (volts), and the preset voltage interval is 1mv. According to 1mv, 2.5V-3.3V is divided into 800 voltage intervals, that is, the preset dimension is 800; if the target cycle number interval N is 90, the third preset battery cycle number is the 10th cycle, and the fourth preset battery cycle number is 10+90=100, that is, the 100th cycle. For the bth sample battery, the third battery capacity data is Q10, and the fourth battery capacity data is Q100. According to the 800 voltage intervals, 800 Q100-Q10 can be obtained. After vectorization processing of 800 Q100-Q10, sample battery capacity feature data x with 800 dimensions is obtained. b The fourth battery capacity data Q100 of the sample battery corresponds to the second target sample battery remaining life, which is marked as y b , the bth training sample (x b , y b ) is constructed, where b=1, 2, …, p, p is the total number of sample batteries, and a training sample data set consisting of p sample batteries is obtained, that is, the training sample data.

[0135] If the third preset battery cycle number is the first cycle, the fourth preset battery cycle number is 1+90=91.

[0136] The third preset battery cycle number can be flexibly selected, and its value should be in the charge and discharge cycle number during the period from the beginning of the sample battery life to the end of the sample battery life. The fourth preset battery cycle number is obtained based on the sum of the third preset battery cycle number and the target cycle number interval, that is, the cycle interval between the fourth preset battery cycle number and the third preset battery cycle number is the target cycle number interval.

[0137] The preset dimension is determined based on the preset voltage interval and the preset voltage interval.

[0138] The sample battery can be a battery with the same specification parameters, and the battery with the same specification parameters corresponds to the same charge-discharge voltage interval and the same target cycle number interval N.

[0139] The battery with different specification parameters has different target cycle number intervals N due to the influence of the manufacturing process and the specification parameters, and the corresponding charge-discharge voltage interval should be within the preset voltage interval; but the target cycle number interval N can be calculated by the above method, and the target cycle number interval N at this time can be the average of a plurality of target cycle number intervals N.

[0140] The above data is only exemplary and does not limit the scope of protection of the present disclosure, and those skilled in the art can select flexibly according to actual needs.

[0141] The construction method of the battery remaining life prediction model of the embodiment can accurately calculate the target cycle number interval by the Pearson correlation coefficient method, process the preset voltage interval by the preset voltage interval, the preset voltage interval and the target cycle number interval, obtain a plurality of sample battery capacity feature data corresponding to the preset dimension, and then based on the sample battery capacity feature data and the second target sample battery remaining life, construct a training sample data set. On the basis of simplifying the sample data, the multi-dimensional information of the sample data is ensured, which greatly simplifies the sample data, makes the whole model training simpler, and at the same time can accurately predict the actual remaining life of the battery, and predict the confidence and confidence interval corresponding to the actual remaining life, realize the prediction of the probability and distribution of the actual remaining life.

[0142] In an optional embodiment, the preset model includes but is not limited to an Ngboost model.

[0143] The construction method of the battery remaining life prediction model of the present disclosure will be discussed in detail below taking the preset model as an Ngboost model.

[0144] The bth training sample is (x b ,y b ), where b=1, 2, …, p, p is the total number of sample batteries, and the training sample data composed of p sample batteries is The parameters involved in the model training process are: the total number of training iterations M, the learning rate v, the probability distribution P θ (y) with parameters θ, and the basic learning machine h, θ={μ,σ} represents a group of parameters of the normal distribution, where μ represents the mean of the normal distribution, and σ represents the variance of the normal distribution. The most commonly used scoring rule is the log score S, and we get the maximum likelihood estimation (MLE) when the log score S is minimized.

[0145] where S(0, y) = -log P θ (y), y is the actual remaining life predicted by the battery remaining life prediction model.

[0146] Initialize 0 as marginal value, similar to initialize as mean, the goal is to find a set of optimal parameters 0, which satisfies the following formula:

[0147]

[0148] Assume that we have M rounds of training, for the mth round, m = 1, 2,..., M, for each training sample (x b , y b ), calculate the natural gradient g bm , where b = 1, 2,..., p, where:

[0149]

[0150] where F is the Fisher information, and g bm is the prediction, the iteration m trains a set of base learners h m (x), and at the same time, the weight p m of the weak learner is calculated, where:

[0151]

[0152] Then update the model to get 0 m = 0 m-1 + v p m h m (x); here the learning rate v is also called step size, which is used to prevent model overfitting.

[0153] Get the final battery remaining life prediction model:

[0154]

[0155] where y is the actual remaining life predicted by the battery remaining life prediction model, x is the target battery capacity feature data corresponding to the actual battery, since 0 = {m, s} represents a set of parameters of normal distribution, where m represents the mean of normal distribution, s represents the variance of normal distribution, therefore, the confidence and confidence interval of the actual remaining life can be known based on 0.

[0156] In addition, Ngboost is based on tree model, based on node purity improvement (Gain) for calculation, which can quantify the importance of input features.

[0157] Suppose we have a regression tree model, which contains multiple nodes. For a certain feature j, we can calculate its Gain at a certain node t as:

[0158]

[0159] where:

[0160] Var b is the variance of the target variable (predicted value in regression problem) of all samples at node t.

[0161] Var a is the sum of the weighted average variance of the target variable of all samples at the two child nodes after splitting node t by feature j.

[0162] D is the total number of samples at node t.

[0163] Then, we can take the weighted average of the Gain value of each node by feature j to get the importance score of the feature in a single tree.

[0164] Finally, we average the importance scores of a single tree in all trees to get the final importance score of the feature. Through the above method, we quantify the importance of the sample battery capacity feature data of several preset dimensions of the sample battery, and map it to the corresponding preset voltage interval to know which feature has the greatest impact on the result and which voltage value is important. When determining the preset voltage interval, it is necessary to ensure that the preset voltage interval covers these important voltages.

[0165] If the preset voltage interval is 2.5V-3.3V and the preset dimension is 800, the 800-dimension sample battery capacity feature data is mapped to the 2.5V-3.3V voltage interval, as shown in Figure 6 The horizontal coordinate is the voltage, the left vertical coordinate is the feature importance, and the right vertical coordinate is the battery capacity. It can be seen that 3.3V is discharged at the beginning, and the difference between the discharge curve corresponding to the 10th cycle and the discharge curve corresponding to the 100th cycle is larger at 3.0V, and the largest is at 2.9V. It can be seen that the importance of each voltage in the preset voltage interval.

[0166] Differential analysis is often used to explain the battery aging mechanism, which can effectively analyze the phase change process of lithium ion insertion or extraction from the electrode. The commonly used methods are incremental capacity analysis (ICA) and differential capacity analysis (DCA). Due to the difference value processing of voltage in the process of battery data processing, the local capacity interval voltage change is 0, so we express the difference form as follows:

[0167]

[0168] The tree model is used to quantify the importance of the features, which helps to determine which voltage position has a more important influence on the prediction of the RUL in the entire discharge process. The importance of the input features corresponding to the voltage position is mapped to the original discharge curve. At the same time, since the dv / dq curve or the feature referred to as the IC (incremental capacity) curve is particularly crucial for determining the battery health state, the mapping relationship is also performed on the IC curve; as shown in Figure 7

[0169] The construction method of the battery remaining life prediction model of the embodiment is based on the Ngboost model, and the battery remaining life prediction model is trained. Since the Ngboost model uses the conditional probability P(y|x) as the optimization target, instead of the traditional optimal solution E(y|x) as the optimization target, compared with various algorithm models based on traditional data driving, the Ngboost model can give the expectation and the corresponding confidence interval of the confidence interval, and can give a relatively accurate result under the condition of less data. In addition, the model can provide the importance ranking of the input features, and can provide strong support for the battery degradation mechanism analysis.

[0170] Figure 8 The actual remaining life predicted by the battery remaining life prediction model of the present disclosure is shown, and compared with the actual remaining life. The horizontal coordinate is the battery number in the test set, the vertical coordinate is the actual remaining life of the battery, the diamond box represents the actual remaining life predicted by the battery remaining life prediction model, and x(true) represents the actual remaining life. The battery remaining life prediction model can not only predict the actual remaining life RUL of the battery, but also give the distribution range of the RUL. It can be seen that most of the x in the test set falls within the corresponding diamond box, that is, the actual remaining life of most of the batteries falls within the prediction range of the battery remaining life prediction model of the present disclosure, which reflects the accuracy of the prediction of the battery remaining life prediction model of the present disclosure.

[0171] The battery remaining life prediction model of the present disclosure can output the actual remaining life of the actual battery, and the probability of the actual life falling into a certain prediction interval, for example, the predicted actual remaining life of a certain battery is 1000 cycles, and it has a 95% probability of falling into 900 cycles to 1100 cycles.

[0172] Figure 9 ​The battery remaining life prediction model of the present disclosure is compared with the existing GBDT (Gradient Boosting Decision Tree) model and the existing ElasticNET (Elastic Net) model in terms of the error of the predicted actual remaining life. Figure 9 In the figure, the abscissa represents the battery number in the test set, the ordinate represents the actual remaining life of the battery, Ngboost_Error represents the single error of the actual remaining life predicted by the battery remaining life prediction model of the present disclosure, Ngboost_Mean_Error represents the average error of the actual remaining life predicted by the battery remaining life prediction model of the present disclosure; GBDT_Error represents the single error of the actual remaining life predicted by the GBDT model, GBDT_Mean_Error represents the average error of the actual remaining life predicted by the GBDT model; ElasticNET_Error represents the single error of the actual remaining life predicted by the ElasticNET model, ElasticNET_Mean_Error represents the average error of the actual remaining life predicted by the ElasticNET model. It can be seen from the comparison that the battery remaining life prediction model of the present disclosure reduces the RUL average error to below 80 cycles, and the corresponding average error ratio is about 9%, which is 47.3% higher in accuracy than the ElasticNET model and 20% higher in accuracy than the GBDT model.

[0173] The preset model of the present disclosure can also be other models as long as it can realize numerical prediction and prediction of the confidence and confidence interval corresponding to the numerical value.

[0174] In an optional embodiment, the battery aging data set includes the remaining life of the sample battery under the constant current charging and variable frequency discharging working condition.

[0175] The battery aging data set can be constructed according to the real use scenario. In the field of electric vehicles, the working condition is stable when the electric vehicle is charging, and the working condition is complex during discharging, that is, the working condition of the lithium battery is mainly constant current charging and variable frequency discharging. For example, when constructing the battery aging data set, the charging process uses constant current 0.5C charging, and the discharging can be performed using different rates and different steps, so as to accurately predict the remaining life of the lithium battery of the electric vehicle.

[0176] In an optional embodiment, the preset voltage interval is within the full charging and full discharging voltage interval of the sample battery.

[0177] If the sample battery is a battery with the same specification parameters, it corresponds to the same full charge and full discharge voltage, at this time, it should be ensured that the preset voltage interval is within the full charge and full discharge voltage interval of the sample battery. If the sample battery is a battery with different specification parameters, it corresponds to different full charge and full discharge voltages, at this time, it should also be ensured that the preset voltage interval is within the full charge and full discharge voltage interval of the sample battery. Since the full charge and full discharge voltage interval is particularly important for predicting the remaining life of the battery, when determining the preset voltage interval, it is necessary to ensure that the preset voltage interval is within the full charge and full discharge voltage interval.

[0178] In an optional embodiment, the target cycle number interval includes a target charge cycle number interval or a target discharge cycle number interval.

[0179] The battery aging data set can correspond to the charging condition of the sample battery, at this time, the target cycle number interval corresponds to the target charge cycle number interval, and the training sample data is constructed according to the data of the sample battery during the charging period, so as to predict the actual remaining life of the actual battery.

[0180] The battery aging data set can correspond to the discharging condition of the sample battery, at this time, the target cycle number interval corresponds to the target discharge cycle number interval, and the training sample data is constructed according to the data of the sample battery during the discharging period, so as to predict the actual remaining life of the actual battery.

[0181] Embodiment 2

[0182] The embodiment provides a battery remaining life prediction method, as shown in Figure 10 The method comprises:

[0183] S5, obtaining initial charge and discharge data corresponding to the actual battery.

[0184] S6, preprocessing the initial charge and discharge data to obtain target battery capacity feature data.

[0185] S7, inputting the target battery capacity feature data into the battery remaining life prediction model to output the actual remaining life corresponding to the actual battery, and the confidence and confidence interval corresponding to the actual remaining life.

[0186] The battery remaining life prediction model is obtained based on the construction method of the battery remaining life prediction model in embodiment 1.

[0187] Specifically, the same preprocessing method as the battery aging data set in the model construction process in embodiment 1 is used to preprocess the initial charge and discharge data corresponding to the actual battery, so as to obtain target battery capacity feature data of a plurality of preset dimensions.

[0188] For example, the preset dimension sample battery capacity feature data in the model construction process corresponds to 800 dimensions, and the preset dimension target battery capacity feature data in the embodiment also corresponds to 800 dimensions, so as to ensure that the input data in the model construction process and the model using process correspond to each other.

[0189] The battery remaining life prediction method of the embodiment is based on the model constructed by the battery remaining life prediction model construction method in embodiment 1 to predict the battery remaining life, can accurately predict the actual remaining life of the battery, and predict the confidence and confidence interval corresponding to the actual remaining life, and realize the prediction of the probability and distribution of the actual remaining life.

[0190] Embodiment 3

[0191] The embodiment provides a battery remaining life prediction model construction system, as shown in Figure 11 The construction system comprises:

[0192] A data set acquisition module 1 is configured to acquire battery aging data sets corresponding to a plurality of sample batteries.

[0193] The battery aging data set comprises sample remaining life and sample charge-discharge data corresponding to the sample batteries from the beginning of the life to the end of the life.

[0194] A cycle interval acquisition module 2 is configured to acquire a target cycle interval based on the battery aging data set.

[0195] A sample data acquisition module 3 is configured to preprocess the battery aging data set based on the target cycle interval to acquire training sample data.

[0196] A model training module 4 is configured to input the training sample data into a preset model to train a battery remaining life prediction model.

[0197] The battery remaining life prediction model is configured to predict the actual remaining life and the confidence and confidence interval corresponding to the actual remaining life according to the input target battery capacity feature data of the actual battery.

[0198] In an optional embodiment, the cycle interval acquisition module 2 comprises:

[0199] A first capacity data acquisition unit 21 is configured to acquire first battery capacity data corresponding to a first preset battery cycle number and second battery capacity data corresponding to a second preset battery cycle number based on the sample charge-discharge data of the sample battery.

[0200] The second preset battery cycle number is greater than the first preset battery cycle number.

[0201] The remaining life obtaining unit 22 is configured to obtain a first target sample battery remaining life corresponding to the second battery capacity data of the sample battery.

[0202] The cycle interval obtaining unit 23 is configured to calculate a target cycle interval based on the first battery capacity data, the second battery capacity data and the first target sample battery remaining life.

[0203] In an optional embodiment, the cycle interval obtaining unit 23 is configured to calculate a capacity difference value of the second battery capacity data and the first battery capacity data; take a logarithm of a variance of the capacity difference value to obtain sample voltage data; and calculate the target cycle interval based on the sample voltage data and the first target sample battery remaining life by using a Pearson correlation coefficient method.

[0204] In an optional embodiment, the sample data obtaining module 3 comprises:

[0205] The preset voltage interval obtaining unit 31 is configured to obtain a preset voltage interval corresponding to the sample battery, a third preset battery cycle number and a fourth preset battery cycle number.

[0206] The fourth preset battery cycle number is obtained based on a sum of the third preset battery cycle number and the target cycle interval.

[0207] The second capacity data obtaining unit 32 is configured to divide the preset voltage interval based on the preset voltage interval, and obtain third battery capacity data and fourth battery capacity data corresponding to the divided preset voltage interval.

[0208] The third battery capacity data is capacity data of the sample battery corresponding to the third preset battery cycle number, and the fourth battery capacity data is capacity data of the sample battery corresponding to the fourth preset battery cycle number.

[0209] The feature data obtaining unit 33 is configured to obtain sample battery capacity feature data of a plurality of preset dimensions of the sample battery based on the third battery capacity data and the fourth battery capacity data.

[0210] The training sample data obtaining unit 34 is configured to obtain training sample data based on the sample battery capacity feature data and a second target sample battery remaining life corresponding to the fourth battery capacity data of the sample battery.

[0211] In an optional embodiment, the preset model comprises an Ngboost model.

[0212] In an optional embodiment, the process of the plurality of sample batteries from the beginning of the life to the end of the life corresponds to the battery capacity from the rated capacity to 80% of the rated capacity.

[0213] In an optional embodiment, the battery aging data set comprises the remaining life of the sample battery under the condition of constant current charging and variable frequency discharging.

[0214] In an optional embodiment, the preset voltage interval is within the full charge full discharge voltage interval of the sample battery.

[0215] In an optional embodiment, the target cycle number interval comprises a target charge cycle number interval or a target discharge cycle number interval.

[0216] The construction system of the battery remaining life prediction model of the present embodiment corresponds to the construction method of the battery remaining life prediction model of embodiment 1, and the working principle of the construction system is the same as that of the construction method. Therefore, the specific content will not be repeated here.

[0217] The construction system of the battery remaining life prediction model of the present embodiment obtains the target cycle number interval through the battery aging data set, pre-processes the battery aging data set based on the target cycle number interval, obtains the training sample data, inputs the training sample data into the preset model, trains to obtain the battery remaining life prediction model, and predicts the actual remaining life corresponding to the actual battery and the confidence and confidence interval corresponding to the actual remaining life. Only the target cycle number interval needs to be determined, and the training sample data is constructed according to the charge-discharge data corresponding to the target cycle number interval, which greatly simplifies the sample data, makes the whole model training more simple, and at the same time can accurately predict the actual remaining life of the battery and the confidence and confidence interval corresponding to the actual remaining life, and realizes the prediction of the probability and distribution of the actual remaining life.

[0218] Embodiment 4

[0219] The present embodiment provides a battery remaining life prediction system, as shown in Figure 12 The system comprises:

[0220] An initial data acquisition module 5 is configured to acquire initial charge-discharge data corresponding to an actual battery.

[0221] A target data acquisition module 6 is configured to pre-process the initial charge-discharge data to obtain target battery capacity feature data.

[0222] A prediction module 7 is configured to input the target battery capacity feature data into a battery remaining life prediction model to output actual remaining life corresponding to the actual battery, and confidence and confidence interval corresponding to the actual remaining life.

[0223] The battery remaining life prediction model is constructed based on the construction system of the battery remaining life prediction model of embodiment 3.

[0224] The battery residual life prediction system of the embodiment corresponds to the battery residual life prediction method of embodiment 2, and the working principle of the battery residual life prediction system is the same as that of the battery residual life prediction method, which will not be repeated here.

[0225] The battery residual life prediction system of the embodiment is based on the model constructed by the battery residual life prediction model construction system in embodiment 3, and can accurately predict the actual residual life of the battery, and predict the confidence and confidence interval corresponding to the actual residual life, thereby realizing the prediction of the probability and distribution of the actual residual life.

[0226] Embodiment 5

[0227] The embodiment provides an electronic device, Figure 13 A structural schematic diagram of an electronic device provided by the embodiment is shown. The electronic device includes a memory, a processor, and a computer program stored in the memory and used for executing on the processor. When the processor executes the program, the construction method of the battery residual life prediction model in embodiment 1 and / or the battery residual life prediction method in embodiment 2 is realized. Figure 13 The electronic device 80 shown is only an example and should not limit the functions and use range of the embodiments of the present disclosure.

[0228] As Figure 13 shown, the electronic device 80 can be in the form of a general computing device, for example, it can be a server device. The components of the electronic device 80 can include but are not limited to: the above-mentioned at least one processor 81, the above-mentioned at least one memory 82, a bus 83 connecting different system components including the memory 82 and the processor 81.

[0229] The bus 83 includes a data bus, an address bus, and a control bus.

[0230] The memory 82 can include volatile memory, such as random access memory (RAM) 821 and / or cache memory 822, and can further include read-only memory (ROM) 823.

[0231] The memory 82 can further include programs / utilities 825 having a set of (at least one) program modules 824, such as operating systems, one or more application programs, other program modules, and program data, each of which or some combination of which can include the implementation of a network environment.

[0232] The processor 81 performs various function applications and data processing by executing the computer programs stored in the memory 82, such as the construction method of the battery remaining life prediction model in Embodiment 1 and / or the battery remaining life prediction method in Embodiment 2.

[0233] The electronic device 80 can also communicate with one or more external devices 84 (such as a keyboard, a pointing device, etc.) via an input / output (I / O) interface 85. Further, the model generation device 80 can communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 86. As Figure 13 illustrated, the network adapter 86 communicates with the other modules of the model generation device 80 via the bus 83. It should be appreciated that although not shown, other hardware and / or software modules could be used in conjunction with the model generation device 80, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID (Redundant Array of Independent Disks) systems, tape drives, and data archival storage systems, etc.

[0234] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the foregoing detailed description, such a division is merely exemplary and not mandatory. Indeed, according to an implementation of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into a plurality of units / modules.

[0235] Embodiment 6

[0236] The present embodiment provides a computer readable storage medium, having stored thereon a computer program, the program being executed by a processor to implement the steps in the construction method of the battery remaining life prediction model in Embodiment 1 and / or the steps in the battery remaining life prediction method in Embodiment 2.

[0237] More specifically, the readable storage medium can include, but is not limited to, a portable disc, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0238] In a possible implementation, the present disclosure can also be implemented in the form of a program product, which includes program codes for causing a terminal device to perform the steps in the construction method of the battery remaining life prediction model in Embodiment 1 or the steps in the battery remaining life prediction method in Embodiment 2 when the program product is executed on the terminal device.

[0239] programmable logic arrays, field programmable gate arrays, processors, controllers, micro-controllers, microprocessors, other electronic digital turbines, or similar compute logic devices. The program code can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's device, partly on the user's device, as a stand-alone software package, partly on the user's device and partly on a remote device or entirely on the remote device. In the latter scenario, the remote device can be connected to the user's device through any type of network, including a local area network ("LAN"), a wide area network ("WAN"), the Internet, or the like.

[0240] Although the specific embodiments of the present disclosure have been described above, those skilled in the art will understand that the present disclosure is only illustrative and the scope of protection of the present disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to the embodiments without departing from the principles and essence of the present disclosure, and such changes and modifications fall within the scope of protection of the present disclosure.

Claims

1. A method for constructing a battery remaining life prediction model, characterized in that: The construction method comprises: Obtaining a battery aging data set corresponding to a number of sample batteries; the battery aging data set includes the corresponding sample remaining life and sample charge and discharge data of the sample batteries from the beginning to the end of their life; Obtaining a target cycle interval based on the battery aging dataset; Preprocessing the battery aging dataset based on the target cycle interval to obtain training sample data; Inputting the training sample data into a preset model to train and obtain the battery remaining life prediction model; The battery remaining life prediction model is used to predict the corresponding actual remaining life, and the confidence level and confidence interval corresponding to the actual remaining life according to the input target battery capacity characteristic data of the actual battery.

2. The construction method according to claim 1, characterized in that The step of obtaining a target cycle interval based on the battery aging data set includes: Based on the sample charge and discharge data corresponding to the sample battery, obtaining first battery capacity data corresponding to a first preset number of battery cycles and second battery capacity data corresponding to a second preset number of battery cycles of the sample battery; Wherein, the second preset number of battery cycles is greater than the first preset number of battery cycles; Obtaining a first target sample battery remaining life corresponding to the second battery capacity data of the sample battery; The target cycle interval is calculated based on the first battery capacity data, the second battery capacity data, and the first target sample battery remaining life.

3. The construction method according to claim 2, characterized in that The step of calculating the target cycle interval based on the first battery capacity data, the second battery capacity data and the first target sample battery remaining life includes: Calculating a capacity difference between the second battery capacity data and the first battery capacity data; Taking the logarithm of the variance of the capacity difference to obtain sample voltage data; The target cycle interval is calculated based on the sample voltage data and the first target sample battery remaining life using a Pearson correlation coefficient method.

4. The construction method according to claim 2, characterized in that The step of preprocessing the battery aging data set based on the target cycle interval to obtain training sample data includes: Obtaining a preset voltage range, a third preset battery cycle number, and a fourth preset battery cycle number corresponding to the sample battery; The fourth preset battery cycle number is obtained based on the sum of the third preset battery cycle number and the target cycle number interval; Dividing the preset voltage interval based on a preset voltage interval, and acquiring third battery capacity data and fourth battery capacity data corresponding to the divided preset voltage interval; The third battery capacity data is the capacity data of the sample battery corresponding to the third preset battery cycle number, and the fourth battery capacity data is the capacity data of the sample battery corresponding to the fourth preset battery cycle number; Based on the third battery capacity data and the fourth battery capacity data, obtaining sample battery capacity characteristic data of several preset dimensions of the sample battery; The training sample data is obtained based on the sample battery capacity characteristic data and the second target sample battery remaining life corresponding to the fourth battery capacity data of the sample battery.

5. The construction method according to claim 1, characterized in that The preset model includes an Ngboost model; And / or, the process from the beginning to the end of the life of the several sample batteries corresponds to the battery capacity decaying from the rated capacity to 80% of the rated capacity; And / or, the battery aging data set includes the remaining life of the sample battery under constant current charging and variable frequency discharging conditions.

6. The construction method according to claim 4, characterized in that The preset voltage range is within the full charge and discharge voltage range of the sample battery; And / or, the target cycle number interval includes a target charge cycle number interval or a target discharge cycle number interval.

7. A method for predicting remaining battery life, characterized in that: The method comprises: Obtain the initial charge and discharge data corresponding to the actual battery; Preprocessing the initial charge and discharge data to obtain target battery capacity characteristic data; Inputting the target battery capacity characteristic data into a battery remaining life prediction model to output an actual remaining life corresponding to the actual battery, and a confidence level and confidence interval corresponding to the actual remaining life; The battery remaining life prediction model is obtained based on the method for constructing the battery remaining life prediction model according to any one of claims 1 to 6.

8. A system for constructing a battery remaining life prediction model, characterized in that: The build system includes: A data set acquisition module is used to acquire battery aging data sets corresponding to a number of sample batteries; the battery aging data sets include the remaining service life and sample charge and discharge data of the corresponding sample batteries from the beginning to the end of their life; A cycle interval acquisition module, configured to acquire a target cycle number interval based on the battery aging dataset; a sample data acquisition module, which pre-processes the battery aging data set based on the target cycle number interval to obtain training sample data; A model training module, configured to input the training sample data into a preset model to train and obtain the battery remaining life prediction model; The battery remaining life prediction model is used to predict the corresponding actual remaining life, and the confidence level and confidence interval corresponding to the actual remaining life according to the input target battery capacity characteristic data of the actual battery.

9. A battery remaining life prediction system, characterized in that: The system comprises: An initial data acquisition module is used to obtain initial charge and discharge data corresponding to the actual battery; a target data acquisition module, configured to pre-process the initial charge and discharge data to obtain target battery capacity characteristic data; A prediction module, configured to input the target battery capacity characteristic data into a battery remaining life prediction model to output an actual remaining life corresponding to the actual battery, and a confidence level and confidence interval corresponding to the actual remaining life; The battery remaining life prediction model is obtained based on the battery remaining life prediction model construction system as described in claim 8.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and configured to run on the processor, wherein: When the processor executes the computer program, it implements the method for constructing a battery remaining life prediction model as described in any one of claims 1 to 6, and / or the battery remaining life prediction method as described in claim 7.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When executed by a processor, the computer program implements the method for constructing a battery remaining life prediction model as described in any one of claims 1 to 6, and / or the battery remaining life prediction method as described in claim 7.

Images