A lithium battery soh estimation method based on polynomial reconstruction charge-discharge curve

Abstract

The application discloses a lithium battery SOH estimation method based on a polynomial reconstruction charge-discharge curve, and the method comprises the following steps: obtaining a lithium battery aging data set; data preprocessing; training and obtaining an SOH estimation model; training and obtaining a charge-discharge curve reconstruction model; reconstructing a missing charge-discharge curve; and online estimating the SOH of the lithium battery. The constant-current charging voltage curve, the constant-voltage charging current curve and the constant-current discharging voltage curve are characterized by a polynomial fitting function, and a corresponding charge-discharge curve reconstruction model is constructed based on the coefficients of the polynomial fitting function. When a kind of curve appears a serious data missing condition, the polynomial coefficient of the polynomial fitting function of the missing curve can be estimated through the charge-discharge curve reconstruction model, and the reconstructed curve is further characterized by the function curve corresponding to the polynomial fitting function, so as to complete the reconstruction of the missing curve.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of lithium batteries, in particular to a lithium battery SOH estimation method based on polynomial reconstruction of charge-discharge curves. BACKGROUND

[0002] Lithium batteries have become the mainstream power supply solution in electric vehicles, portable electronic devices and energy storage systems. The state of health (SOH) of the battery, as a core parameter for measuring its aging and performance degradation, is of great significance for accurately evaluating the remaining life and reliability of the battery.

[0003] Today's data-driven lithium battery SOH estimation method is a research hotspot. In the data-driven lithium battery SOH estimation method, researchers extract a variety of health features from the current-voltage curves at various stages for lithium battery SOH estimation. The most commonly used current-voltage curves include constant-current charging voltage curves, constant-voltage charging current curves and constant-current discharging voltage curves. However, in real applications, sensor failures, data transmission distortion problems and other problems can cause serious missing problems in the current curve or voltage curve at a certain stage, thereby preventing the extraction of complete and effective health features and affecting the SOH estimation accuracy. To solve this problem, researchers have proposed a curve reconstruction method based on neural networks. The complete curve is directly used as the input of the neural network, and the missing curve is used as the output of the neural network, thereby completing the reconstruction of the missing curve. However, this method results in a large input and output dimension of the neural network, requires high accuracy of the neural network, is more susceptible to noise interference, and the high output dimension can cause the neural network to ignore the reconstruction accuracy of some data points in the missing curve. SUMMARY

[0004] The present application overcomes the problems existing in the above-mentioned curve reconstruction method based on neural networks, and provides a lithium battery SOH estimation method based on polynomial reconstruction of charge-discharge curves.

[0005] The technical scheme of the present application is as follows:

[0006] S1: Obtain a lithium battery aging data set, the specific steps including: performing a cyclic charge-discharge experiment on a plurality of same type lithium batteries in a constant-current-constant-voltage charging and constant-current discharging manner; in the charge-discharge cycle, collecting constant-current charging voltage curves, constant-voltage charging current curves and constant-current discharging voltage curves, and the collected data should include the time, current value and voltage value corresponding to each sampling point; recording the SOH value corresponding to each cycle; and constructing a data set D with the constant-current charging voltage curves, constant-voltage charging current curves, constant-current discharging voltage curves and SOH values;

[0007] S2: data preprocessing, that is, first data cleaning is performed on the data in the data set D described in step S1, and then time adjustment is performed on the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the data set, and the specific steps of the time adjustment are as follows: the time of all sampling points of each stage curve is subtracted from the time of the first sampling point of the curve of the stage;

[0008] S3: training and obtaining an SOH estimation model, that is, first the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the data set D after data preprocessing described in step S2 are extracted for health features, and then the data set is divided into a training set and a test set; then an SOH estimation model based on an artificial neural network is constructed, all the extracted health features are used as the input of the artificial neural network model, and SOH is used as the output of the artificial neural network model, the training set is used to train the artificial neural network model, and the test set is used to evaluate the performance of the model; if the test accuracy requirement is met, the model is saved, otherwise the hyperparameters of the model are modified and retrained until the test accuracy requirement is met;

[0009] S4: training and obtaining a charge-discharge curve reconstruction model, that is, first a polynomial fitting function with the smallest fitting error is selected for each of the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the data set D described in step S1, then the three curves are fitted by using the corresponding polynomial fitting functions to obtain the polynomial coefficients corresponding to the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve; then three charge-discharge curve reconstruction models are constructed, the three charge-discharge curve reconstruction models include a constant current charging voltage curve reconstruction model, a constant voltage charging current curve reconstruction model and a constant current discharging voltage curve reconstruction model, and the three charge-discharge curve reconstruction models are all composed of multiple-input multiple-output multilayer fully connected layers; the polynomial coefficients corresponding to the curve to be reconstructed are sequentially taken as the output of the model, and the polynomial coefficients corresponding to the other two curves are taken as the input of the model, and the three charge-discharge curve reconstruction models are trained, and after the training is completed, the three charge-discharge curve reconstruction models are saved, and the specific steps are as follows:

[0010] S401: taking the constant current charging voltage curve of all samples in the data set D after data preprocessing described in step S2 as a fitting object, using a polynomial fitting function for fitting, taking the minimum average fitting error of all samples as the target, searching for the best polynomial degree, determining a polynomial fitting function with the minimum average fitting error, and taking the polynomial fitting function as the polynomial fitting function of the constant current charging voltage curve;

[0011] S402: Take the constant-voltage charging current curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant-voltage charging current curve;

[0012] S403: Take the constant-current discharge voltage curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant-current discharge voltage curve;

[0013] S404: The polynomial coefficient set of the polynomial fitting function of the constant-current charging voltage curve forms a data set A; the polynomial coefficient set of the polynomial fitting function of the constant-voltage charging current curve forms a data set B; and the polynomial coefficient set of the polynomial fitting function of the constant-current discharge voltage curve forms a data set C;

[0014] S405: Construct a constant-current charging voltage curve reconstruction model Model1, the model Model1 is a multi-input and multi-output multi-layer full connection layer model, data set B and data set C are taken as the input of the model, data set A is taken as the output of the model, thereby training the constant-current charging voltage curve reconstruction model Model1, and saving the model Model1 after training is completed;

[0015] S406: Construct a constant-voltage charging current curve reconstruction model Model2, the model Model2 is a multi-input and multi-output multi-layer full connection layer model, data set A and data set C are taken as the input of the model, data set B is taken as the output of the model, thereby training the constant-voltage charging current curve reconstruction model Model2, and saving the model Model2 after training is completed;

[0016] S407: Construct a constant-current discharge voltage curve reconstruction model Model3, the model Model3 is a multi-input and multi-output multi-layer full connection layer model, data set A and data set B are taken as the input of the model, data set C is taken as the output of the model, thereby training the constant-current discharge voltage curve reconstruction model Model3, and saving the model Model3 after training is completed;

[0017] S5: reconstruct the missing charge-discharge curve, the specific steps are: in the actual application occasion of estimating SOH, collect constant current charging voltage curve, constant voltage charging current curve and constant current discharging voltage curve, the battery management system further detects whether a curve in the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve appears serious data missing condition, the serious data missing condition refers to that there is long time, continuous multiple sampling period data loss in the collected curve data; if it is detected that a curve in the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve appears serious data missing condition, then the corresponding model in the constant current charging voltage curve reconstruction model, the constant voltage charging current curve reconstruction model and the constant current discharging voltage curve reconstruction model in step S4 is used to reconstruct the curve with serious data missing condition; if there is no serious data missing condition, then directly proceed to step S6;

[0018] S6: online estimate the SOH of lithium battery, extract the health characteristics of the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve respectively, input into the SOH estimation model based on artificial neural network in step S3, and output the SOH estimation value.

[0019] In the scheme, the specific steps of reconstructing the missing charge-discharge curve in step S5 are as follows:

[0020] Collect the constant current charging voltage curve, if the battery management system detects that the constant current charging voltage curve appears serious data missing condition, then first fit the collected constant voltage charging current curve and constant current discharging voltage curve according to the polynomial fitting function of the constant voltage charging current curve and the polynomial fitting function of the constant current discharging voltage curve obtained in step S4, obtain the corresponding polynomial coefficients; input the obtained polynomial coefficients into the constant current charging voltage curve reconstruction model obtained in step S4, output the polynomial coefficient set of the reconstructed constant current charging voltage curve Then further obtain the polynomial coefficient set The corresponding polynomial fitting function Poly cc , substitute the constant current charging cutoff voltage into the polynomial fitting function Poly cc to obtain the corresponding time t1, and the function curve segment corresponding to the argument range [0, t1] of the polynomial fitting function Poly cc is taken as the reconstructed constant current charging voltage curve.

[0021] If the battery management system detects that the constant-voltage charging current curve has serious data missing, the constant-current charging voltage curve and the constant-current discharging voltage curve collected are fitted according to the polynomial fitting function of the constant-current charging voltage curve and the polynomial fitting function of the constant-voltage charging current curve obtained in step S4, and corresponding polynomial coefficients are obtained; the obtained polynomial coefficients are input into the constant-voltage charging current curve reconstruction model obtained in step S4, and a polynomial coefficient set of the reconstructed constant-voltage charging current curve is output Then the polynomial coefficient set is further obtained The corresponding polynomial fitting function Poly cv The constant-voltage charging cutoff current is substituted into the polynomial fitting function Poly cv to obtain the corresponding time t2, and the function curve segment corresponding to the argument range [0, t2] of the polynomial fitting function Poly cv is taken as the reconstructed constant-voltage charging current curve.

[0022] The constant-current discharging voltage curve is collected, and if the battery management system detects that the constant-current discharging voltage curve has serious data missing, the constant-current charging voltage curve and the constant-voltage charging current curve collected are fitted according to the polynomial fitting function of the constant-current charging voltage curve and the polynomial fitting function of the constant-voltage charging current curve obtained in step S4, and corresponding polynomial coefficients are obtained; the obtained polynomial coefficients are input into the constant-current discharging voltage curve reconstruction model obtained in step S4, and a polynomial coefficient set of the reconstructed constant-current discharging voltage curve is output Then the polynomial coefficient set is further obtained The corresponding polynomial fitting function Poly dc The constant-current discharging cutoff voltage is substituted into the polynomial fitting function Poly dc to obtain the corresponding time t3, and the function curve segment corresponding to the argument range [0, t3] of the polynomial fitting function Poly dc is taken as the reconstructed constant-current discharging voltage curve.

[0023] Compared with the prior art, the beneficial effects of the technical scheme of the present application are:

[0024] The application provides a lithium battery SOH estimation method based on polynomial reconstruction of charge-discharge curves. In the face of serious missing of the charge-discharge curves, a constant-current charging voltage curve reconstruction model, a constant-voltage charging current curve reconstruction model and a constant-current discharging voltage curve reconstruction model are established. The three models are all neural networks composed of full connection layers, the output of the reconstruction model is the polynomial coefficient of the polynomial fitting function corresponding to the missing curve, and the input of the reconstruction model is the polynomial coefficient of the polynomial fitting function corresponding to the complete curve. When a certain charge-discharge curve in the constant-current charging voltage curve, the constant-voltage charging current curve and the constant-current discharging voltage curve is seriously missing in actual application, the polynomial coefficients of the corresponding polynomial fitting functions are obtained by polynomial fitting of the other two complete curves, the polynomial coefficients of the complete curve are input into the corresponding reconstruction model, the polynomial coefficients of the missing curve are output, and then the curve of the polynomial fitting function corresponding to the polynomial coefficients is used as the reconstructed curve. The method for reconstructing the curve through the polynomial can not only significantly reduce the complexity of the reconstruction model, but also can avoid the problem of unbalanced reconstruction accuracy of the data points in the missing curve by using the polynomial fitting function curve to represent the overall shape of the battery charge-discharge curve, and can reduce the interference of noise. In the case of serious data missing, the method can generate a smooth and accurate reconstructed curve. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0026] Figure 1 The steps of the lithium battery SOH estimation method based on polynomial reconstruction of charge-discharge curves. DETAILED DESCRIPTION

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0028] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, therefore, the protection scope of the present application is not limited by the specific embodiments disclosed below.

[0029] In one specific embodiment, as shown in Figure 1 A lithium battery SOH estimation method based on polynomial reconstruction of charge-discharge curves, comprising the following steps:

[0030] S1: Obtain a lithium battery aging dataset, the specific steps comprising: performing a cyclic charging and discharging experiment on a plurality of same type lithium batteries in a constant current-constant voltage charging and constant current discharging manner; in the charging and discharging cycle, collecting a constant current charging voltage curve, a constant voltage charging current curve and a constant current discharging voltage curve, the collected data including the time, current value and voltage value corresponding to each sampling point; recording the SOH value corresponding to each cycle; and constructing the constant current charging voltage curve, the constant voltage charging current curve, the constant current discharging voltage curve and the SOH value into a dataset D;

[0031] S2: data preprocessing, that is, first performing data cleaning on the dataset D described in step S1, and then performing time adjustment on the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the dataset, the specific steps of the time adjustment being: subtracting the time of the first sampling point of the curve in each stage from the time of all sampling points of the curve in the stage;

[0032] S3: training and obtaining an SOH estimation model, that is, first extracting health features from the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the dataset D after data preprocessing described in step S2, then dividing the dataset into a training set and a test set; then constructing an SOH estimation model based on an artificial neural network, taking all the extracted health features as the input of the artificial neural network model and taking SOH as the output of the artificial neural network model, training the artificial neural network model using the training set and evaluating the model performance using the test set; if the test accuracy requirement is met, saving the model, otherwise modifying the hyperparameters of the model and retraining until the test accuracy requirement is met;

[0033] S4: training and obtaining a charging and discharging curve reconstruction model, that is, first selecting a polynomial fitting function with the smallest fitting error for each of the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the dataset D described in step S1, then fitting the three curves with the corresponding polynomial fitting functions to obtain the polynomial coefficients corresponding to the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve; then constructing three charging and discharging curve reconstruction models, the three charging and discharging curve reconstruction models comprising a constant current charging voltage curve reconstruction model, a constant voltage charging current curve reconstruction model and a constant current discharging voltage curve reconstruction model, the three charging and discharging curve reconstruction models each being composed of a multi-input multi-output multi-layer fully connected layer; taking the polynomial coefficients corresponding to the curve to be reconstructed as the output of the model and the polynomial coefficients corresponding to the other two curves as the input of the model, training the three charging and discharging curve reconstruction models, and saving the three charging and discharging curve reconstruction models after training, the specific steps being as follows:

[0034] S401: Take the constant current charging voltage curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant current charging voltage curve;

[0035] S402: Take the constant voltage charging current curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant voltage charging current curve;

[0036] S403: Take the constant current discharging voltage curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant current discharging voltage curve;

[0037] S404: The polynomial coefficients of the polynomial fitting functions of all constant current charging voltage curves form a data set A; the polynomial coefficients of the polynomial fitting functions of all constant voltage charging current curves form a data set B; the polynomial coefficients of the polynomial fitting functions of all constant current discharging voltage curves form a data set C;

[0038] S405: Construct a constant current charging voltage curve reconstruction model Model1, the model Model1 is a multi-input and multi-output multi-layer full connection layer model, data set B and data set C are taken as the input of the model, data set A is taken as the output of the model, so as to train the constant current charging voltage curve reconstruction model Model1, and save the model Model1 after training;

[0039] S406: Construct a constant voltage charging current curve reconstruction model Model2, the model Model2 is a multi-input and multi-output multi-layer full connection layer model, data set A and data set C are taken as the input of the model, data set B is taken as the output of the model, so as to train the constant voltage charging current curve reconstruction model Model2, and save the model Model2 after training;

[0040] S407: Construct a constant current discharge voltage curve reconstruction model Model3, the model Model3 is a multi-input multi-output multi-layer full connection layer model, data set A and data set B are used as inputs of the model, and data set C is used as an output of the model, so as to train the constant current discharge voltage curve reconstruction model Model3, and save the model Model3 after training is completed;

[0041] S5: Reconstruct the missing charge-discharge curve, and the specific steps are as follows: in the actual application of estimating SOH, the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharge voltage curve are collected, and the battery management system further detects whether a curve in the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharge voltage curve appears a serious data missing condition, the serious data missing condition refers to that there is a long-time and continuous multiple sampling period data loss in the collected curve data; if it is detected that a curve in the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharge voltage curve appears a serious data missing condition, the corresponding model in the constant current charging voltage curve reconstruction model, the constant voltage charging current curve reconstruction model and the constant current discharge voltage curve reconstruction model in step S4 is used to reconstruct the curve with the serious data missing condition; if there is no serious data missing condition, step S6 is directly performed;

[0042] S6: On-line estimation of lithium battery SOH, the health characteristics of the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharge voltage curve are extracted and input into the SOH estimation model based on the artificial neural network in step S3, and the SOH estimation value is output.

[0043] The specific steps of reconstructing the missing charge-discharge curve in step S5 are as follows:

[0044] The constant current charging voltage curve is collected, if the battery management system detects that the constant current charging voltage curve has a continuous missing voltage interval of 0.2V and above, the constant voltage charging current curve and the constant current discharge voltage curve collected are first fitted according to the polynomial fitting function of the constant voltage charging current curve and the polynomial fitting function of the constant current discharge voltage curve obtained in step S4, and the corresponding polynomial coefficients are obtained; the obtained polynomial coefficients are input into the constant current charging voltage curve reconstruction model obtained in step S4, and the polynomial coefficient set of the reconstructed constant current charging voltage curve is output Then the polynomial coefficient set is further obtained The corresponding polynomial fitting function Poly cc The constant current charging cutoff voltage is substituted into the polynomial fitting function Poly cc , and the corresponding time t1 is obtained, and the polynomial fitting function Poly ccthe function curve segment corresponding to the argument range [0, t2] of the polynomial fitting function Poly

[0045] If the battery management system detects that the constant-voltage charging current curve has a continuous missing current interval of 0.3 A or more, the constant-current charging voltage curve and the constant-current discharging voltage curve collected are fitted according to the polynomial fitting functions of the constant-current charging voltage curve and the constant-voltage charging current curve obtained in step S4, to obtain corresponding polynomial coefficients; the obtained polynomial coefficients are input into the constant-voltage charging current curve reconstruction model obtained in step S4, and a polynomial coefficient set of the reconstructed constant-voltage charging current curve is output Then the polynomial coefficient set is further obtained The corresponding polynomial fitting function Poly cv The constant-voltage charging cutoff current is substituted into the polynomial fitting function Poly cv to obtain the corresponding time t2, and the function curve segment corresponding to the argument range [0, t2] of the polynomial fitting function Poly cv is taken as the reconstructed constant-voltage charging current curve.

[0046] If the battery management system detects that the constant-current discharging voltage curve has a continuous missing voltage interval of 0.2 V or more, the constant-current charging voltage curve and the constant-voltage charging current curve collected are fitted according to the polynomial fitting functions of the constant-current charging voltage curve and the constant-voltage charging current curve obtained in step S4, to obtain corresponding polynomial coefficients; the obtained polynomial coefficients are input into the constant-current discharging voltage curve reconstruction model obtained in step S4, and a polynomial coefficient set of the reconstructed constant-current discharging voltage curve is output Then the polynomial coefficient set is further obtained The corresponding polynomial fitting function Poly dc The constant-current discharging cutoff voltage is substituted into the polynomial fitting function Poly dc to obtain the corresponding time t3, and the function curve segment corresponding to the argument range [0, t3] of the polynomial fitting function Poly dc is taken as the reconstructed constant-current discharging voltage curve.

Claims

1. A lithium battery SOH estimation method based on a polynomial reconstruction charge-discharge curve, characterized by, The specific steps are as follows: S1: Obtain a lithium battery aging data set, the specific steps comprising: performing a cyclic charging and discharging experiment on a plurality of same type lithium batteries in a constant current-constant voltage charging and constant current discharging manner; in the charging and discharging cycle, collecting a constant current charging voltage curve, a constant voltage charging current curve and a constant current discharging voltage curve, the collected data including the time, current value and voltage value corresponding to each sampling point; recording the SOH value corresponding to each cycle; and constructing a data set D from the constant current charging voltage curve, the constant voltage charging current curve, the constant current discharging voltage curve and the SOH value; S2: data preprocessing, that is, first performing data cleaning on the data set D described in step S1, and then performing time adjustment on the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the data set, the specific steps of the time adjustment being: subtracting the time of the first sampling point of the curve in each stage from the time of all sampling points of the curve in the stage; S3: training and obtaining an SOH estimation model, that is, first extracting health features from the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the data set D after data preprocessing described in step S2, then dividing the data set into a training set and a test set; then constructing an SOH estimation model based on an artificial neural network, taking all the extracted health features as the input of the artificial neural network model and taking SOH as the output of the artificial neural network model, training the artificial neural network model using the training set and evaluating the model performance using the test set; if the test accuracy requirement is met, the model is saved, otherwise the hyperparameters of the model are modified and retrained until the test accuracy requirement is met; S4: training and obtaining a charging and discharging curve reconstruction model, that is, first selecting a polynomial fitting function with the smallest fitting error for each of the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve in the data set D described in step S1, then fitting the three curves with the corresponding polynomial fitting functions to obtain the polynomial coefficients corresponding to the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharging voltage curve; then constructing three charging and discharging curve reconstruction models, the three charging and discharging curve reconstruction models comprising a constant current charging voltage curve reconstruction model, a constant voltage charging current curve reconstruction model and a constant current discharging voltage curve reconstruction model, the three charging and discharging curve reconstruction models each being composed of a multi-input multi-output multi-layer fully connected layer; taking the polynomial coefficients corresponding to the curve to be reconstructed as the output of the model and the polynomial coefficients corresponding to the other two curves as the input of the model, training the three charging and discharging curve reconstruction models, and saving the three charging and discharging curve reconstruction models after training is completed, the specific steps being as follows: S401: Take the constant current charging voltage curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant current charging voltage curve; S402: Take the constant voltage charging current curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant voltage charging current curve; S403: Take the constant current discharging voltage curve of all samples in the data set D after data preprocessing in step S2 as the fitting object, use the polynomial fitting function for fitting, take the minimum average fitting error of all samples as the target, search for the best polynomial degree, determine a polynomial fitting function with the minimum average fitting error, and take the polynomial fitting function as the polynomial fitting function of the constant current discharging voltage curve; S404: The polynomial coefficients of the polynomial fitting functions of all constant current charging voltage curves form a data set A; the polynomial coefficients of the polynomial fitting functions of all constant voltage charging current curves form a data set B; the polynomial coefficients of the polynomial fitting functions of all constant current discharging voltage curves form a data set C; S405: Construct a constant current charging voltage curve reconstruction model Model1, the model Model1 is a multi-input and multi-output multi-layer full connection layer model, data set B and data set C are taken as the input of the model, data set A is taken as the output of the model, so as to train the constant current charging voltage curve reconstruction model Model1, and save the model Model1 after training; S406: Construct a constant voltage charging current curve reconstruction model Model2, the model Model2 is a multi-input and multi-output multi-layer full connection layer model, data set A and data set C are taken as the input of the model, data set B is taken as the output of the model, so as to train the constant voltage charging current curve reconstruction model Model2, and save the model Model2 after training; S407: Construct a constant current discharging voltage curve reconstruction model Model3, the model Model3 is a multi-input and multi-output multi-layer full connection layer model, data set A and data set B are taken as the input of the model, data set C is taken as the output of the model, so as to train the constant current discharging voltage curve reconstruction model Model3, and save the model Model3 after training; S5: Reconstruct the missing charge-discharge curve, the specific steps are: in the actual application occasion of estimating SOH, collect constant current charging voltage curve, constant voltage charging current curve and constant current discharge voltage curve, the battery management system further detects whether a certain curve in the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharge voltage curve appears serious data missing condition, the serious data missing condition refers to the existence of long time, continuous multiple sampling period data loss in the collected curve data; if it is detected that a certain curve in the constant current charging voltage curve, the constant voltage charging current curve and the constant current discharge voltage curve appears serious data missing condition, then the corresponding model in the constant current charging voltage curve reconstruction model, the constant voltage charging current curve reconstruction model and the constant current discharge voltage curve reconstruction model in step S4 is used to reconstruct the curve with serious data missing condition; if there is no serious data missing condition, then directly proceed to step S6; S6: Online estimate the SOH of lithium battery, extract the health characteristics of constant current charging voltage curve, constant voltage charging current curve and constant current discharge voltage curve respectively, input into the SOH estimation model based on artificial neural network in step S3, and output SOH estimation value.

2. The lithium battery SOH estimation method based on polynomial reconstruction charge-discharge curve according to claim 1, characterized in that, The specific steps of reconstructing the missing charge-discharge curve in step S5 are as follows: If the battery management system detects that the constant-current charging voltage curve has serious data missing, the constant-voltage charging current curve and the constant-current discharging voltage curve collected are fitted according to the polynomial fitting function of the constant-voltage charging current curve and the polynomial fitting function of the constant-current discharging voltage curve obtained in step S4 to obtain corresponding polynomial coefficients; the obtained polynomial coefficients are input into the constant-current charging voltage curve reconstruction model obtained in step S4 to output a polynomial coefficient set of the reconstructed constant-current charging voltage curve Then further obtain a polynomial coefficient set The corresponding polynomial fitting function Poly cc The constant-current charging cutoff voltage is substituted into the polynomial fitting function Poly cc The corresponding time t1 is calculated, and the function curve segment corresponding to the argument range [0, t1] of the polynomial fitting function Poly cc is taken as the reconstructed constant-current charging voltage curve; If the battery management system detects that the constant-voltage charging current curve has serious data missing, the constant-current charging voltage curve and the constant-current discharging voltage curve collected are fitted according to the polynomial fitting functions of the constant-current charging voltage curve and the constant-current discharging voltage curve obtained in step S4 to obtain corresponding polynomial coefficients; the obtained polynomial coefficients are input into the constant-voltage charging current curve reconstruction model obtained in step S4 to output a polynomial coefficient set of the reconstructed constant-voltage charging current curve Then a polynomial coefficient set is further obtained The corresponding polynomial fitting function Poly cv The constant-voltage charging cutoff current is substituted into the polynomial fitting function Poly cv to obtain the corresponding time t2, and the function curve segment corresponding to the argument range [0, t2] of the polynomial fitting function Poly cv is taken as the reconstructed constant-voltage charging current curve; If the battery management system detects that the constant-current discharge voltage curve has serious data missing, the constant-current charging voltage curve and the constant-voltage charging current curve collected are fitted according to the polynomial fitting function of the constant-current charging voltage curve and the polynomial fitting function of the constant-voltage charging current curve obtained in step S4 to obtain corresponding polynomial coefficients; the obtained polynomial coefficients are input into the constant-current discharge voltage curve reconstruction model obtained in step S4 to output a polynomial coefficient set of the reconstructed constant-current discharge voltage curve Then further obtain the polynomial coefficient set The corresponding polynomial fitting function Poly dc The constant-current discharge cutoff voltage is substituted into the polynomial fitting function Poly dc The corresponding time t3 is calculated, and the function curve segment corresponding to the argument range [0, t3] of the polynomial fitting function Poly dc is taken as the reconstructed constant-current discharge voltage curve.

Images