A method for identifying the inflection point of lithium battery capacity based on dynamic grey prediction modeling

Abstract

The present invention relates to the field of battery management systems, and specifically discloses a method for identifying the inflection point of lithium battery capacity based on dynamic gray prediction modeling, including: establishing a new discrete gray prediction model based on the double exponential degradation characteristics of the capacity of the lithium battery; solving the parameters and discrete recursive formula of the gray prediction model, using 40% of the charge and discharge cycle data of the lithium battery's entire life cycle to train the model, and obtain the predicted degradation trajectory of its entire life cycle; integrating four identification methods based on the geometric feature fusion of the predicted trajectory to identify the inflection point position of the capacity, thereby solving the problem of lithium battery capacity prediction and health management based on small sample sequence modeling, and providing an important reference for the design of a reasonable battery replacement strategy. The method proposed by the present invention can well capture the double exponential degradation law of the capacity and obtain a higher prediction accuracy, so that battery managers can better grasp the aging of the battery and implement a reasonable battery replacement strategy.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric vehicle battery management systems, and in particular relates to a lithium battery capacity inflection point identification method based on dynamic grey prediction modeling. Background Art

[0002] As a form of energy storage, lithium batteries are widely used in electronic devices, new energy vehicles, and even spacecraft due to their high energy density, small size, and convenience. In the battery management system, lithium battery degradation prediction is an important link. For battery managers, this is related to the safe use of batteries, cascade utilization, and the formulation of replacement strategies, making battery degradation prediction a hot research issue in battery management systems.

[0003] Since batteries exhibit a complex nonlinear degradation process, a large number of prediction methods have been developed. Existing battery capacity predictions are divided into the following three categories: (1) Model-based modeling methods. This type of model is mainly established based on the intrinsic reaction of battery degradation, which can be specifically divided into electrochemical models, equivalent circuit models and empirical degradation models. (2) Data-driven prediction models. In the context of the big data era, data-driven models such as deep learning, neural networks, and LSTM use collected experimental data to extract battery degradation features to learn capacity degradation. There is no need to understand the accurate physical and chemical mechanism parameters of battery failure. The law of lithium battery performance degradation can be directly mined from data such as lithium battery voltage and current. It has strong applicability and is widely welcomed. (3) Hybrid prediction methods. Hybrid prediction methods combine model-based and data-driven methods, such as combining Gaussian processes with deep learning, particle filtering with interactive multi-model fusion, principal component analysis with neural networks, and first performing modal decomposition and then using time series models. However, in addition to normal nonlinear degradation, some superlinear trends often appear in the aging trajectory of the battery. This degradation is generally called the capacity inflection point phenomenon. Simply put, the capacity inflection point is a turning point in the degradation process where the degradation rate changes greatly. According to relevant data statistics, the remaining capacity after the inflection point is less than 10% of the rated capacity, and the effective working platform accounts for more than 90% of the total capacity. In other words, the appearance of the inflection point indicates that the battery will soon reach the exhaustion of available capacity. Therefore, avoiding or delaying the capacity inflection point is crucial to ensure the extension of battery life. At the same time, the location of the inflection point poses a challenge to accurate battery health estimation, even batteries with the same health status may have completely different remaining service life. Being able to identify the capacity inflection point as early as possible has important practical significance for the entire battery life cycle management.

[0004] The identification of the inflection point of battery capacity requires accurate aging trajectories. As the battery cycles through charge and discharge, the continuous transfer of internal lithium ions between the positive and negative electrodes, the loss of active substances, and the complex electrochemical reactions cause the battery capacity to degrade. Model-based methods require repeated experimental testing of the same type of battery over a long period of time to track its degradation process, and it is difficult to develop new mechanism models. Although some relatively ideal results have been obtained for popular data-driven models through continuous learning and parameter adjustment of big data samples, due to the continuous development of battery technology, the metal composition, operating temperature conditions, and fast charging technology of the battery have a great impact on battery degradation. Collecting a large amount of data not only has a great computational burden, but also cannot explain the consistency of the capacity degradation problem. Most data-driven machine learning algorithms are also subject to certain limitations in this regard. Therefore, it is particularly important to model capacity degradation based on small sample data. Summary of the invention

[0005] The purpose of the invention is to address the deficiencies of the above-mentioned background technology, propose a method for identifying the inflection point of lithium battery capacity based on dynamic gray prediction modeling, construct a new discrete gray prediction model, predict the degradation of lithium battery capacity, and identify the degradation inflection point of capacity in combination with non-numerical methods. As a special type of data-driven model, the gray prediction model is often used to predict the remaining life of lithium batteries. The gray prediction model uses the cumulative generation operator to perform data mining on sequences with unclear evolution laws. It is not only suitable for characterizing sequences with a single change law, but also suitable for predicting sequences with mixed trend changes, and has strong adaptability. The nonlinear degradation prediction of lithium-ion batteries is of great value in battery management systems, and the capacity inflection point is an important signal of the health status of lithium batteries in power vehicles. The degradation process before and after the inflection point is regarded as two stages with different degradation characteristics. Therefore, the inflection point has become a key issue that cannot be ignored in degradation.

[0006] To achieve the above object, the present invention adopts the following technical solution: a method for identifying the inflection point of lithium battery capacity based on dynamic grey prediction modeling, comprising:

[0007] S1. Collect the lithium battery capacity data after each charge and discharge cycle and calculate the degradation characteristics;

[0008] S2. According to the degradation characteristics of lithium battery capacity, analyze the degradation characteristics of lithium battery capacity and construct a grey prediction model suitable for the degradation characteristics;

[0009] S3, solving the parameter estimation and discrete recursive formula of the grey prediction model constructed in step S2 to obtain the predicted degradation curve of the lithium battery capacity;

[0010] S4. Use the non-numerical geometric feature integration method to identify the inflection point where the lithium battery capacity degradation rate changes greatly in the predicted degradation curve.

[0011] Further, step S1 includes:

[0012] Collect and clean the data of lithium battery capacity degradation characteristics, and calculate the degradation rate of lithium battery capacity after each charge and discharge cycle; use discrete time series to represent lithium battery capacity, and record a moment for each charge and discharge; remember that the lithium battery capacity sequence from moment 1 to moment n is: Q(1), Q(2), ..., Q(n);

[0013] At the same time, the degradation amount ΔQ(i) and degradation rate λ(i) of the lithium battery capacity after each charge and discharge cycle are calculated, and the formula is expressed as:

[0014] ΔQ(i)=Q(i-1)-Q(i), i=1,2,...,k;

[0015]

[0016] Among them, Q(i) represents the lithium battery capacity at time i, and Q(i-1) represents the lithium battery capacity at time i-1.

[0017] Further, step S2 includes:

[0018] According to the lithium battery capacity sequence Q(1), Q(2), ..., Q(n) obtained in step S1, the degradation characteristics of the lithium battery capacity are analyzed:

[0019] When the lithium battery capacity Q(i) at any time i satisfies the following formula:

[0020] Q(i)=ai+b;

[0021] Where a and b are both constants; it is considered that the capacity of the lithium battery exhibits linear degradation characteristics;

[0022] When the lithium battery capacity Q(i) at any time i satisfies the following formula:

[0023]

[0024] Where Q(i+1) is the capacity of the lithium battery at time i+1, and c is a constant; it is considered that the capacity of the lithium battery exhibits exponential degradation characteristics;

[0025] When the capacity of the lithium battery does not meet the above two degradation characteristics, but meets the existence of a capacity inflection point, the degradation is slow before the inflection point, and the degradation is fast after the inflection point. From the perspective of the charge and discharge cycle of the entire lithium battery life cycle, it presents a double exponential degradation characteristic. A new discrete gray prediction model is constructed, and the formula is expressed as follows:

[0026] Q (1) (k+1)=β1Q (1) (k)+Q (1) (k-1)+β2k+β3;

[0027] Among them, Q (1) (k) represents the cumulative capacity of the lithium battery at time k, and its formula is expressed as:

[0028]

[0029] Q (1) (k+1), Q (1) (k-1) is the sum of the lithium battery capacity at the next moment and the previous moment of moment k; β1, β2, and β3 are the three parameter estimates to be solved.

[0030] Further, step S3 includes:

[0031] S31, taking the minimum error square sum as the objective function, the least square method is used to find the parameter estimation matrix P, and the specific calculation formula is:

[0032] P=(B T B) -1 B T Y;

[0033] Among them, the specific formulas of Y and B are:

[0034]

[0035] Among them, Q (1) (n) represents the cumulative capacity of the lithium battery at time n, Q (1) (n-2) represents the accumulation of the lithium battery capacity at the two moments before moment n;

[0036] The obtained parameter estimation matrix P is expressed as:

[0037]

[0038] The three parameter estimates β1, β2, and β3 can be solved;

[0039] S32, bring the parameter estimation value P into the constructed discrete grey prediction model to obtain the first-order cumulative sequence prediction value of the lithium battery capacity at time k

[0040] S33, using the first-order cumulative reduction operator as a discrete recursive formula to obtain the predicted value of the lithium battery capacity at time k The formula is:

[0041]

[0042] By connecting the predicted values ​​at each discrete moment, we can obtain the predicted degradation curve of the lithium battery capacity.

[0043] More specifically, the mean absolute percentage error MAPE, standard deviation STD, and root mean square percentage error RMSPE are used as evaluation criteria to evaluate the accuracy of the predicted degradation curve of the lithium battery capacity obtained by the constructed grey prediction model. The specific formula is as follows:

[0044]

[0045] Among them, Q(k) is the capacity of the lithium battery at time k, is the predicted value of the lithium battery capacity at time k, v, j are the end time and start time intercepted in the capacity sequence.

[0046] Further, step S4 includes:

[0047] Step S4 uses four geometric feature integration methods to identify the inflection point of the lithium battery capacity based on the predicted degradation curve of the lithium battery capacity; let the predicted degradation curve of the lithium battery capacity be l, then the specific process of step S4 includes:

[0048] S41, Kneedle method: Connect the straight line l' between the starting point and the end point of the predicted degradation curve l, and the formula is expressed as:

[0049]

[0050] Among them, x(end) and x(start) represent the abscissas of the end point and the starting point respectively, and y(end) and y(start) represent the ordinates of the end point and the starting point respectively;

[0051] Find the point farthest from the straight line l' in the predicted degradation curve l of the lithium battery capacity, which is the predicted inflection point 1 (k1, l (k1)) of the lithium battery capacity. The formula is:

[0052]

[0053] Where dist(.) represents the Euclidean distance from the point on the predicted degradation curve l to the straight line l';

[0054] S42, Bacon-Watts method: The fitting straight lines near the starting point and the end point on the predicted degradation curve l are recorded as l1 and l2 respectively, and the intersection of the two straight lines is the lithium battery capacity prediction inflection point 2(k2,l(k2)), that is, {(k2,l(k2))|k2∈l1∩l2};

[0055] S43, Tangent-ratio method: Take the intersection of the tangents at any point on the fitting curve near the starting point and the end point on the predicted degradation curve l as the lithium battery capacity prediction inflection point 3 (k3, l (k3)), and the formula is expressed as:

[0056] {(k3,l(k3))|k3∈l3∩l4,K(l3)=l′1(x0),K(l4)=l′2(x′0)};

[0057] Where K(.) is the slope of the straight line, l′1 and l′2 represent the fitting curves near the starting point and the end point, respectively, x0 and x′0 are arbitrary points on l1 and l2, respectively, l′1(x0) and l′2(x′0) represent the slopes of any selected points on the fitting curve, respectively, l3 and l4 are the tangent lines of l′1 and l′2 at points x0 and x′0, respectively;

[0058] S44, Bisector method: The intersection of the set of points with equal distances between the fitting straight lines l1 and l2 near the starting point and the end point and the curve is used to determine the position of the inflection point 4 (k4, l(k4)) of the lithium battery capacity prediction. The formula is expressed as:

[0059] {(k4,l(k4))|k4∈l∩(x,y)|dist((x,y),l1)=dist((x,y),l2)};

[0060] Among them, (x, y) represents a set of points, and dist(.) represents the Euclidean distance from a point to a line;

[0061] S45, determine the position of the final inflection point; integrate the four predicted inflection points of Kneedle, Bacon-Watts, Tangent-ratio and Bisector, and use the mean method to identify the position of the final lithium battery capacity inflection point.

[0062] More specifically, 40% of the lithium battery capacity degradation characteristic data obtained from the charge and discharge cycle of the lithium battery throughout its life cycle is used as a training set to train the constructed grey prediction model, and the remaining 60% is used as a test set to test the predicted degradation curve of the lithium battery capacity and obtain the predicted inflection point of the lithium battery capacity; then the real lithium battery inflection point obtained from all the data is used as a standard to evaluate the accuracy of the constructed grey prediction model.

[0063] The present invention adopts the above technical solution and has the following beneficial effects:

[0064] 1. Construct a new discrete grey prediction model for the degradation characteristics of lithium battery capacity. Based on small sample data, the new model can well capture the double exponential degradation law of capacity and achieve high prediction accuracy. The grey prediction model does not require prior knowledge of battery degradation experience, does not require assumptions about the degradation law of battery degradation, and does not require the corresponding selection of specific model forms. At the same time, the entire life cycle prediction can be obtained using the first 40% of experimental data.

[0065] 2. Based on the degradation curve obtained by the grey prediction model, the identification engineering technology of the battery capacity inflection point is completed, which can help battery managers better understand the aging of the battery and formulate a reasonable battery replacement strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A schematic flow chart of the method proposed in the present invention;

[0067] Figure 2 A specific flow chart for plotting the predicted degradation curve and identifying the inflection point of lithium battery capacity;

[0068] Figure 3 The prediction effect of the central difference operator on the grey model under different noise levels;

[0069] Figure 4 Predicted degradation curve result diagram of lithium battery capacity;

[0070] Figure 5 Lithium battery capacity inflection point prediction and identification results. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0072] Although the steps in the present invention are arranged with numbers, they are not used to limit the order of the steps. Unless the order of the steps is clearly stated or the execution of a certain step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" used in this article involves and covers any and all possible combinations of one or more of the associated listed items.

[0073] Please refer to Figures 1 to 5 , a specific implementation of this embodiment is given:

[0074] like Figure 1 As shown in FIG. 1 , a method for identifying the inflection point of lithium battery capacity based on dynamic grey prediction modeling proposed by the present invention is given, such as Figure 2As shown, the method includes two parts, namely, predictive degradation curve drawing and inflection point identification; specifically includes the following steps:

[0075] S1. Collect the lithium battery capacity data after each charge and discharge cycle and calculate the degradation characteristics;

[0076] As a preferred implementation mode of step S1, it specifically includes:

[0077] Data collection and cleaning of lithium battery capacity degradation characteristic information are performed to calculate the degradation rate of lithium battery capacity after each charge and discharge cycle. The degradation process of lithium battery is mainly based on each charge and discharge cycle to describe the capacity change of lithium battery. In essence, it can be regarded as a time series. Therefore, discrete time series is used to represent the capacity of lithium battery. Each charge and discharge is recorded as a moment. The lithium battery capacity sequence from the 1st moment to the nth moment is: Q(1), Q(2), ..., Q(n).

[0078] At the same time, the degradation amount ΔQ(i) and degradation rate λ(i) of the lithium battery capacity after each charge and discharge cycle are calculated, and the formula is expressed as:

[0079] ΔQ(i)=Q(i-1)-Q(i), i=1,2,...,k;

[0080]

[0081] Among them, Q(i) represents the lithium battery capacity at time i, and Q(i-1) represents the lithium battery capacity at time i-1.

[0082] S2. According to the degradation characteristics of lithium battery capacity, analyze the degradation characteristics of lithium battery capacity and construct a grey prediction model for degradation characteristics;

[0083] As a preferred implementation of step S2, it specifically includes:

[0084] According to the lithium battery capacity sequence Q(1), Q(2), ..., Q(n) obtained in step S1, the degradation characteristics of the lithium battery capacity are analyzed:

[0085] When the lithium battery capacity Q(i) at any time i satisfies the following formula:

[0086] Q(i)=ai+b;

[0087] Where a and b are both constants; it is considered that the capacity of the lithium battery exhibits linear degradation characteristics;

[0088] When the lithium battery capacity Q(i) at any time i satisfies the following formula:

[0089]

[0090] Where Q(i+1) is the capacity of the lithium battery at time i+1, and c is a constant; it is considered that the capacity of the lithium battery exhibits exponential degradation characteristics;

[0091] When the capacity of the lithium battery does not meet the above two degradation characteristics, but meets the existence of a capacity inflection point, the degradation is slow before the inflection point, and the degradation is fast after the inflection point. From the perspective of the charge and discharge cycle of the entire lithium battery life cycle, it presents a double exponential degradation characteristic. A new discrete gray prediction model is constructed, and the formula is expressed as follows:

[0092] Q (1) (k+1)=β1Q (1) (k)+Q (1) (k-1)+β2k+β3;

[0093] Among them, Q (1) (k) represents the cumulative capacity of the lithium battery at time k, Q (1) (k+1), Q (1) (k-1) is the sum of the lithium battery capacity at the next moment and the previous moment of time k; β1, β2, β3 are the three parameter estimates to be solved; Q (1) The formula for (k) is:

[0094]

[0095] This formula reflects that the nature of lithium battery degradation is the accumulation of capacity degradation in each charge and discharge cycle;

[0096] The new grey prediction model constructed in the method proposed in the present invention is an improvement on the discrete non-homogeneous grey prediction model NDGM on the basis that the traditional discrete grey model has not yet demonstrated the characterization and prediction of the double exponential degradation sequence, wherein the general form of the NDGM model is:

[0097] Q (1) (k+1)=β1Q (1) (k)+β2k+β3;

[0098] The existing NDGM model can only be used for sequence prediction of linear degradation characteristics and exponential degradation characteristics, while the new discrete grey prediction model, based on the original NDGM, displays the original sequence and cumulative sequence information at the current moment;

[0099] More specifically, since the discrete grey prediction model is the discretization of the continuous grey prediction model, it is a special form obtained by using differential discretization on the continuous model; usually Perform forward difference and get In this embodiment, in order to emphasize the importance of the adjacent information at this time point, the central difference operator is used in the above formula, that is:

[0100]

[0101] This is consistent with most sequence evolution laws in the objective world. The original sequence information at the current moment depends on the change information at the previous moment and the accumulation until the previous moment. It also proves that the modeling mechanism of the discrete grey prediction model of the present invention is reasonable and correct.

[0102] The influence of the central difference operator on the prediction effect of the grey model under different noise levels is as follows: Figure 3 As shown. The prediction accuracy of the model at different step sizes is also shown. Since the data sample capacity is fixed, different prediction step sizes will cause changes in the data sample length in the simulation phase. Through numerical simulation experiments, it can be found that the central difference model is generally better than the original form, and the error in both the simulation phase and multi-step prediction is smaller than the original form.

[0103] S3, solving the parameter estimation and discrete recursive formula of the grey prediction model constructed in step S2 to obtain the predicted degradation curve of the lithium battery capacity;

[0104] As a preferred implementation of step S3, it specifically includes:

[0105] S31, taking the minimum error square sum as the objective function, the least square method is used to find the parameter estimation matrix P, and the specific calculation formula is:

[0106] P=(B T B) -1 B T Y;

[0107] Among them, the specific formulas of Y and B are:

[0108]

[0109] Among them, Q (1) (n) represents the cumulative capacity of the lithium battery at time n, Q (1) (n-2) represents the accumulation of the lithium battery capacity at the two moments before moment n;

[0110] The obtained parameter estimation matrix P is expressed as:

[0111]

[0112] The three parameter estimates β1, β2, and β3 can be solved;

[0113] S32, bring the parameter estimation value P into the constructed discrete grey prediction model to obtain the first-order cumulative sequence prediction value of the lithium battery capacity at time k

[0114] S33, using the first-order cumulative reduction operator as a discrete recursive formula to obtain the predicted value of the lithium battery capacity at time k The formula is:

[0115]

[0116] By connecting the predicted values ​​at each discrete moment, we can obtain the predicted degradation curve of the lithium battery capacity.

[0117] S4. Using a non-numerical geometric feature integration method, the inflection point position where the capacity degradation rate of the lithium battery changes greatly in the predicted degradation curve is identified;

[0118] As a preferred implementation of step S4, it specifically includes:

[0119] Based on the predicted degradation curve of the lithium battery capacity, four geometric feature integration methods are used to identify the inflection point of the lithium battery capacity; the predicted degradation curve of the lithium battery capacity is denoted as l, and the specific process of step S4 includes:

[0120] S41, Kneedle method: Connect the straight line l' between the starting point and the end point of the predicted degradation curve l, and the formula is expressed as:

[0121]

[0122] Among them, x(end) and x(start) represent the abscissas of the end point and the starting point respectively, and y(end) and y(start) represent the ordinates of the end point and the starting point respectively;

[0123] Find the point farthest from the straight line l' in the predicted degradation curve l of the lithium battery capacity, which is the predicted inflection point 1 (k1, l (k1)) of the lithium battery capacity. The formula is:

[0124]

[0125] Where dist(.) represents the Euclidean distance from the point on the predicted degradation curve l to the straight line l';

[0126] S42, Bacon-Watts method: The fitting straight lines near the starting point and the end point on the predicted degradation curve l are recorded as l1 and l2 respectively, and the intersection of the two straight lines is the lithium battery capacity prediction inflection point 2(k2,l(k2)), that is, {(k2,l(k2))|k2∈l1∩l2};

[0127] S43, Tangent-ratio method: Take the intersection of the tangents at any point on the fitting curve near the starting point and the end point on the predicted degradation curve l as the lithium battery capacity prediction inflection point 3 (k3, l (k3)), and the formula is expressed as:

[0128] {(k3,l(k3))|k3∈l3∩l4,K(l3)=l′1(x0),K(l4)=l′2(x′0)};

[0129] Where, K(.) is the slope of the straight line, l′1 and l′2 represent the fitting curves near the starting point and the end point, respectively, x0 and x′0 are arbitrary points on l1 and l2, respectively, l′1(x0) and l′2(x′0) represent the slopes of any selected points on the fitting curve, l3 and l4 are the tangent lines of l′1 and l′2 at points x0 and x′0, respectively;

[0130] S44, Bisector method: The intersection of the set of points with equal distances between the fitting straight lines l1 and l2 near the starting point and the end point and the curve is used to determine the position of the inflection point 4 (k4, l(k4)) of the lithium battery capacity prediction. The formula is expressed as:

[0131] {(k4,l(k4))|k4∈l∩(x,y)|dist((x,y),l1)=dist((x,y),l2)};

[0132] Among them, (x, y) represents a set of points, and dist(.) represents the Euclidean distance from a point to a line;

[0133] S45, determine the position of the final inflection point; integrate the four predicted inflection points of Kneedle, Bacon-Watts, Tangent-ratio and Bisector, and use the mean method to identify the position of the final lithium battery capacity inflection point.

[0134] In addition, in this embodiment, the mean absolute percentage error MAPE, standard deviation STD, and root mean square percentage error RMSPE are used as evaluation criteria to evaluate the accuracy of the predicted degradation curve of the lithium battery capacity obtained by the grey prediction model constructed in step S3. The specific formula is as follows:

[0135]

[0136] Among them, Q(k) is the capacity of the lithium battery at time k, is the predicted value of the lithium battery capacity at time k, v, j are the end time and start time intercepted in the capacity sequence.

[0137] In this embodiment, lithium batteries are used to verify the effect, and three lithium batteries of the same material in different fast charging modes are selected as research objects in the actual experiment. After a new lithium battery is put into use, the battery capacity degrades relatively slowly in the early stage and the performance is relatively stable. However, after several charge and discharge cycles, the degradation rate shows a significant and substantial decline, causing the battery to quickly reach the end of its life. After initially extracting the battery capacity data, we found that its degradation curve showed an obvious turning point. 40% of the full life charge and discharge cycle data of each battery is selected as the training set, and the lithium battery capacity prediction degradation curve results are shown as follows. Figure 4 Its accuracy evaluation is shown in Table 1.

[0138] Table 1. Lithium battery capacity prediction degradation curve accuracy evaluation table

[0139]

[0140] It can be seen from the table that the error between the predicted results of the predicted degradation curve and the real data curve is small. In the training stage, the model error is controlled at 1%, with high accuracy. In the long-term prediction stage, MAPE can also be controlled at 5%. Based on the performance of the prediction model, such accuracy can far meet the requirements.

[0141] Based on the same experimental environment as above, the effect of inflection point identification is also verified, and the results are as follows Figure 5 As shown, the specific number of charge and discharge cycles where the inflection point of the lithium battery capacity is located is also given in Table 2 below:

[0142] Table 2. Table of charge and discharge cycle times for the inflection point of lithium battery capacity

[0143] Battery 1 Battery 2 Battery 3 The real turning point 393 382 347 Predicting inflection points 427 379 367

[0144] From a numerical point of view, the real inflection point based on the total life cycle is roughly located at about 70% of the total life cycle. The experimental results show that the capacity inflection point identified based on the predicted degradation curve is very close to the real inflection point. In particular, the real inflection point of battery 2 basically coincides with the predicted inflection point. It is effective to use 40% of the sample data to train the model and the remaining 60% of the data to draw the predicted degradation curve and identify the inflection point, which reflects the effectiveness and rationality of the method proposed in the present invention.

[0145] In summary, the method proposed in the present invention aims at the degradation characteristics of lithium battery capacity and constructs a new discrete gray prediction model. Based on small sample data, the new model can well capture the double exponential degradation law of capacity and can obtain higher prediction accuracy. The gray prediction model does not require prior knowledge of battery degradation experience, does not require assumptions about the degradation law of battery degradation, and does not require the corresponding selection of specific model forms. At the same time, the entire life cycle prediction can be obtained using the first 40% of experimental data.

[0146] Based on the degradation curve obtained by the grey prediction model, the identification engineering technology of the battery capacity inflection point is completed, which can help battery managers better understand the aging of the battery and formulate a reasonable battery replacement strategy.

[0147] In this specification, the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" refer to at least one embodiment or example described in combination with specific features, structures, materials or characteristics. These described specific features, structures, materials or characteristics can be combined in an appropriate manner in one or more embodiments or examples. In addition, the technician can combine and combine different embodiments or examples and their features described in this specification without contradiction.

[0148] The logic and / or steps shown in the flowchart or described in other ways can be regarded as a sequence of executable instructions for implementing logical functions. These instructions can be implemented in any computer-readable medium for use by instruction execution systems, devices or equipment. These systems, devices or equipment include processor systems or other systems capable of receiving and executing instructions.

[0149] The above embodiments have described the principles and implementations of the present invention in detail, and have used specific examples to illustrate the working principles thereof. These examples are only used to help understand the method of the present invention and its core concept. Meanwhile, according to the concept of the present invention, the actual implementation methods and application scope may vary. Therefore, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for identifying the inflection point of lithium battery capacity based on dynamic grey prediction modeling, characterized in that: The following steps are involved: S1, collecting lithium battery capacity data after each charge and discharge cycle, and calculating degradation characteristics; Step S1 includes: Collect and clean the data of lithium battery capacity degradation characteristics, and calculate the degradation rate of lithium battery capacity after each charge and discharge cycle; use discrete time series to represent lithium battery capacity, and record a moment for each charge and discharge; remember that the lithium battery capacity sequence from moment 1 to moment n is: Q(1), Q(2), ..., Q(n); At the same time, the degradation amount ΔQ(i) and degradation rate λ(i) of the lithium battery capacity after each charge and discharge cycle are calculated, and the formula is expressed as: ΔQ(i)=Q(i-1)-Q(i), i=1,2,...,k; Among them, Q(i) represents the lithium battery capacity at time i, and Q(i-1) represents the lithium battery capacity at time i-1; S2. Analyze the degradation characteristics of lithium battery capacity according to the degradation characteristics of lithium battery capacity, and construct a grey prediction model suitable for the degradation characteristics; Step S2 includes: According to the lithium battery capacity sequence Q(1), Q(2), ..., Q(n) obtained in step S1, the degradation characteristics of the lithium battery capacity are analyzed: When the lithium battery capacity Q(i) at any time i satisfies the following formula: Q(i)=ai+b; Where a and b are both constants; it is considered that the capacity of the lithium battery exhibits linear degradation characteristics; When the lithium battery capacity Q(i) at any time i satisfies the following formula: Where Q(i+1) is the capacity of the lithium battery at time i+1, and c is a constant; it is considered that the capacity of the lithium battery exhibits exponential degradation characteristics; When the capacity of the lithium battery does not meet the above two degradation characteristics, but meets the existence of a capacity inflection point, the degradation rate λ(i) before the inflection point is slower, the degradation amount ΔQ(i) is smaller, and the degradation rate λ(i) after the inflection point is faster, and the degradation amount ΔQ(i) is larger. From the perspective of the charge and discharge cycle of the entire lithium battery life cycle, it presents a double exponential degradation characteristic. A new discrete gray prediction model is constructed, and the formula is expressed as follows: Q (1) (k+1)=β1Q (1) (k)+Q (1) (k-1)+β2k+β3; Among them, Q (1) (k) represents the cumulative capacity of the lithium battery at time k, and its formula is expressed as: Q (1) (k+1), Q (1) (k-1) is the sum of the lithium battery capacity at the next moment and the previous moment of time k; β1, β2, and β3 are the three parameter estimates to be solved; S3, solving the parameter estimation and discrete recursive formula of the grey prediction model constructed in step S2 to obtain the predicted degradation curve of the lithium battery capacity; step S3 includes: S31, taking the minimum error square sum as the objective function, the least square method is used to find the parameter estimation matrix P, and the specific calculation formula is: P=(B T B) -1 B T Y; Among them, the specific formulas of Y and B are: Among them, Q (1) (n) represents the cumulative capacity of the lithium battery at time n, Q (1) (n-2) represents the accumulation of the lithium battery capacity at the two moments before moment n; The obtained parameter estimation matrix P is expressed as: The three parameter estimates β1, β2, and β3 can be solved; S32, bring the parameter estimation value P into the constructed discrete grey prediction model to obtain the first-order cumulative sequence prediction value of the lithium battery capacity at time k S33, using the first-order cumulative reduction operator as a discrete recursive formula to obtain the predicted value of the lithium battery capacity at time k The formula is: Connect the predicted values ​​at each discrete moment to obtain the predicted degradation curve of the lithium battery capacity; S4. Use the non-numerical geometric feature integration method to identify the inflection point where the lithium battery capacity degradation rate changes greatly in the predicted degradation curve.

2. The method for identifying the inflection point of lithium battery capacity based on dynamic grey prediction modeling according to claim 1, characterized in that: The mean absolute percentage error (MAPE), standard deviation (STD), and root mean square percentage error (RMSPE) are used as evaluation criteria to evaluate the accuracy of the predicted degradation curve of lithium battery capacity obtained by the proposed new grey prediction model. The specific formula is as follows: Among them, Q(k) is the capacity of the lithium battery at time k, is the predicted value of the lithium battery capacity at time k, v, j are the end time and start time intercepted in the capacity sequence.

3. The method for identifying the inflection point of lithium battery capacity based on dynamic grey prediction modeling according to claim 1, characterized in that: Step S4: Based on the predicted degradation curve of the lithium battery capacity, four geometric feature integration methods are used to identify the inflection point of the lithium battery capacity; The predicted degradation curve of the lithium battery capacity is denoted as 1, and the specific process of step S4 includes: S41, Kneedle method: Connect the straight line l' between the starting point and the end point of the predicted degradation curve l, and the formula is expressed as: Among them, x(end) and x(start) represent the abscissas of the end point and the starting point respectively, and y(end) and y(start) represent the ordinates of the end point and the starting point respectively; Find the point farthest from the straight line l' in the predicted degradation curve l of the lithium battery capacity, which is the predicted inflection point 1 (k1, l (k1)) of the lithium battery capacity. The formula is: Where dist(.) represents the Euclidean distance from the point on the predicted degradation curve l to the straight line l'; S42, Bacon-Watts method: The fitting straight lines near the starting point and the end point on the predicted degradation curve l are recorded as l1 and l2 respectively, and the intersection of the two straight lines is the lithium battery capacity prediction inflection point 2(k2,l(k2)), that is, {(k2,l(k2))|k2∈l1∩l2}; S43, Tangent-ratio method: Take the intersection of the tangents at any point on the fitting curve near the starting point and the end point on the predicted degradation curve l as the lithium battery capacity prediction inflection point 3 (k3, l (k3)), and the formula is expressed as: {(k3,l(k3))|k3∈l3∩l4,K(l3)=l′1(x0),K(l4)=l′2(x′0)}; Where K(.) is the slope of the straight line, l′1 and l′2 represent the fitting curves near the starting point and the end point, respectively, using polynomial fitting, x0 and x′0 are arbitrary points on l1 and l2, respectively, l′1(x0) and l′2(x′0) represent the slopes of any selected points on the fitting curve, l3 and l4 are the tangent lines of l′1 and l′2 at points x0 and x′0, respectively; S44, Bisector method: The intersection of the set of points with equal distances between the fitting straight lines l1 and l2 near the starting point and the end point and the curve is used to determine the position of the inflection point 4 (k4, l(k4)) of the lithium battery capacity prediction. The formula is expressed as: {(k4,l(k4))|k4∈l∩(x,y)|dist((x,y),l1)=dist((x,y),l2)}; Among them, (x, y) represents a set of points, and dist(.) represents the Euclidean distance from a point to a line; S45, determine the position of the final inflection point; integrate the four predicted inflection points of Kneedle, Bacon-Watts, Tangent-ratio and Bisector, and use the mean method to identify the position of the final lithium battery capacity inflection point.

4. The method for identifying the inflection point of lithium battery capacity based on dynamic grey prediction modeling according to claim 1, characterized in that: 40% of the lithium battery capacity degradation characteristic data obtained from the charge and discharge cycle of the lithium battery throughout its life cycle is used as a training set to train the constructed grey prediction model, and the remaining 60% is used as a test set to verify the predicted degradation curve of the lithium battery capacity and obtain the predicted lithium battery capacity inflection point. The real lithium battery inflection point obtained from all the data is used as a standard to evaluate the accuracy of the constructed grey prediction model.

Images