Method and device for modeling hysteresis characteristic of battery and storage medium

Abstract

The invention discloses a method and device for modeling hysteresis characteristics of a battery and a storage medium, and belongs to the field of battery modeling. The invention relates to a method for modeling a hysteresis characteristic of a battery, which comprises the following steps of: modeling the hysteresis characteristic of the battery to obtain an initial model, and discretizing the initial model to obtain a discrete hysteresis characteristic model of the battery; measuring the OCV and the SOC when the charging and discharging direction of the battery only changes once to obtain a first-order rotation curve of the battery; and performing parameter identification on the discrete battery hysteretic characteristic model based on the battery first-order rotation curve to obtain a target battery hysteretic characteristic model. The constructed target battery hysteretic characteristic model can distinguish the reversible component and the irreversible component of the OCV, the OCV reconstruction precision and the SOC estimation precision are improved, the accuracy when the OCV is used for battery sorting and matching is improved, and meanwhile the hardware cost is not increased.

Description

Technical Field

[0001] The present invention belongs to the field of battery modeling, and more specifically, relates to a method, device and storage medium for modeling battery hysteresis characteristics. Background Art

[0002] Due to the core advantages of lithium-ion batteries such as high energy density, long cycle life, low self-discharge rate, and no memory effect, they are currently widely used in portable devices, electric vehicles, grid energy storage, aerospace and other industrial fields. Compared with ternary lithium batteries, lithium iron phosphate batteries have better thermal stability, which can effectively reduce the risk of safety accidents such as fire and explosion. At the same time, the manufacturing cost is relatively low, which is suitable for scenarios with high requirements for economy and safety. However, due to the combined influence of multiple factors such as stress accumulation during the phase change of the lithium iron phosphate positive electrode and the mass transfer characteristics of the graphite negative electrode, the OCV-SOC curve of the lithium iron phosphate battery shows a hysteresis phenomenon, that is, at the same SOC, the OCV measured during the charging and discharging process does not overlap, which directly leads to the fact that there is no one-to-one mapping relationship between SOC and OCV, which increases the difficulty of SOC estimation.

[0003] In response to the above problems, the most commonly used solution is to obtain the average OCV-SOC curve by taking the average value, and then use it as an important basis for estimating SOC under dynamic conditions. For batteries with significant voltage platforms such as lithium iron phosphate batteries, liquid metal batteries, and lithium sulfur batteries, due to the existence of a persistent and flat voltage platform, the deviation and noise introduced in the battery voltage and current sampling process will cause serious degradation of the SOC estimation accuracy. At this time, the disadvantages of applying the average OCV-SOC curve will become apparent: taking the OCV of a lithium iron phosphate battery at 50% SOC as an example, the measured values ​​obtained during the charging and discharging process have a difference of about 24mV. If this difference is reflected in the average OCV-SOC curve, the corresponding SOC range is 33% to 66%. This means that if the hysteresis voltage is ignored, the SOC estimation error will be as high as 15% or more.

[0004] In Chinese invention patent CN201510578190.X, a method for estimating the state of charge based on the hysteresis characteristic of the open circuit voltage is disclosed. The method determines the initial parameters of the adaptive model of the hysteresis characteristic of the open circuit voltage based on the Preisach operator according to the hysteresis characteristic curve training, and establishes the adaptive model of the hysteresis characteristic of the open circuit voltage based on the Preisach operator, and estimates the state of charge of the lithium-ion battery online based on this. In Chinese invention patent CN200980105680.0, a method for calculating the state of charge of the battery using the Preisach model is disclosed. However, these two methods only directly use the classic Preisach model to fit the experimental data, and use it for real-time prediction of the battery hysteresis voltage after completing parameter identification, which is not enough to accurately describe the real physical law of the change of OCV with SOC. In fact, during the charge and discharge process, the physical mechanism that drives the dynamic change of OCV can be divided into two types: first, the change of lithium concentration will cause the change of electrode potential. This change is completely reversible, and there is a definite functional relationship between the ideal electrode equilibrium potential and the lithium concentration. The ideal OCV is determined only by the current SOC and has nothing to do with the historical change trajectory of the SOC. Secondly, due to the existence of the hysteresis effect, at a certain SOC, the measured OCV will always deviate from the ideal OCV, and the deviation between the two is closely related to the charge and discharge history. This results in differences in OCV when the battery reaches the same SOC under different charge and discharge paths, and this change is irreversible.

[0005] From the perspective of model accuracy, the classic Preisach model can only accurately reflect the irreversible component of OCV changes, but does not include the reversible component in the modeling scope. It cannot fully reflect the two different mechanisms that drive OCV changes, thereby reducing the accuracy of the model. From the perspective of model application, directly using the classic Preisach model can only obtain the actual value of the open circuit voltage after integrating the two factors. In some application scenarios, it is also necessary to obtain a theoretical OCV value that is only related to the lithium concentration in order to analyze the relevant characteristics of the battery. For example, when performing battery sorting and matching through OCV testing, the measured OCV includes the influence of historical SOC, which reduces the accuracy of battery sorting and matching.

[0006] Therefore, the existing methods cannot accurately describe the true law of OCV changes with SOC, the accuracy of OCV curve reconstruction is low, the battery SOC cannot be estimated with high precision, and the relevant characteristics of the battery cannot be accurately analyzed in application scenarios that rely on OCV for battery electrochemical characteristics analysis. Summary of the invention

[0007] In view of the defects of the related art, the purpose of the present invention is to provide a method, device and storage medium for modeling the hysteresis characteristics of a battery, aiming to solve the problems of being unable to accurately describe the true law of OCV changes with SOC, low accuracy of OCV curve reconstruction, inability to estimate SOC with high accuracy, and inability to accurately analyze the relevant characteristics of the battery in application scenarios that rely on OCV for battery electrochemical characteristics analysis.

[0008] To achieve the above object, the present invention provides a method for modeling battery hysteresis characteristics, comprising:

[0009] S1. Modeling the battery hysteresis characteristics to obtain an initial model, and discretizing the initial model to obtain a discrete battery hysteresis characteristics model; the initial model is:

[0010]

[0011] OCV1t=∫∫ 0≤α＜β≤1 μα,βγ αβ [SoC t|dαdβ

[0012]

[0013] Among them, OCV max and OCV min are the maximum and minimum values ​​of OCV, the SOC of the battery is the model input, and the OCV of the battery is the model output. OCV1 t and OCV2 t represent the irreversible component and reversible component of OCV change, respectively. The irreversible component represents the OCV change caused by the hysteresis effect, and the reversible component represents the OCV change caused by the change of lithium concentration. γ αβ and γ αα is the hysteresis operator, γ αβ is a finite width, γ αα is zero width, μα, β and να are weight functions used to measure the contribution of each hysteresis operator to the output, α and β are the thresholds of the input in the descending and ascending directions respectively;

[0014] S2, measure the OCV and SOC when the battery charge and discharge direction changes only once, and obtain the first-order rotation curve of the battery;

[0015] S3. Perform parameter identification on a discrete battery hysteresis characteristic model based on the first-order rotation curve of the battery to obtain a target battery hysteresis characteristic model.

[0016] Optionally, the discretizing the initial model to obtain a discrete battery hysteresis characteristic model includes:

[0017] After discretizing the irreversible component OCV1t of the initial model, the expression is:

[0018]

[0019] Among them, α n ,β n and α n-1 ,β n For two adjacent charge-discharge turning points, Fα n-1 ,β n -Fα n ,β n is the contribution of two adjacent charge-discharge turning points to the total hysteresis effect, k is the total charge-discharge conversion order, each charge-discharge conversion order includes one charge-to-discharge and one discharge-to-charge, C1 is a constant term;

[0020] The reversible component OCV2t of the initial model is converted into a linear combination of weight functions by the trapezoidal method, and the expression is:

[0021]

[0022] Wherein, C2 is a constant term, p is the equal division of the complete SOC interval (i.e., 0≤SOC≤1), that is, the interval 0≤SOC≤11 is divided into p equal divisions, and i is the closest equal division point when SOCt is rounded down, that is, the closest i-th equal division point when SOCt is rounded down;

[0023] The discrete battery hysteresis characteristic model is:

[0024]

[0025] Among them, C represents a constant term.

[0026] Optionally, the discretizing the irreversible component OCV1 t of the initial model comprises:

[0027] The irreversible component OCV1t of the initial model is converted into an integral sum on a right-angled trapezoid through a geometric relationship, and then converted into a linear combination of integrals on a right-angled triangle with the charge and discharge turning point as a right-angle vertex and the straight line α=β as the hypotenuse;

[0028] Among them, α n ,β n is the coordinate of the right angle vertex in the Preisach triangle corresponding to the nth-order charge-discharge conversion, Fα n-1 ,β n The charge-discharge turning point α n-1 ,β n is the integral of the weight function on the right triangle with the right vertex and the straight line α = β as the hypotenuse, Fα n ,β n The charge-discharge turning point α n ,βn It is the integral of the weight function on the right triangle with the right vertex and the straight line α=β as the hypotenuse.

[0029] Optionally, S3 specifically includes:

[0030] S31, discretizing the weight functions μα, β in the discrete battery hysteresis characteristic model;

[0031] S32. Based on the experimental data obtained from the first-order rotation curve test of the battery, the constrained linear least squares method is used to perform parameter identification on the discretized weight function μα, β, and the weight function is solved to obtain the target battery hysteresis characteristic model.

[0032] Optionally, S32 specifically includes:

[0033] S321, setting the objective function to be the square sum of the errors between the measured OCV in the first-order rotation curve and the OCV predicted by the discrete battery hysteresis characteristic model;

[0034] S322, setting constraint conditions including boundary constraints and inequality constraints; wherein the boundary constraint is: vα≥0; the inequality constraint is: ∫∫ Tαβ μ,βdαd≥0; where Tα,β is a right triangle with α,β as right-angle vertices and the straight line α=β as the hypotenuse;

[0035] S323. Solve the discretized weight function in the discrete battery hysteresis characteristic model based on the objective function and the constraint conditions to obtain a target battery hysteresis characteristic model.

[0036] In a second aspect, the present invention further provides a battery SOC estimation method, comprising:

[0037] Establishing an equivalent circuit model of the battery to be tested and identifying model parameters, wherein the OCV in the equivalent circuit model is divided into a reversible component OCV2 and an irreversible component OCV1; wherein OCV1 is obtained by the historical charge and discharge turning points of the battery to be tested and a target battery hysteresis characteristic model, wherein the target battery hysteresis characteristic model is constructed by the method for modeling battery hysteresis characteristics as described in any one of the first aspects;

[0038] Obtaining the OCV2-SOC relationship of the battery to be tested according to the target battery hysteresis characteristic model;

[0039] A spatial state equation is established according to the equivalent circuit model, and the OCV2-SOC relationship OCV2=f(SOC) is used as the observation equation. The SOC of the battery to be tested is estimated by combining the state space equation and the observation equation through a filtering algorithm.

[0040] In a third aspect, the present invention further provides a battery matching method, comprising:

[0041] Obtaining an OCV2-SOC relationship of the battery to be tested according to the target battery hysteresis characteristic model; the target battery hysteresis characteristic model is constructed by using the method for modeling battery hysteresis characteristics as described in any one of the first aspects;

[0042] The difference between the OCV2 of different batteries under test at the same SOC is calculated according to the OCV2-SOC relationship of the batteries under test, and different batteries under test whose difference is less than a threshold are judged to be batteries under test with consistent performance, and are matched.

[0043] In a fourth aspect, the present invention further provides a device for modeling battery hysteresis characteristics, comprising:

[0044] The model building module is used to model the battery hysteresis characteristics to obtain an initial model, and discretize the initial model to obtain a discrete battery hysteresis characteristic model; the initial model is:

[0045]

[0046] OCV1 t=∫∫ 0≤α≤β≤1 μ,α,β,γ αβ SoC t|dαdβ

[0047]

[0048] Among them, OCV max and OCV min are the maximum and minimum values ​​of OCV, the SOC of the battery is the model input, and the OCV of the battery is the model output. OCV1 t and OCV2 t represent the irreversible component and reversible component of OCV change, respectively. The irreversible component represents the OCV change caused by the hysteresis effect, and the reversible component represents the OCV change caused by the change of lithium concentration. γ αβ and γ αα is the hysteresis operator, γ αβ is a finite width, γ αα is zero width, μα, β and vα are weight functions used to measure the contribution of each hysteresis operator to the output, α and β are the thresholds of the input in the descending and ascending directions respectively;

[0049] The curve acquisition module is used to measure the OCV and SOC when the battery charge and discharge direction changes only once, and obtain the battery first-order rotation curve;

[0050] The parameter identification module is used to perform parameter identification on the discrete battery hysteresis characteristic model based on the battery first-order rotation curve to obtain a target battery hysteresis characteristic model.

[0051] In a fifth aspect, the present invention further provides a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a processor to implement the method for modeling battery hysteresis characteristics as described in any one of the first aspects when executed.

[0052] Compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects:

[0053] 1. The present invention provides a method for modeling battery hysteresis characteristics. Before the classic Preisach model is used for modeling the OCV hysteresis phenomenon, necessary corrections are made to it, and an item is added to the model that is only related to the input at the current moment and has nothing to do with the historical path, namely, the reversible component. The battery hysteresis characteristic model based on the improved Preisach model can comprehensively reflect the two different mechanisms (reversible component and irreversible component) that drive the OCV change, and can accurately describe the real law of OCV change with SOC. At the same time, the reversible OCV value related only to the lithium concentration and the irreversible OCV value related only to the historical path of charge and discharge are obtained, thereby improving the reconstruction accuracy of OCV.

[0054] 2. The present invention provides a method for modeling battery hysteresis characteristics and gives a specific method for discretizing the battery hysteresis characteristic model, laying the foundation for integrating the model into the core algorithm system of the existing battery management system. The continuous battery hysteresis characteristic model cannot be directly deployed into the embedded system. The discretization of the battery hysteresis characteristic model will not only greatly reduce the computational overhead, but also help to simplify the parameter identification process of the model, and at the same time facilitate the deployment of the model into the embedded system.

[0055] 3. The present invention provides a battery SOC estimation method. The target battery hysteresis characteristic model constructed by the method for modeling the battery hysteresis characteristic provided by the present invention can obtain the OCV2-SOC relationship of the battery to be tested, and the OCV in the equivalent circuit model of the battery to be tested is divided into a reversible component OCV2 and an irreversible component OCV1, wherein OCV1 is obtained by the historical charge and discharge turning point of the battery to be tested and the target battery hysteresis characteristic model. A space state equation is established according to the equivalent circuit model, and the OCV2-SOC relationship OCV2=f(SOC) is used as the observation equation, and then the SOC estimation is completed by combining the state space equation and the observation equation through a filtering algorithm. The reversible OCV (OCV2) eliminates the influence of the hysteresis effect, and OCV2 corresponds to SOC one by one under different charge and discharge paths, making the SOC estimation of batteries (such as iron phosphate batteries) with serious hysteresis characteristics that cannot be ignored more accurate and reliable, while not increasing the hardware cost.

[0056] 4. By using the battery hysteresis characteristic model provided by the present invention, the OCV2 of the battery can be obtained on the battery production line through the target battery hysteresis characteristic model to screen out cells with consistent performance so as to form a battery pack with a high degree of balance. This helps to ensure the consistency and reliability of the battery pack and improve the overall performance and safety of the battery pack. The reversible OCV (OCV2) eliminates the influence of historical behavior and can better reflect the intrinsic electrochemical characteristics of the battery. The battery sorting and matching based on OCV2 is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of a method for modeling battery hysteresis characteristics provided by the present invention;

[0058] Figure 2 (a) is a schematic diagram of the main hysteresis loop of the lithium iron phosphate battery. Figure 2 (b) is a schematic diagram of a small hysteresis loop of a lithium iron phosphate battery;

[0059] Figure 3 It is a schematic diagram of the smallest unit that characterizes the hysteresis phenomenon in the Preisach model;

[0060] Figure 4 It is a schematic diagram of the Preisach triangle;

[0061] Figure 5 It is a schematic diagram of the Preisach triangle step memory curve;

[0062] Figure 6 It is a schematic diagram of the integral domain decomposition during the discretization of the initial model;

[0063] Figure 7 is a schematic diagram of a first-order rotation curve;

[0064] Figure 8 1 is a first-order rotation curve diagram obtained in the experiment, where (a) and (b) correspond to the first-order rotation curve of charge-to-discharge and the first-order rotation curve of discharge-to-charge;

[0065] Fig. 9 It is a first-order rotation curve reconstruction result diagram using parameter identification results. (a) and (b) in the figure respectively represent the first-order rotation curve reconstruction results of charge-to-discharge and discharge-to-charge. The solid line is the first-order rotation curve obtained by battery testing, and the single mark is the OCV predicted by the hysteresis characteristic model of the target battery. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions 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 intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0067] The contents involved in the above embodiment are described below in conjunction with a preferred embodiment.

[0068] Embodiment 1

[0069] like Figure 1 As shown, a method for modeling battery hysteresis characteristics includes:

[0070] S1. Modeling the battery hysteresis characteristics to obtain an initial model, and discretizing the initial model to obtain a discrete battery hysteresis characteristics model; the initial model is:

[0071]

[0072] OCV1 t=∫∫ 0≤α≤β≤1 μ,α,β,γ αβ SoC t|dαdβ

[0073]

[0074] Among them, OCV max and OCV min are the maximum and minimum values ​​of OCV, the SOC of the battery is the model input, and the OCV of the battery is the model output. OCV1 t and OCV2 t represent the irreversible component and reversible component of OCV change, respectively. The irreversible component represents the OCV change caused by the hysteresis effect, and the reversible component represents the OCV change caused by the change of lithium concentration. γ αβ and γ αα is the hysteresis operator, γ αβ is a finite width, γ αα is zero width, μα, β and α are weight functions used to measure the contribution of each hysteresis operator to the output, α and β are the thresholds of the input in the descending and ascending directions respectively;

[0075] S2, measure the OCV and SOC when the battery charge and discharge direction changes only once, and obtain the first-order rotation curve of the battery;

[0076] S3. Perform parameter identification on a discrete battery hysteresis characteristic model based on the first-order rotation curve of the battery to obtain a target battery hysteresis characteristic model.

[0077] Figure 2(a) shows the main hysteresis loop of the lithium iron phosphate battery tested in the laboratory. It can be clearly seen from the figure that the hysteresis effect causes the OCV-SOC relationship to become a multi-valued function. Figure 2 (b) shows the small hysteresis loop of the lithium iron phosphate battery. The solid line is the main hysteresis loop curve, which defines the outer boundary of the small hysteresis loop. Under the current condition of frequent alternation of charge and discharge, due to the existence of hysteresis effect, the OCV-SOC relationship will also form an up and down reciprocating motion trajectory in the main hysteresis loop, tending to the upper and lower boundaries during charging and discharging.

[0078] The principle of the classic Preisach model commonly used in existing methods is as follows: Figure 3 The hysteresis operator shown is the smallest unit that characterizes the hysteresis phenomenon in the Preisach model. The horizontal axis represents the system input and the vertical axis represents the system output. The output of the hysteresis operator has only two states, "1" and "-1". α and β are the thresholds of the input in the falling and rising directions respectively. The output state will only switch when the input crosses the threshold. Hysteresis operator γ αβ The input-output relationship can be expressed as

[0079]

[0080] The macroscopic hysteresis phenomenon can be considered as the result of the joint action of an infinite number of hysteresis operators with different thresholds, where the threshold of each hysteresis operator can be arbitrarily selected within a specific range, as long as they do not exceed the maximum value allowed in theory and α<β. For the OCV hysteresis phenomenon, the input is SOC, whose theoretical value range is 0≤SoC≤1, and the values ​​of the two thresholds must satisfy 0≤α<β≤1.

[0081] Figure 4 The isosceles right triangle shown is the Preisach triangle. For each coordinate point in the Preisach triangle, there is a corresponding hysteresis operator. If a weight function μα, β is used to measure the contribution of each hysteresis operator to the output, the output at any time can be obtained by solving a double integral in the Preisach triangle:

[0082] yt=∫∫ 0≤α＜β≤1 μ,βγ αβ [xt]dαdβ

[0083] Figure 5This is a schematic diagram of the Preisach triangle step memory curve. Assume that the battery starts charging from a fully discharged state. When the SOC increases to xt1, all hysteresis operators that satisfy β≤xt1 take the value of "1", while other hysteresis operators still take the value of "-1". At this time, the Preisach triangle is divided into two parts, upper and lower, by a straight line β=xt1 parallel to the α-axis. The values ​​of the hysteresis operators are "1" and "-1". As the charging process progresses, the upper and lower boundaries will continue to move upward. If at a certain point in the future the battery is not fully charged and turns to discharge, the SOC gradually drops to xt2. At this time, all hysteresis operators that satisfy xt2≤α take the value of "-1". A straight line α=xt2 parallel to the β-axis appears in the Preisach triangle as the left and right boundaries, forming Figure 5 Boundary lines shown.

[0084] As the discharge process continues, the left and right boundaries will continue to shift to the left. After the next charge-discharge conversion, the battery starts charging: 1) If the SOC does not exceed the historical extreme value xt1 during the charging process, the boundary is in a step shape; 2) If the SOC exceeds the historical extreme value xt1 during the charging process, the historical boundary is erased and a new straight boundary is re-formed.

[0085] For some physical systems, the underlying principle of the relationship between input and output includes two parts. On the one hand, the output size is directly related to the current input size. On the other hand, due to different reaction paths and other reasons, the output size shows historical input dependence, that is, the output size is not only related to the current input size, but also to the historical input. Directly using the Preisach model can obtain the output result more accurately, but the output result includes these two aspects and cannot be further distinguished for analysis. For this reason, the Preisach model is improved to ensure that the improved Preisach model can fully reflect the two different mechanisms that drive output changes.

[0086] The improved Preisach model is:

[0087] yt=y1t+y2t

[0088] y1 t=∫∫ 0≤α＜β≤1 μ,βγ αβ [xt]dαdβ

[0089]

[0090] Among them, γ αβ and γ αα is the hysteresis operator, γ αβ is a finite width, γ ααis zero width, μα, β and vα are weight functions used to measure the contribution of each hysteresis operator to the output, α and β are the thresholds of the input in the descending and ascending directions respectively, xt is the model input, yt is the model output, y1 t is the first output item, describing the part of the output that is only related to the historical input, and y2 t is the second output item, describing the component of the output that is only related to the input.

[0091] Compared with the classic Preisach model, the improved Preisach model only adds an integral term to describe the component in the output that is independent of the historical input. At the same time, the basic structure and physical meaning of the second term and the first term remain consistent, but the integral dimension is degenerated from two dimensions to one dimension. From the perspective of the hysteresis operator, the finite width hysteresis operator γ αβ Degenerates into a zero-width hysteresis operator γ αα , when the input xt exceeds the threshold α, γ αα The output state will mutate. From the perspective of weight function, the two-dimensional weight function μα,β degenerates into a one-dimensional weight function να, and the zero-width hysteresis operator γ αα Continuously distributed in the range of 0≤α≤1.

[0092] In the above model, y1 t is the first output term, that is, the traditional Preisach model. According to the memory mechanism of the Preisach model, when the two-dimensional weight function μα,β on the hypotenuse of the Preisach triangle takes 0, its output size only includes the hysteresis effect, that is, the part related to the historical input. y2 t is the second output term, and the second integral is used to describe the component of the output that is only related to the current input. Therefore, in the modified improved Preisach model, the first and second integrals can respectively model the irreversible and reversible components in the output change, thereby fundamentally solving the problem that the classical Preisach model cannot accurately match the output evolution mechanism.

[0093] The battery hysteresis characteristics are modeled based on the improved Preisach model to obtain the initial model; the initial model contains the irreversible component OCV1 t and the reversible component OCV2 t in the OCV change; the irreversible component OCV1 t represents the OCV change caused by the hysteresis effect, and its size is only related to the historical SOC, but not to the current SOC; the reversible component OCV2 t represents the OCV change caused by the change in lithium concentration, and its size is only related to the current SOC, but not to the historical SOC. In order to facilitate subsequent calculations, the initial model is discretized and simplified to obtain a discrete battery hysteresis characteristic model. The OCV and SOC are measured when the battery charge and discharge direction changes only once, and the first-order rotation curve of the battery is obtained. Based on the obtained first-order rotation curve of the battery, the discrete battery hysteresis characteristic model is parameter identified to confirm the unknown parameters in the model, thereby obtaining the target battery hysteresis characteristic model.

[0094] In the above initial model, the first integral term is based on the classic Preisach model, and its output size OCV1 t is used to describe the component of OCV change that is only related to the charge and discharge path. The second integral term is used to describe the component of OCV change that is only related to the input size. It maintains the same basic structure and physical meaning as the first term, but the integral dimension is degenerated from two dimensions to one dimension. From the perspective of the hysteresis operator, the finite width hysteresis operator γ αβ Degenerates into a zero-width hysteresis operator γ αα , when the input SOC t exceeds the threshold α, γ αα The output state will change suddenly. According to the memory mechanism of the improved Preisach model, during the charging and discharging process, the second integral is only related to the SOC at the current moment.

[0095] The continuous battery hysteresis characteristic model cannot be directly deployed into the embedded system. Discretizing the model will not only greatly reduce the computational overhead, but also help simplify the model parameter identification process.

[0096] The Preisach triangle is divided into two parts, where the output states of the two-dimensional hysteresis operator are "1" and "-1", respectively denoted as S + t and S - t, T represents the area covered by the Preisach triangle, then T = S + t∪S - t. It can be obtained that the term in the model that changes with time is the integral of the part of the output state of the two-dimensional hysteresis operator that is "1", that is, S + The integral on t, the boundary of this region is step-shaped, such as Figure 6 As shown, the integral on this area is simplified into several trapezoidal areas Q through geometric relations. nThe sum of the integrals on the trapezoid is converted into the difference of the integrals on the two right triangles through geometric relationships. Each right triangle can be uniquely determined by the right-angle vertex, that is, the turning point of charge and discharge. Therefore, α n ,β n The integral on the right triangle with vertices is expressed as:

[0097]

[0098] Therefore, after discretizing the irreversible component OCV1t in the initial model, the expression is:

[0099]

[0100] Among them, C1 is a constant term.

[0101] Whether the last step of the battery operation is discharge or charge will determine whether the last segment of the step boundary is vertical or horizontal. In the above derivation, it is assumed that the current condition ends in discharge. At this time, α k =SOC t. On the contrary, when the current condition ends during charging, α k =β k =SOC t.

[0102] The complete SOC interval (i.e., 0≤SOC≤1) is divided into two parts, where the output states of the one-dimensional hysteresis operator are “1” and “-1”, respectively denoted as L + t and L - t,[0,SOC t] is L + t, [SOC t, 1] is Lt. It can be obtained that the term that changes with time in the model is the integral of the part of the output state of the one-dimensional hysteresis operator that is "1", that is, L + The integral over t.

[0103] Furthermore, L + The integral discretization on t can be achieved by the trapezoidal method. Divide the complete SOC interval (i.e., 0≤SOC≤1) into p equal parts. Assuming that when SOC t is rounded down, the closest equal division point is the i-th equal division point in the interval (i.e., i / p), then the expression after discretizing the reversible component in the initial model is:

[0104]

[0105] Among them, C2 is a constant term.

[0106] The discrete battery hysteresis characteristic model is:

[0107]

[0108] Among them, C represents a constant term.

[0109] Figure 7 is a schematic diagram of the first-order conversion curve. "First-order" means that the direction of the model input changes only once, that is, only one charge-discharge conversion occurs. The battery starts to charge continuously from the full discharge state until the state of charge reaches SOC1, and then turns to discharge until the state of charge reaches SOC2. The open circuit voltages corresponding to SOC1 and SOC2 are recorded as OCV1 and OCV2 respectively.

[0110] During the charge-discharge conversion process, the hysteresis operator in the triangle T SOC2,SOC1 will change the output, and we can get:

[0111]

[0112] In the above formula, the left side of the equation can be obtained through experiments, while the right side of the equation is a linear function of the discrete battery hysteresis characteristic model parameters.

[0113] Figure 8 This is the first-order conversion curve of charge to discharge obtained in the laboratory. The test method is to first charge at constant current and constant voltage to a fully charged state and let it stand, and then pulse discharge to a fully discharged state. The ΔSOC of a single pulse discharge process is 5%. It is cycled 20 times, and the SOC threshold of each constant current and constant voltage charging process is lowered by 5%.

[0114] The steps for obtaining the first-order rotation curve of the battery are as follows:

[0115] Charge to discharge process:

[0116] 1. Constant current and constant voltage charging to full charge state: charging current is 8A, cut-off voltage is 3.65V, cut-off current is 0.8A, and standing time is 1h;

[0117] 2. Pulse discharge to full discharge state: discharge current is 8A, cut-off voltage is 2V, ΔSOC of a single pulse discharge process is 5%, standing time is 1h, and cycled 20 times;

[0118] 3. Charge with constant current and constant voltage until SOC=95%, then pulse discharge to full discharge state after standing for 1 hour;

[0119] 4. The SOC threshold of the constant current and constant voltage charging process is lowered by 5%, and step 3 is repeated for 19 times.

[0120] Discharge to charge process:

[0121] 1. Constant current discharge to full discharge state: discharge current is 8A, cut-off voltage is 2V, and standing time is 1h;

[0122] 2. Pulse charging to full charge state: charging current is 8A, cut-off voltage is 3.65V, ΔSOC of a single pulse charging process is 5%, standing time is 1h, and cycled 20 times;

[0123] 3. Constant current discharge to SOC = 5%, then pulse charge to full charge state after standing for 1 hour;

[0124] 4. The SOC threshold of the constant current discharge process is increased by 5%, and step 3 is repeated for 19 times.

[0125] Based on the experimental data obtained from the above charging and discharging processes, the first-order rotation curve of the battery is calculated.

[0126] Optionally, S3 specifically includes:

[0127] S31, discretizing the weight functions μα, β in the discrete battery hysteresis characteristic model;

[0128] S32. Based on the experimental data obtained from the first-order rotation curve test of the battery, the constrained linear least squares method is used to perform parameter identification on the discretized weight function μα, β, and the weight function is solved to obtain the target battery hysteresis characteristic model.

[0129] In S31, the continuous weight function μα, β is discretized, specifically including: taking a discrete step length L = 0.05, dividing the horizontal axis and the vertical axis into 20 equal parts, drawing a uniform grid in the Preisach triangle, and dividing the grid point α i ,β j The weight function at is denoted as μα i ,β j , so as to convert the continuous integral in the above formula into a numerical integral, the voltage difference of different data points in the OCV first-order rotation curve is expressed as a linear weighted superposition of discrete weight functions.

[0130] Optionally, S32 specifically includes:

[0131] S321, setting the objective function to be the square sum of the errors between the measured OCV in the first-order rotation curve and the OCV predicted by the discrete battery hysteresis characteristic model;

[0132] S322, setting constraint conditions including boundary constraints and inequality constraints; wherein the boundary constraint is: 0; the inequality constraint is: ∫∫ Tα,β μ,α,βdαdβ≥0;

[0133] S323. Solve the discretized weight function in the discrete battery hysteresis characteristic model based on the objective function and the constraint conditions to obtain a target battery hysteresis characteristic model.

[0134] The constraints for constrained linear least squares are divided into two parts:

[0135] 1) Boundary constraints. The functional relationship between the ideal OCV and SOC is a monotonically increasing function, so να≥00 always holds true.

[0136] 2) Inequality constraints. The OCV in the charging path is higher than the OCV in the discharging path, that is, at any SOC, the hysteresis voltage is positive during charging and negative during discharging. For any α and β in the definition domain, the integral ∫∫ Tα,β ,μ,α,βdαdβ≥0 is always true; where Tα,β is a right triangle with α,β as right-angle vertices and the straight line α=β as the hypotenuse.

[0137] The solution can be completed using MATLAB's lsqlin function to obtain the target battery hysteresis characteristic model.

[0138] In this embodiment, the first-order conversion curve of discharge-to-charge is reconstructed by using the parameter identification result of the first-order conversion curve of charge-to-discharge; the first-order conversion curve of charge-to-discharge is reconstructed by using the parameter identification result of the first-order conversion curve of discharge-to-charge. Fig. 9 In the figure, the solid line is the first-order rotation curve obtained from the battery test, and the single mark is the OCV predicted by the target battery hysteresis characteristic model. It can be seen that the two have a high degree of overlap. The root mean square error (RMSE) and mean absolute error (MAE) of the OCV prediction are shown in Table 1, which shows that the model has a high accuracy.

[0139] Table 1 OCV prediction accuracy in the first-order rotation curve

[0140]

[0141] Furthermore, on the basis of the above-mentioned embodiments, the application of the battery hysteresis characteristic model based on the improved Preisach model includes two aspects: 1. Battery SOC estimation, 2. Battery sorting and matching. When estimating the battery SOC, the OCV2-SOC relationship is obtained in advance through the target battery hysteresis characteristic model, and an equivalent circuit model is established to estimate the battery SOC in combination with the OCV2-SOC relationship. When sorting and matching batteries, the OCV2 is accurately obtained to analyze the relevant electrochemical characteristics of the battery, and the OCV2 of the battery is obtained through the target battery hysteresis characteristic model to screen out cells with consistent performance, thereby performing battery sorting and matching.

[0142] The present invention also provides a battery SOC estimation method, which adopts a target battery hysteresis characteristic model constructed by the above-mentioned method for modeling battery hysteresis characteristics, including:

[0143] Obtaining the OCV2-SOC relationship of the battery to be tested according to the target battery hysteresis characteristic model;

[0144] Establishing an equivalent circuit model of the battery to be tested and identifying model parameters, wherein the OCV in the equivalent circuit model is divided into a reversible component OCV2 and an irreversible component OCV1; wherein OCV1 is obtained from the historical charge and discharge turning points of the battery to be tested and the hysteresis characteristic model of the target battery;

[0145] A spatial state equation is established according to the equivalent circuit model, and the OCV2-SOC relationship OCV2=F(SOC) is used as the observation equation. The SOC estimation is completed by combining the state space equation and the observation equation through a filtering algorithm.

[0146] Among them, according to the accuracy and complexity of the model, a first-order Thevenin equivalent circuit model can be selected, and the first-order equivalent circuit model can be processed and converted into a form that is conducive to model identification. According to the speed and accuracy of the parameter identification algorithm, the model parameter identification can be obtained by using a recursive least squares algorithm with a forgetting factor.

[0147] The reversible component OCV2 eliminates the influence of hysteresis effect. Under different paths, OCV2 corresponds to SOC one by one, thereby improving the accuracy of SOC estimation.

[0148] The present invention also provides a battery matching method, comprising:

[0149] Obtaining the OCV2-SOC relationship of the battery to be tested according to the target battery hysteresis characteristic model; the target battery hysteresis characteristic model is constructed by using the method for modeling battery hysteresis characteristics as described in any one of the above embodiments;

[0150] The difference between the OCV2 of different batteries under test at the same SOC is calculated according to the OCV2-SOC relationship of the batteries under test, and different batteries under test whose difference is less than a threshold are judged to be batteries under test with consistent performance, and are matched.

[0151] The target battery hysteresis characteristic model constructed by the method of modeling the battery hysteresis characteristic can be used on the battery production line to obtain the OCV2 of the battery to be tested at a certain SOC value. The closeness of the OCV2 at the same SOC is used as a screening criterion or one of the screening criteria. When other conditions are the same, the closer the OCV2 of the battery at the same SOC is, the higher the consistency of the two batteries is considered to be, so as to screen out cells with consistent performance to form a battery pack with a higher degree of balance. It helps to ensure the consistency and reliability of the battery pack and improve the overall performance and safety of the battery pack. The reversible OCV (OCV2) eliminates the influence of historical behavior and can better reflect the intrinsic electrochemical characteristics of the battery. The battery sorting and matching based on OCV2 is more accurate.

[0152] In this embodiment, the improved Preisach model is used to model the battery hysteresis characteristics, which solves the problem that the real law of OCV changing with SOC cannot be accurately described, the accuracy of OCV curve reconstruction is low, the battery SOC cannot be estimated with high accuracy, and the relevant characteristics of the battery cannot be accurately analyzed in the application scenario that relies on OCV for battery electrochemical characteristics analysis. An item that is only related to the SOC at the current moment and has nothing to do with the historical path is added to the model, so that the model can fully reflect the two different mechanisms that drive the OCV change, improve the accuracy of the battery hysteresis characteristic model, and when reconstructing the OCV curve, the reversible OCV component and the irreversible OCV component can be obtained at the same time. The reversible OCV component is only related to SOC and is used for SOC estimation, which can make the SOC estimation of batteries with serious hysteresis characteristics that cannot be ignored more accurate and reliable. In the fields of battery sorting and matching, the reversible OCV component OCV2 can more accurately reflect the intrinsic electrochemical characteristics of the battery, thereby making the battery sorting and matching based on OCV2 more accurate.

[0153] Embodiment 2

[0154] The present invention also provides a device for modeling battery hysteresis characteristics, comprising:

[0155] The model building module is used to model the battery hysteresis characteristics to obtain an initial model, and discretize the initial model to obtain a discrete battery hysteresis characteristic model; the initial model is:

[0156]

[0157] OCV1 t=∫∫ 0≤α≤β≤1 μ,α,βγ αβ [SOC t]dαdβ

[0158]

[0159] Among them, OCV max and OCV min are the maximum and minimum values ​​of OCV, the SOC of the battery is the model input, and the OCV of the battery is the model output. OCV1 t and OCV2 t represent the irreversible component and reversible component of OCV change, respectively. The irreversible component represents the OCV change caused by the hysteresis effect, and the reversible component represents the OCV change caused by the change of lithium concentration. γ αβ and γ αα is the hysteresis operator, γ αβ is a finite width, γ αα is zero width, μα, β and vα are weight functions used to measure the contribution of each hysteresis operator to the output, α and β are the thresholds of the input in the descending and ascending directions respectively;

[0160] The curve acquisition module is used to measure the OCV and SOC when the battery charge and discharge direction changes only once, and obtain the battery first-order rotation curve;

[0161] The parameter identification module is used to perform parameter identification on the discrete battery hysteresis characteristic model based on the battery first-order rotation curve to obtain a target battery hysteresis characteristic model.

[0162] An apparatus for modeling battery hysteresis characteristics provided by an embodiment of the present invention is used to execute a method for modeling battery hysteresis characteristics provided by any embodiment of the present invention, and has corresponding beneficial effects.

[0163] Embodiment 3

[0164] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for modeling battery hysteresis characteristics as described in any one of the first embodiments when executed.

[0165] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for modeling battery hysteresis characteristics, characterized in that: include: S1. Modeling the battery hysteresis characteristics to obtain an initial model, and discretizing the initial model to obtain a discrete battery hysteresis characteristics model; the initial model is: OCV1t=∫∫ 0≤α≤β≤1 but, bg αβ [SOC t]dαdβ Among them, OCV max and OCV min are the maximum and minimum values ​​of OCV, the SOC of the battery is the model input, and the OCV of the battery is the model output. OCV1 t and OCV2 t represent the irreversible component and reversible component of OCV change, respectively. The irreversible component represents the OCV change caused by the hysteresis effect, and the reversible component represents the OCV change caused by the change of lithium concentration. γ αβ and γ αα is the hysteresis operator, γ αβ is a finite width, γ αα is zero width, μα, β and να are weight functions used to measure the contribution of each hysteresis operator to the output, α and β are the thresholds of the input in the descending and ascending directions respectively; S2, measure the OCV and SOC when the battery charge and discharge direction changes only once, and obtain the first-order rotation curve of the battery; S3. Perform parameter identification on a discrete battery hysteresis characteristic model based on the first-order rotation curve of the battery to obtain a target battery hysteresis characteristic model.

2. The method according to claim 1, characterized in that The discretization of the initial model to obtain a discrete battery hysteresis characteristic model includes: After discretizing the irreversible component OCV1t of the initial model, the expression is: Among them, α n ,β n and α n-1 ,β n For two adjacent charge-discharge turning points, Fα n-1 ,β n -Fα n ,β n is the contribution of two adjacent charge-discharge turning points to the total hysteresis effect, k is the total charge-discharge conversion order, each charge-discharge conversion order includes one charge-to-discharge and one discharge-to-charge, C1 is a constant term; The reversible component OCV2t of the initial model is converted into a linear combination of weight functions by the trapezoidal method, and the expression is: Wherein, C2 is a constant term, p is the equal division of the complete SOC interval (i.e., 0≤SOC≤1), that is, the interval 0≤SOC≤1 is divided into p equal divisions, and i is the closest equal division point when SOCt is rounded down, that is, the closest i-th equal division point when SOCt is rounded down; The discrete battery hysteresis characteristic model is: Among them, C represents a constant term.

3. The method according to claim 2, characterized in that The discretizing the irreversible component OCV1 t of the initial model comprises: The irreversible component OCV1t of the initial model is converted into an integral sum on a right-angled trapezoid through a geometric relationship, and then converted into a linear combination of integrals on a right-angled triangle with the charge and discharge turning point as a right-angle vertex and the straight line α=β as the hypotenuse; Among them, α n ,β n is the coordinate of the right angle vertex in the Preisach triangle corresponding to the nth-order charge-discharge conversion, Fα n-1 ,β n The charge-discharge turning point α n-1 ,β n is the integral of the weight function on the right triangle with the right vertex and the straight line α = β as the hypotenuse, Fα n ,β n The charge-discharge turning point α n ,β n It is the integral of the weight function on the right triangle with the right vertex and the straight line α=β as the hypotenuse.

4. The method according to claim 1, characterized in that S3 specifically includes: S31, discretizing the weight functions μα, β in the discrete battery hysteresis characteristic model; S32. Based on the experimental data obtained from the first-order rotation curve test of the battery, the constrained linear least squares method is used to perform parameter identification on the discretized weight function μα, β, and the discretized weight function is solved to obtain the target battery hysteresis characteristic model.

5. The method according to claim 4, characterized in that S32 specifically includes: S321, setting the objective function to be the square sum of the errors between the measured OCV in the first-order rotation curve and the OCV predicted by the discrete battery hysteresis characteristic model; S322. Set constraint conditions including boundary constraints and inequality constraints; wherein the boundary constraint is: να≥0; the inequality constraint is: ∫∫ Tα，β μα, βdαdβ≥0; where Tα,β is a right triangle with α,β as right-angle vertices and the straight line α=β as the hypotenuse; S323. Solve the discretized weight function in the discrete battery hysteresis characteristic model based on the objective function and the constraint conditions to obtain a target battery hysteresis characteristic model.

6. A battery SOC estimation method, characterized in that: include: Establish an equivalent circuit model of the battery to be tested and identify model parameters, wherein the OCV in the equivalent circuit model is divided into a reversible component OCV2 and an irreversible component OCV1; wherein OCV1 is obtained by the historical charge and discharge turning points of the battery to be tested and a target battery hysteresis characteristic model, and the target battery hysteresis characteristic model is constructed by the method for modeling battery hysteresis characteristics as described in any one of claims 1 to 5; Obtaining the OCV2-SOC relationship of the battery to be tested according to the target battery hysteresis characteristic model; A spatial state equation is established according to the equivalent circuit model, and the OCV2-SOC relationship OCV2=f(SOC) is used as the observation equation. The SOC of the battery to be tested is estimated by combining the state space equation and the observation equation through a filtering algorithm.

7. A battery matching method, characterized in that: include: Obtaining the OCV2-SOC relationship of the battery to be tested according to the target battery hysteresis characteristic model; The target battery hysteresis characteristic model is constructed by using the method for modeling battery hysteresis characteristics as described in any one of claims 1 to 5; The difference between the OCV2 of different batteries under test at the same SOC is calculated according to the OCV2-SOC relationship of the batteries under test, and different batteries under test whose difference is less than a threshold are judged to be batteries under test with consistent performance, and are matched.

8. A device for modeling battery hysteresis characteristics, characterized in that: include: The model building module is used to model the battery hysteresis characteristics to obtain an initial model, and discretize the initial model to obtain a discrete battery hysteresis characteristic model; the initial model is: OCV1t=∫∫ 0≤α≤β≤1 but, bg αβ [SOC t]dαdβ Among them, OCV max and OCV min are the maximum and minimum values ​​of OCV, the SOC of the battery is the model input, and the OCV of the battery is the model output. OCV1 t and OCV2 t represent the irreversible component and reversible component of OCV change, respectively. The irreversible component represents the OCV change caused by the hysteresis effect, and the reversible component represents the OCV change caused by the change of lithium concentration. γ αβ and γ αα is the hysteresis operator, γ αβ is a finite width, γ αα is zero width, μα, β and να are weight functions used to measure the contribution of each hysteresis operator to the output, α and β are the thresholds of the input in the descending and ascending directions respectively; The curve acquisition module is used to measure the OCV and SOC when the battery charge and discharge direction changes only once, and obtain the battery first-order rotation curve; The parameter identification module is used to perform parameter identification on the discrete battery hysteresis characteristic model based on the battery first-order rotation curve to obtain a target battery hysteresis characteristic model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for modeling battery hysteresis characteristics according to any one of claims 1 to 5 when the computer instructions are executed.

Images