CN105203963A

CN105203963A - Charge state estimation method based on open-circuit voltage hysteretic characteristics

Info

Publication number: CN105203963A
Application number: CN201510578190.XA
Authority: CN
Inventors: 戴海峰; 魏学哲; 朱乐涛; 孙泽昌
Original assignee: Tongji University
Current assignee: Tongji University
Priority date: 2015-09-11
Filing date: 2015-09-11
Publication date: 2015-12-30
Anticipated expiration: 2035-09-11
Also published as: CN105203963B

Abstract

The invention relates to a charge state estimation method based on open-circuit voltage hysteretic characteristics. The method is used for estimating the charge state of a lithium ion battery online. The method comprises the following steps that 1, a hysteretic characteristic curve of open-circuit voltage and the charge state of the lithium ion battery is obtained offline; 2, initial parameters of an open-circuit voltage hysteretic characteristic self-adaptive model based on a Preisach operator is determined through training according to the hysteretic characteristic curve, and the open-circuit voltage hysteretic characteristic self-adaptive model based on the Preisach operator is established; 3, the charge state of the lithium ion battery is estimated online according to the open-circuit voltage hysteretic characteristic self-adaptive model based on the Preisach operator, and the charge state of the lithium ion battery at the present moment is obtained. Compared with the prior art, the charge state estimation method based on the open-circuit voltage hysteretic characteristics has the advantages of being accurate in modeling, capable of improving the accuracy, accurate in estimation and the like.

Description

Method for estimating state of charge based on open-circuit voltage hysteresis characteristic

Technical Field

The invention relates to a state of charge estimation method, in particular to a state of charge estimation method based on open-circuit voltage hysteresis characteristics.

Background

The power battery system is used as a key part and is increasingly applied to the fields of electric automobiles, power energy storage and the like. In the application process, a Battery Management System (BMS) is required to monitor the battery state, prevent overcharge and overdischarge, and prolong the service life of the battery. Among these, accurate estimation of SOC (state of charge) is particularly critical. Most SOC estimation methods are obtained by using a correspondence relationship between SOC and open circuit voltage OCV (open circuit voltage), such as an open circuit voltage method and a model-based SOC estimation method. The description of the OCV and SOC correspondence is the core basis of these SOC estimation methods. The open-circuit voltage and the OCV in the lithium ion battery are not completely in one-to-one correspondence, but have a hysteresis relationship (the OCV in the charging process is greater than the OCV in the discharging process under the same SOC).

The traditional modeling method for the open-circuit voltage hysteresis characteristic of the lithium ion battery introduces more simplification, so that the modeling precision of a hysteresis model is low, and SOC estimation is influenced. The hysteresis model modeling method based on the Preisach operator is already applied to open-circuit voltage hysteresis modeling of a nickel-metal hydride battery (NiMH), but the discrete Preisach model cannot be well applied to the lithium ion battery because the open-circuit voltage hysteresis relation of the lithium ion battery is not obvious and has no symmetry compared with the NiMH battery.

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provide the state of charge estimation method based on the hysteresis characteristic of the open-circuit voltage, which is accurate in modeling, accuracy improving and estimation.

The purpose of the invention can be realized by the following technical scheme:

a state of charge estimation method based on open-circuit voltage hysteresis characteristics is used for estimating the state of charge of a lithium ion battery on line and comprises the following steps:

1) obtaining a hysteresis characteristic curve of the open circuit voltage and the charge state of the lithium ion battery in an off-line manner;

2) determining initial parameters of the open-circuit voltage hysteresis characteristic adaptive model based on the Preisach operator according to the hysteresis characteristic curve training, and establishing the open-circuit voltage hysteresis characteristic adaptive model based on the Preisach operator;

3) and estimating the charge state of the lithium ion battery on line according to the open-circuit voltage hysteresis characteristic self-adaptive model based on the Preisach operator to obtain the charge state of the lithium ion battery at the current moment.

The initial parameters of the open-circuit voltage hysteresis characteristic adaptive model based on the Preisach operator in the step 2) comprise the mesh division number N of the Preisach triangle and the initial value mu of the weight vector.

The open-circuit voltage hysteresis characteristic self-adaptive model based on the Preisach operator in the step 2) is as follows:

wherein,is t_kState of charge value, ω (t), corresponding to the open circuit voltage at that moment_k) Is t_kThe hysteresis state vector, mu (t), corresponding to the open circuit voltage value in the Preisach triangle at the moment_k) Is t_kThe hysteresis weight vectors of all meshes in the time preiach triangle.

The step 3) specifically comprises the following steps:

31) on-line acquisition of last time t_k-1Weight vector mu (t) in the Preisach triangular mesh of_k-1) And the current time t_kValue of hysteresis state ω (t)_k) Calculating to obtain the prior state of charge value at the current momentComprises the following steps:

32) according to battery capacity Q₀Last moment t_k-1State of charge value ofAnd the current time t_kPrior state of charge value ofCalculating to obtain a current estimation value I at the current moment_cal(t_k) Comprises the following steps:

Δt＝t_k-t_k-1；

33) according to the actual current value I of the current moment measured actually_m(t_k) And current estimate I_cal(t_k) Obtaining the current error value eta _ current (t) at the current moment_k) Comprises the following steps:

η_current(t_k)＝I_m(t_k)-I_cal(t_k)；

34) according to the current error value eta _ current (t) of the current moment_k) And prior state of charge value at the current timeAnd obtaining the current time weight vector increment by adopting a minimum mean square error methodAnd calculates a weight value mu (t) of the current time_k) Comprises the following steps:

μ(t_k)＝μ(t_k-1)+λη(t_k)ω(t_k)

wherein lambda is a step factor, and lambda belongs to [0,1 ];

35) according to the weight vector mu (t) of the current time_k-1) And a hysteresis state value ω (t)_k) Obtaining the posterior charge state value of the current moment through the open-circuit voltage hysteresis characteristic self-adaptive modelI.e., the state of charge of the lithium ion battery at the present time, and returns to step 31) to perform the state of charge estimation of the lithium ion battery at the next time.

In the step 34), the weight vector increment at the current momentThe calculation formula of (A) is as follows:

wherein λ is the step lengthFactor, and λ ∈ [0,1]]，η(t_k) Is t_kError values, ω (t), of the calculated and measured currents at times_k) Is t_kThe hysteretic state vector in the triangle at time Preisach.

Compared with the prior art, the invention has the following advantages:

firstly, accurate modeling: according to the method, errors of the measured current and the calculated current are introduced, the weight values corresponding to the grids in the Preisach triangle are adjusted in a self-adaptive mode at each moment, and the open-circuit voltage hysteresis characteristics of the lithium ion battery are accurately modeled through iterative calculation of the prior charge state value, the hysteresis state value and the current actual value.

Secondly, improving the precision: according to the invention, the modeling precision is improved through the improvement of the algorithm, the measured current which plays a core role in the algorithm can be directly obtained through the original current sensor in the battery management system, and the hardware cost is not increased while the precision is improved.

Thirdly, accurate estimation: according to the method, the charge state of the lithium ion battery with serious hysteresis characteristics (such as a lithium iron phosphate battery) can be estimated more accurately and reliably by estimating the charge state of the lithium ion battery with hysteresis characteristics.

Drawings

Fig. 1 is a schematic diagram of an OCV-SOC hysteresis characteristic curve of a lithium ion battery.

Fig. 2 is a diagram of the basic preiach operator.

FIG. 3 is a diagram of Preisach triangle and ladder memory curves.

FIG. 4 is a schematic diagram of a discrete Preisach triangle and mesh.

FIG. 5 is a flow chart of lithium ion battery OCV-SOC hysteresis model training.

FIG. 6 is a flow chart of an application of an OCV-SOC adaptive discrete Preisach model of a lithium ion battery.

Detailed Description

The invention is described in detail below with reference to the figures and specific embodiments.

Example (b):

the invention mainly aims to establish an accurate modeling method for hysteresis characteristics of open-circuit voltage of a lithium ion battery, so that the method can be finally used for accurately estimating the SOC of the lithium ion battery.

In order to achieve the above objects and other advantages of the present invention, as embodied and broadly described herein, there is provided a modeling method for an open-circuit voltage hysteresis characteristic of a lithium ion battery based on current regulation and a discrete preiach operator, where the conventional modeling method for an open-circuit voltage hysteresis characteristic has a disadvantage that a large error and a high precision are difficult to guarantee, and is not suitable for the present invention, the present invention takes a weight value corresponding to a grid triangularly divided by preiach as a time variable on the basis of modeling the hysteresis characteristic by the discrete preiach operator, obtains a current deviation value by obtaining a priori SOC defined at each time and a current value in the battery, obtains an increment of a change in the weight value at each time by combining a least mean square error theory (LMS), obtains a weight value at a current time by summing the weight value at the previous time, and further obtains a hysteresis state at the current time to obtain a posteriori SOC output at the current time, the initial values of the number of grid divisions and the weight vector in the model are determined off-line through an OCV-SOC hysteresis curve of the lithium ion battery.

According to a preferred embodiment of the invention, the complete implementation steps are as follows: 1) obtaining an OCV-SOC hysteresis characteristic curve of the lithium ion battery in an off-line manner in an experiment; 2) determining the number of Preisach triangular mesh divisions and the initial value of the corresponding weight vector in an off-line manner; 3) when the online application is carried out, the weight value corresponding to the Preisach triangular grid is regarded as time-varying, and the current value is obtained by adding the weight value at the previous moment and the variable quantity at the current moment; 4) changing the hysteresis state value of each grid of the Preisach triangle according to the OCV value at the current moment, and obtaining a prior SOC value at the current moment according to the weight value at the previous moment; 5) obtaining a current estimation value at the current moment by the prior SOC value, the SOC value at the last moment and the battery capacity, and comparing the current estimation value with the actual current to obtain an error value; 6) the current error value at the current moment is combined with a least mean square error (LMS) theory to obtain the increment of the weight value at the current moment and the weight value at the current moment; 7) and combining the weighted value of the current moment with the hysteresis state value of the current moment to obtain a posterior SOC value of the current moment, namely the SOC value of the current moment, and repeating the process at the next moment.

Fig. 1 shows an experimental lithium ion battery OCV-SOC hysteresis characteristic curve (OCV is fully static for more than 3 hours) for determining initial parameters of the hysteresis model provided by the present invention, in order to fully cover the hysteresis characteristic of the lithium ion battery, during the experimental measurement, the battery is discharged from a fully charged state, and OCV values are recorded every 5% SOC change and after static placement until SOC is equal to 0 fully discharged state; then entering a charging state, changing every 5% of SOC, standing, recording an OCV value until the SOC is 95%, converting to discharge, and changing to charge until the SOC is 5%; the process is circulated and ensures that each state of charge is reduced by 5% compared with the SOC value of the last state of charge, and each state of discharge is increased by 5% compared with the SOC value of the last state of discharge until the SOC is 50% of the end of the experiment; throughout the process, OCV was recorded after each 5% change in SOC.

As shown in fig. 2-4, the preiach operators (α, β) describe a relationship between input and output, that is, the output is 1 when the input is greater than the threshold α, the output is-1 when the input is less than the threshold β (β < α), and the output is constant when the input is between the thresholds α and β, because the input has a finite amplitude in the hysteresis relationship, as shown in fig. 3, all preiach operators (α, β) form a right triangle in a two-dimensional plane, history information of hysteresis characteristics can be represented by a memory curve in the triangle, and the memory curve is a staircase curve, and the formation rule is: when the input increases with time, the stepped curve reflects as a segment of the triangle that rises horizontally; when the input is reduced along with time, the step-shaped curve is reflected as a line segment which is vertically moved to the left in the triangle, the hysteresis state value of the left lower area of the step-shaped curve in the triangle is 1, the hysteresis state value of the right upper area of the step-shaped curve in the triangle is-1, meanwhile, each point in the triangle corresponds to a weight value, and the output value is obtained by integrating the product of the hysteresis state values of all the points in the triangle and the weight values.

In practical application, the continuous preisac triangle needs to be discretized, that is, the triangle is divided in the horizontal direction and the vertical direction (generally, equal-length division) to finally form a plurality of rectangular grids, a hysteresis state value corresponding to each grid is determined by hysteresis state values of all points included in the grid, the values are located in an interval [ -1, +1], each grid also corresponds to a weight value, the model needs to be trained before application, an OCV-SOC hysteresis relation curve obtained through experiments is used for training to obtain the number of grid divisions and the weight value corresponding to each grid, and a flow chart of a training process is shown in fig. 5.

The initial parameters of the model mainly comprise the side length division number n of the Preisach triangle in the discrete Preisach model and the hysteresis weight initial value mu corresponding to each grid in the Preisach triangle under the division. The Preisach triangle is a vertex coordinate (u) on the alpha-beta plane_min,u_max) (wherein u_minAnd u_maxOpen circuit voltage values corresponding to 0 and 100%, respectively, states of charge) are positioned on α ═ β. The selection of the side length division number n of the Preisach triangle is determined by the final precision requirement of the model, and the initial value of n is generally more than or equal to 20. The division results of the vertical and horizontal sides may be respectively expressed as α_i<α_i+1I is 1,2, …, n (where α is)₁＝u_min，α_n+1＝u_max) And beta_i<β_i+1I ═ 1,2, …, n (where β is₁＝u_min，β_n+1＝u_max). Finally, the total number of N ═ N (N +1)/2 meshes is generated in the whole Preisach triangle. Each grid may be represented as S_i(i-1)/2+j＝{(β,α)|β_j≤β＜β_j+1,α_i≤α＜α_i+1Where j ≦ i, i ≦ 1,2, …, n, j ≦ 1,2, …, n. Each grid in the discrete Preisach model corresponds to a weight value mu (called hysteresis weight, which needs to be obtained by off-line computation). And (3) taking the OCV value in the hysteresis characteristic experiment curve of the lithium ion battery as a model input u, updating each grid under each input, and generating a corresponding hysteresis state value omega under the input. The determination of the value of ω needs to be based on the staircase curve in the preiach triangle. The staircase curve is determined by an input value u (a horizontal ascending straight line is generated in the Preisach triangle when u is increased, a vertical left-moving straight line is generated in the Preisach triangle when u is decreased, and a staircase curve is generated in the Preisach triangle along with the increasing and decreasing of the input u), the hysteresis state omega corresponding to the grid S below the staircase curve is equal to 1, the hysteresis state omega corresponding to the grid S above the staircase curve is equal to-1, and when the staircase curve passes through the interior of the grid S, the corresponding hysteresis state omega is equal to the area of the part below the staircase curve in the grid minus the area of the part below the staircase curve; defining all the hysteresis states ω as a hysteresis state vector ω (ω ═ ω₁,ω₂,…,ω_N]^T) All the hysteresis weights μ are a hysteresis weight vector μ (μ ═ μ₁,μ₂,…,μ_N]^T) The corresponding output SOC at each input is equal to the product of the hysteretic state vector ω and the hysteretic weight vector μ, i.e.Finally, a linear equation system with unknown number as the hysteresis weight vector mu is obtained from input and output values obtained by all experiments, and mu is determined through off-line numerical operation.

As shown in fig. 6, in order to improve the modeling accuracy of the hysteresis curve of the lithium ion battery, the present invention provides a modeling method that takes the weight value corresponding to each grid as a time variable and can be adaptively changed, when the method is specifically applied, the weight value obtained by model training is taken as an initial weight value, each time first updates the hysteresis state values corresponding to all grids in a discrete preiach triangle according to an OCV input value, obtains an SOC value (defined as a prior SOC at the current time) by weighted summation of the hysteresis state values and the weight values obtained at the previous time, and determines a current estimation value at the current time according to the known battery capacity and the SOC value obtained at the previous time, compares the estimation value with the current value detected by the BMS to obtain an error value, compares the error value with the current hysteresis state value and a constant (the constant takes a value between 0 and 1 according to the minimum mean square error theory, a proper value can be selected by multiple values) as the increment of the current time weighted value compared with the previous time weighted value, the obtained current time weighted value and the hysteresis state value at the current time are weighted and summed, the posterior SOC at the current time is obtained and is the corresponding SOC output under the current OCV input, and the process is repeated at the next time.

Claims

1. A state of charge estimation method based on open-circuit voltage hysteresis characteristics is used for estimating the state of charge of a lithium ion battery on line, and is characterized by comprising the following steps:

2. The method according to claim 1, wherein the initial parameters of the adaptive model of the open circuit voltage hysteresis characteristic based on the Preisach operator in the step 2) include mesh division number N of the Preisach triangle and initial value μ of the weight vector.

3. The method according to claim 1, wherein the adaptive model of the open circuit voltage hysteresis characteristic based on the preiach operator in step 2) is:

4. The method for estimating the state of charge based on the open-circuit voltage hysteresis characteristic according to claim 3, wherein the step 3) specifically comprises the following steps:

Δt＝t_k-t_k-1；

η_current(t_k)＝I_m(t_k)-I_cal(t_k)；

μ(t_k)＝μ(t_k-1)+λη(t_k)ω(t_k)

wherein lambda is a step factor, and lambda belongs to [0,1 ];

5. The method according to claim 4, wherein in step 34), the weight vector at the current time is incrementedThe calculation formula of (A) is as follows:

wherein, λ is step factor, and λ is ∈ [0,1]]，η(t_k) Is t_kError values, ω (t), of the calculated and measured currents at times_k) Is t_kThe hysteretic state vector in the triangle at time Preisach.