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

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

A kind of method of estimation of the state-of-charge based on open-circuit voltage hysteretic characteristic

Technical field

The present invention relates to a kind of method of estimation of state-of-charge, especially relate to a kind of method of estimation of the state-of-charge based on open-circuit voltage hysteretic characteristic.

Background technology

Electrokinetic cell system is more and more applied in the field such as electric automobile and power energy storage as the parts of key.In application process, need battery management system (BatteryManagementSystem, BMS) to monitor battery status, prevent super-charge super-discharge from extending battery.At this wherein, the accurate estimation of SOC (state-of-charge) is particularly crucial.Most SOC method of estimation utilizes the corresponding relation of SOC and open-circuit voltage OCV (OpenCircuitVoltage, open-circuit voltage) to obtain, as open-circuit voltage method, based on the SOC method of estimation etc. of model.It is the key foundation of these SOC methods of estimation to the description of OCV and SOC corresponding relation.Open-circuit voltage and OCV also incomplete one_to_one corresponding in lithium ion battery, but there is hysteretic relationship (under same SOC, the OCV of charging process is greater than the OCV of discharge process).

Introduce more simplification in the modeling method of traditional lithium ion battery open-circuit voltage hysteretic characteristic, make the modeling accuracy of hysteretic behavior low, thus affect SOC estimation.The hysteretic behavior modeling method based on Preisach operator adopted herein, apply in the open-circuit voltage hysteresis modeling of Ni-MH battery (NiMH), but because lithium ion battery compares NiMH battery open circuit voltage hysteretic relationship not significantly and also without symmetry, make discrete Preisach model to be better applied in lithium ion battery.

Summary of the invention

Object of the present invention be exactly provide to overcome defect that above-mentioned prior art exists a kind of modeling accurately, improve precision, estimate accurately based on the method for estimation of the state-of-charge of open-circuit voltage hysteretic characteristic.

Object of the present invention can be achieved through the following technical solutions:

Based on a method of estimation for the state-of-charge of open-circuit voltage hysteretic characteristic, for the state-of-charge of On-line Estimation lithium ion battery, comprise the following steps:

1) off-line obtains the hysteretic characteristic curve of lithium ion battery open-circuit voltage and state-of-charge;

2) determine the initial parameter of the open-circuit voltage hysteretic characteristic adaptive model based on Preisach operator according to the training of hysteretic characteristic curve, and set up the open-circuit voltage hysteretic characteristic adaptive model based on Preisach operator;

3) according to the state-of-charge based on the open-circuit voltage hysteretic characteristic adaptive model On-line Estimation lithium ion battery of Preisach operator, the state-of-charge of current time lithium ion battery is obtained.

Described step 2) in the initial parameter of the open-circuit voltage hysteretic characteristic adaptive model based on Preisach operator comprise the initial value μ of the leg-of-mutton stress and strain model number N of Preisach and weight vectors.

Described step 2) in based on the open-circuit voltage hysteretic characteristic adaptive model of Preisach operator be:

${\mathrm{SOC}}_{t_k}\left(t_k\right)={\mathrm{\ω}}^T\left(t_k\right)\mathrm{\μ}\left(t_k\right)$

Wherein, for t _kthe SOC that moment open-circuit voltage is corresponding, ω (t _k) be t _kthe hysteresis state vector that moment open-circuit voltage values is corresponding in Preisach triangle, μ (t _k) be t _kthe hysteresis weight vectors of all grids in moment Preisach triangle.

Described step 3) specifically comprise the following steps:

31) the upper moment t of online acquisition _k-1preisach triangular mesh in weight vectors μ (t _k-1) and current time t _khysteresis state value ω (t _k), calculate the priori SOC of current time for:

${\mathrm{SOC}}_{t_{k-1}}\left(t_k\right)={\mathrm{\ω}}^T\left(t_k\right)\mathrm{\μ}\left(t_{k-1}\right);$

32) according to battery capacity Q ₀, a upper moment t _k-1sOC with current time t _kpriori SOC calculate the current estimation value I of current time _cal(t _k) be:

$I_{cal}\left(t_k\right)=Q_0\[{\mathrm{SOC}}_{t_{k-1}}\left(t_k\right)-{\mathrm{SOC}}_{t_{k-1}}\left(t_{k-1}\right)\]/\mathrm{\Δ}t$

Δt＝t _k-t _k-1；

33) according to the current actual value I of the current time of actual measurement _m(t _k) and current estimation value I _cal(t _k) obtain the current error value η _ current (t of current time _k) be:

η_current(t _k)＝I _m(t _k)-I _cal(t _k)；

34) according to the current error value η _ current (t of current time _k) and the priori SOC of current time and adopt LMSE method to obtain current time weight vectors increment and calculate the weighted value μ (t of current time _k) be:

μ(t _k)＝μ(t _k-1)+λη(t _k)ω(t _k)

$\mathrm{\η}\left(t_k\right)=\frac{\mathrm{\Δ}t}{Q_0\mathrm{\η}\_current\left(t_k\right)}$

Wherein λ is step factor, and λ ∈ [0,1];

35) according to the weight vectors μ (t of current time _k-1) and hysteresis state value ω (t _k) the posteriority SOC of current time is obtained by open-circuit voltage hysteretic characteristic adaptive model the i.e. state-of-charge of the lithium ion battery of current time, and return step 31) state-of-charge that carries out the lithium ion battery of subsequent time estimates.

Described step 34) in, current time weight vectors increment calculating formula be:

${\mathrm{\Δ\μ}}_{t_k}=\mathrm{\λ}\mathrm{\η}\left(t_k\right)\mathrm{\ω}\left(t_k\right)$

Wherein, λ is step factor, and λ ∈ [0,1], η (t _k) be t _kthe error amount of moment calculating current and measurement electric current, ω (t _k) be t _khysteresis state vector in moment Preisach triangle.

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

One, modeling is accurate: the present invention measures the error of electric current and calculating current by introducing, each moment to Preisach triangle in weighted value corresponding to grid carry out Automatic adjusument, by priori SOC and hysteresis state value and current flow actual value by iterative computation, lithium ion battery open-circuit voltage hysteretic characteristic is made accurately to carry out modeling.

Two, precision is improved: the present invention is the precision being improved modeling by the improvement of algorithm, and the measurement electric current played the role of a nucleus in algorithm directly obtains by current sensor original in battery management system, while precision improves, do not increase hardware cost.

Three, estimate accurately: the present invention is the estimation by carrying out state-of-charge to lithium ion battery hysteretic characteristic, the state-of-charge of the lithium ion battery (as ferric phosphate lithium cell) seriously can not ignored for hysteretic characteristic is estimated more accurately and reliably.

Accompanying drawing explanation

Fig. 1 is lithium ion battery OCV-SOC hysteretic characteristic curve synoptic diagram.

Fig. 2 is basic Preisach operator schematic diagram.

Fig. 3 is Preisach triangle and stepped appearance memory curve schematic diagram.

Fig. 4 is discrete Preisach triangle and grid schematic diagram.

Fig. 5 is lithium ion battery OCV-SOC hysteretic behavior training process flow diagram.

Fig. 6 is lithium ion battery OCV-SOC adaptive discrete Preisach models applying process flow diagram.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

Embodiment:

Fundamental purpose of the present invention sets up a kind of accurate modeling method of lithium ion battery open-circuit voltage hysteretic characteristic, thus finally can be used in the accurate estimation of charge states of lithium ion battery SOC, another object of the present invention is the modeling method by proposing is obtained under hysteresis existence OCV value by known accurate SOC, thus can be further used for analyzing the polarizing voltage of battery and impedance etc., for battery management system provides more information.

In order to realize the present invention's target as above and other advantages, as described particularly and widely here, a kind of modeling method of the lithium ion battery open-circuit voltage hysteretic characteristic based on Current adjustment and discrete Preisach operator is provided, the shortcoming that the larger precision of error is difficult to ensure is there is in traditional open-circuit voltage hysteretic characteristic modeling, and be not suitable for the present invention, the present invention at discrete Preisach operator on the basis of hysteretic characteristic modeling, weighted value corresponding to the grid of Preisach tessellation is considered as variations per hour, by asking for the current value in each moment undefined priori SOC and battery, obtain the deviate of electric current, the increment of weighted value change is inscribed when this value obtains each in conjunction with least mean-square error theory (LMS), sued for peace with the weighted value in a upper moment and obtain the weighted value of current time, further combined with the hysteresis state of current time, obtain the SOC output that posteriority SOC is current time, wherein in model, the number of stress and strain model and the initial value of weight vectors are determined by lithium ion battery OCV-SOC hysteresis loop off-line.

According to a preferred embodiment of the invention, complete implementation step is as follows: 1) test off-line and obtain lithium ion battery OCV-SOC hysteretic characteristic curve; 2) off-line determines the number of Preisach triangular grids and the initial value of corresponding weight vectors; 3) become when the weighted value that during application on site, Preisach triangular mesh is corresponding is considered as, by a upper moment weighted value, currency adds that current time variable quantity obtains; 4) according to the hysteresis state value of each grid of current time OCV value change Preisach triangle, and the priori SOC value of current time was obtained according to the weighted value in a upper moment; 5) obtained the current estimation value of current time by priori SOC value, the SOC value in a upper moment, battery capacity, compare with actual current and obtain error amount; 6) the current error value of current time obtains the increment of current time weighted value and the weighted value of current time in conjunction with least mean-square error (LMS) theory; 7) weighted value of current time is the SOC value of current time in conjunction with the posteriority SOC value that the hysteresis state value of current time obtains current time, and subsequent time repeats said process.

Figure 1 shows that lithium ion battery OCV-SOC hysteretic characteristic curve (the OCV sufficient standing of experiment test, time is greater than 3 hours), this empirical curve is for determining the initial parameter of the hysteretic behavior that the present invention proposes, in order to fully cover the hysteretic characteristic of lithium ion battery, first from the electric discharge of battery fully charged state during measuring, after 5%SOC changes and leaves standstill, record OCV value is until SOC=0 completely puts state; Then enter charged state every 5%SOC change with after recording OCV value to SOC=95% after leaving standstill, transfer electric discharge to, until change charging into again after 5%SOC; The circulation of this process is carried out and is ensured that each charged state reduces 5% than the SOC value of a upper charged state, and each discharge condition increases by 5% than the SOC value of a upper discharge condition, until SOC=50% experiment terminates; In whole process, OCV leaves standstill and record after SOC often changes 5%.

As in Figure 2-4, Preisach operator (α, β) describe a kind of relation of input and output, namely exporting when input is greater than threshold alpha is 1, export when input is less than threshold value beta (β < α) as-1, when inputting between threshold alpha and β, output valve is constant, because in hysteretic relationship, input value amplitude is limited, as shown in Figure 3, all Preisach operator (α, β) constitute a right-angle triangle in two dimensional surface, the historical information of hysteretic characteristic represents by a memory curve in this triangle, memory curve is a step curve, its formation rule is: when input increases in time, step curve is reflected as the line segment that in triangle, level rises, when input reduces in time, step curve is reflected as the line segment vertically moved to left in triangle, triangle interior step curve lower left region hysteresis state value is 1, right regions hysteresis state value is-1, the simultaneously corresponding weighted value of triangle interior every bit, output valve by triangle hysteresis state value a little and weighted value product integral obtain.

When practical application, need to carry out discretize to this continuous P reisach triangle, namely with vertical direction, (generally isometric divide) is divided to this triangle in the horizontal direction, the multiple rectangular node of final formation, hysteresis state value corresponding to each grid by this grid comprise hysteresis state value a little determine, this value is positioned at interval [-1, + 1], the corresponding equally weighted value of each grid, need this model training before application, training adopts the OCV-SOC hysteretic relationship curve of testing and obtaining, obtain the number of stress and strain model and weighted value corresponding to each grid, the process flow diagram of training process as shown in Figure 5.

The initial parameter of model mainly comprises Preisach triangle edges dash in discrete Preisach model and divides the hysteresis weight initial value μ that in Preisach triangle, each grid is corresponding under number n and this division.Preisach triangle is that in alpha-beta plane, an apex coordinate is (u _min, u _max) (wherein u _minand u _maxto be state-of-charge be respectively 0 and 100% time corresponding open-circuit voltage values) hypotenuse is positioned at isosceles right triangle on α=β.Preisach triangle edges dash divides the accuracy requirement chosen by model is final of number n to determine, the initial value generally getting n is more than or equal to 20.The division result of vertical edge and horizontal sides can be expressed as α _i< α _i+1, i=1,2 ..., n (wherein α ₁=u _min, α _n+1=u _max) and β _i< β _i+1, i=1,2 ..., n (wherein β ₁=u _min, β _n+1=u _max).Raw N=n (n+1)/2 grid of common property in final whole Preisach triangle.Each grid can be expressed as S _{i (i-1)/2+j}=(β, α) | β _j≤ β < β _j+1, α _i≤ α < α _i+1wherein j≤i, i=1,2 ..., n, j=1,2 ..., n.The all corresponding weighted value μ of each grid (be called hysteresis weight, need calculated off-line to obtain) in discrete Preisach model.Get in lithium ion battery hysteretic characteristic empirical curve that OCV value is as mode input u, under each input, each grid all upgrades, hysteresis state value ω corresponding under producing this input.The determination of ω value needs according to the step curve in Preisach triangle.This step curve is determined (in Preisach triangle, to produce a level rising straight line when u increases by input value u, in Preisach triangle, produce one vertically to move to left straight line when u reduces, in Preisach triangle, a step curve is generated) along with input u constantly increases and decreases, the hysteresis state ω that grid S below step curve is corresponding equals 1, the hysteresis state ω that grid S more than step curve is corresponding equals-1, when step curve passes the inside of grid S, the area of the part that corresponding hysteresis state ω equals in grid below step curve deducts the area of the part below step curve, defining all hysteresis state ω is a hysteresis state vector ω (ω=[ω ₁, ω ₂..., ω _n] ^t), all hysteresis weight mu are a hysteresis weight vectors μ (μ=[μ ₁, μ ₂..., μ _n] ^t), the lower corresponding output SOC of each input equals the product of hysteresis state vector ω and hysteresis weight vectors μ, namely the input and output value finally obtained by all experiments obtains the system of linear equations that a unknown number is hysteresis weight vectors μ, determines μ by off-line numerical operation.

As shown in Figure 6, in order to improve the modeling accuracy of lithium ion battery hysteresis loop, the present invention proposes and weighted value corresponding for each grid is considered as variations per hour and the modeling method of energy adaptively changing, the weighted value that model training obtains when embody rule the method is considered as weight initial value, first each moment upgrades according to OCV input value the hysteresis state value that in discrete Preisach triangle, all grids are corresponding, the weighted value weighted sum obtained by hysteresis state value and a upper moment obtains SOC value (being defined as the priori SOC of current time), and current time current estimation value can be determined according to the SOC value that known battery capacity and a upper moment obtain, the current value ratio that this estimated value and BMS detect comparatively obtains error amount, theoretical according to least mean-square error, this error amount and current hysteresis state value and constant are (this constant value between zero and one, choose desired value by repeatedly value) product compare the increment of a upper moment weighted value as current time weighted value, by the hysteresis state value weighted sum of the current time weighted value that obtains and current time, the posteriority SOC obtaining current time is current OCV and inputs lower corresponding SOC output, subsequent time repeats this process.

Claims (5)

1., based on a method of estimation for the state-of-charge of open-circuit voltage hysteretic characteristic, for the state-of-charge of On-line Estimation lithium ion battery, it is characterized in that, comprise the following steps:

1) off-line obtains the hysteretic characteristic curve of lithium ion battery open-circuit voltage and state-of-charge;

2) determine the initial parameter of the open-circuit voltage hysteretic characteristic adaptive model based on Preisach operator according to the training of hysteretic characteristic curve, and set up the open-circuit voltage hysteretic characteristic adaptive model based on Preisach operator;

3) according to the state-of-charge based on the open-circuit voltage hysteretic characteristic adaptive model On-line Estimation lithium ion battery of Preisach operator, the state-of-charge of current time lithium ion battery is obtained.

2. the method for estimation of a kind of state-of-charge based on open-circuit voltage hysteretic characteristic according to claim 1, it is characterized in that, described step 2) in the initial parameter of the open-circuit voltage hysteretic characteristic adaptive model based on Preisach operator comprise the initial value μ of the leg-of-mutton stress and strain model number N of Preisach and weight vectors.

3. the method for estimation of a kind of state-of-charge based on open-circuit voltage hysteretic characteristic according to claim 1, is characterized in that, described step 2) in based on the open-circuit voltage hysteretic characteristic adaptive model of Preisach operator be:

${\mathrm{SOC}}_{t_k}\left(t_k\right)={\mathrm{\&omega;}}^T\left(t_k\right)\mathrm{\&mu;}\left(t_k\right)$

Wherein, for t _kthe SOC that moment open-circuit voltage is corresponding, ω (t _k) be t _kthe hysteresis state vector that moment open-circuit voltage values is corresponding in Preisach triangle, μ (t _k) be t _kthe hysteresis weight vectors of all grids in moment Preisach triangle.

4. the method for estimation of a kind of state-of-charge based on open-circuit voltage hysteretic characteristic according to claim 3, is characterized in that, described step 3) specifically comprise the following steps:

31) the upper moment t of online acquisition _k-1preisach triangular mesh in weight vectors μ (t _k-1) and current time t _khysteresis state value ω (t _k), calculate the priori SOC of current time for:

${\mathrm{SOC}}_{t_{k-1}}\left(t_k\right)={\mathrm{\&omega;}}^T\left(t_k\right)\mathrm{\&mu;}\left(t_{k-1}\right);$

32) according to battery capacity Q ₀, a upper moment t _k-1sOC with current time t _kpriori SOC calculate the current estimation value I of current time _cal(t _k) be:

$I_{cal}\left(t_k\right)=Q_0\&lsqb;{\mathrm{SOC}}_{t_{k-1}}\left(t_k\right)-{\mathrm{SOC}}_{t_{k-1}}\left(t_{k-1}\right)\&rsqb;/\mathrm{\&Delta;}t$

Δt＝t _k-t _k-1；

33) according to the current actual value I of the current time of actual measurement _m(t _k) and current estimation value I _cal(t _k) obtain the current error value η _ current (t of current time _k) be:

η_current(t _k)＝I _m(t _k)-I _cal(t _k)；

34) according to the current error value η _ current (t of current time _k) and the priori SOC of current time and adopt LMSE method to obtain current time weight vectors increment and calculate the weighted value μ (t of current time _k) be:

μ(t _k)＝μ(t _k-1)+λη(t _k)ω(t _k)

$\mathrm{\&eta;}\left(t_k\right)=\frac{\mathrm{\&Delta;}t}{Q_0\mathrm{\&eta;}\_current\left(t_k\right)}$

Wherein λ is step factor, and λ ∈ [0,1];

35) according to the weight vectors μ (t of current time _k-1) and hysteresis state value ω (t _k) the posteriority SOC of current time is obtained by open-circuit voltage hysteretic characteristic adaptive model the i.e. state-of-charge of the lithium ion battery of current time, and return step 31) state-of-charge that carries out the lithium ion battery of subsequent time estimates.

5. the method for estimation of a kind of state-of-charge based on open-circuit voltage hysteretic characteristic according to claim 4, is characterized in that, described step 34) in, current time weight vectors increment calculating formula be:

${\mathrm{\&Delta;\&mu;}}_{t_k}=\mathrm{\&lambda;}\mathrm{\&eta;}\left(t_k\right)\mathrm{\&omega;}\left(t_k\right)$

Wherein, λ is step factor, and λ ∈ [0,1], η (t _k) be t _kthe error amount of moment calculating current and measurement electric current, ω (t _k) be t _khysteresis state vector in moment Preisach triangle.