CN108490357A - Lithium battery residual capacity prediction technique based on mechanism-data-driven model - Google Patents

Abstract

The invention discloses a method for predicting the remaining capacity of a lithium battery based on a mechanism-data-driven model. First, a degradation mechanism model of the remaining capacity of the lithium battery and its charge-discharge cycle is constructed; the original remaining capacity data of the lithium battery is subjected to convex optimization and noise reduction processing, based on The data with high reliability obtained by preprocessing is used to identify the unknown parameters in the mechanism model by using the least square method, so as to obtain an accurate prediction model expression, and realize the prediction of the remaining capacity of the lithium battery based on the mechanism model. Then establish a modeling error data-driven model based on LSSVM, and feed back the modeling error estimated by LSSVM to the prediction result of the mechanism model, so as to realize the high-precision prediction of the remaining capacity of the lithium battery. The method of the invention is applicable to the prediction of the remaining capacity of the lithium battery under different working conditions. In the prediction process, the external environment information of the battery and the influence of the working condition of the circuit on its life degradation are comprehensively considered, which is more consistent with the actual situation.

Description

Lithium battery remaining capacity prediction method based on mechanism-data-driven model

technical field

The invention relates to a battery capacity prediction method, in particular to a mechanism-data-driven model-based lithium battery remaining capacity prediction method.

Background technique

Lithium batteries are widely used in consumer electronics, electric vehicles, aerospace and other fields because of their advantages such as long cycle life, low self-discharge rate, and good safety performance. However, the failure of lithium batteries during the degradation process will directly affect the normal operation of machinery and equipment, and even cause serious safety accidents and property losses. The remaining capacity of the battery refers to the electrical energy stored in the battery when it is fully charged under different cycle periods. As the number of charge and discharge cycles increases, the remaining capacity of the battery presents a declining trend, and the degree of life degradation can be predicted according to the change trend of capacity, which provides an important guarantee for the safe and reliable operation of the energy storage system. The existing lithium battery capacity prediction methods are usually only applicable to the prediction of the lithium battery capacity working under a certain working condition, and there are certain limitations to the prediction of the lithium battery capacity working under different working conditions. On the other hand, during the actual degradation process of lithium batteries in the healthy state, they are often affected by external conditions, mainly including environmental factors and circuit working conditions. High-precision prediction of battery capacity.

For this reason, the present invention provides a method for predicting the remaining capacity of a lithium battery based on a mechanism-data-driven model. This method is applicable to the prediction of the capacity of a lithium battery working under different working conditions. The impact of external environmental information and circuit working conditions on its life degradation is more consistent with the actual situation, providing a new idea for the prediction of the health status of lithium batteries.

Contents of the invention

The purpose of the present invention is to provide a method for predicting the remaining capacity of lithium batteries based on a mechanism-data-driven model, which is used to predict the remaining capacity of lithium batteries, and provides guarantee for realizing efficient and accurate prediction and health management.

In order to achieve the above object, the solution of the present invention is:

A method for predicting the remaining capacity of a lithium battery based on a mechanism-data-driven model, including the following steps (1) to (5):

(1) The working condition information of the lithium battery mainly includes the external environment information and the circuit working information, wherein the environmental information includes the ambient temperature T and the relative humidity W, and the circuit working information includes the charging current I _in , the charging voltage V _in , Discharge current I _out , discharge cut-off voltage V _out ; set the ambient temperature T as T′, the relative humidity W as W′, the charging current I _in as I _i ′ _n , and the charging voltage V _in as V ′ _in , the discharge current I _out is set to I′ _out , and the discharge cut-off voltage V _out is set to V’ _out under the standard working conditions, take m pieces of the same lithium battery and carry out cycle charge and discharge experiments at the same time, monitor and record each Lithium battery charge-discharge cycle C and the corresponding remaining capacity q _ac of the lithium battery under the C-th charge-discharge cycle, so as to calculate the average value of the remaining capacity of m lithium batteries corresponding to the C-th charge-discharge cycle Among them, a is the number of the lithium battery under test, and a=1, 2, 3...m, C=1, 2, 3...t;

(2) Establish a lithium battery residual capacity degradation mechanism model: q _c = k ₁ Ck ₂ e ^αC + k ₃ Q ₁ , where (α, k ₁ , k ₂ , k ₃ ) are unknown parameters of the model, and Q ₁ is the lithium battery Initial capacity, q _c is the corresponding remaining capacity of the lithium battery under the Cth charge-discharge cycle; the mean value of the remaining capacity under different charge-discharge cycles obtained in step (1) The average initial capacity of m lithium batteries can be obtained Then the model expression of the remaining capacity degradation mechanism of the lithium battery under standard working conditions is obtained:

(3) For the remaining capacity data of the lithium battery obtained in step (1) Carry out convex optimization noise reduction processing, the processed data As the basis for the identification of model parameters; use the least squares method to identify the unknown parameters (Δ, k ₁ , k ₂ , k ₃ ) of the model in step (1), so as to obtain the specific expression of the degradation model: Where (α ^* , k ₁ ^* , k ₂ ^* , k ₃ ^* ) is the parameter identification result; thus, the specific expression of the general model for the remaining capacity degradation mechanism of this type of lithium battery under standard working conditions is: q _C ＝ k ₁ ^* Ck ₂ ^* e ^α*C +k ₃ ^* Q ₁ ;

(4) Another n=500 lithium batteries of the same type as those in step (1) were placed in n different non-standard working conditions and carried out the cycle charge and discharge test simultaneously, that is, each lithium battery was tested in different charge and discharge cycles. The ambient temperature, relative humidity, charging current, charging voltage, discharging current, and discharge cut-off voltage in the cycle are T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc , monitor and record the initial capacity value q _j.1 of each lithium battery, the charge-discharge cycle C, and the remaining capacity value q _jc under the C-th charge-discharge cycle; bring the initial capacity value q _j.1 into the step (3) In the general model of the degradation mechanism of the remaining capacity of the lithium battery, the predicted value of the remaining capacity of the lithium battery based on the degradation mechanism model is obtained to get the model prediction error Based on the Δq _jc obtained from the test and the multivariate information of the corresponding operating conditions (C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ), the modeling based on LSSVM is established The error prediction model, denoted as Δq _jC =f(C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ), can be based on the current operation of the lithium battery Δq _jC is obtained by multivariate information of working conditions; among them, j is the serial number of the lithium battery tested in this step, j=1,2,3...n;

(5) Monitor the initial capacity value q _x.1 of the lithium battery to be tested and the current operating conditions (C, T _xc , W _xc , I _in.xc , V _in.xc , I _out.xc , V _out.xc ), to obtain the predicted value of the remaining capacity of the lithium battery under test converted to the standard working condition And the model prediction error Δq _xC , and then obtain the predicted value of the remaining capacity of the lithium battery under test under the C charge-discharge cycle Where x is the serial number of the lithium battery to be tested.

In the method for predicting the remaining capacity of lithium batteries based on the mechanism-data-driven model of the present invention, the lithium battery is subjected to a cycle charge and discharge experiment under standard working conditions in the step (1), which means that the ambient temperature is T′ and the relative humidity is W 'Under the condition of _I'in , first charge with a constant current of I'in until the voltage reaches _V'in , and then discharge with a constant current of _I'out until the cut-off voltage is _V'out , and this cycle is carried out; the remaining capacity refers to different charge and discharge cycles Under the cycle, the electric energy stored by the lithium battery when it is fully charged. The mean value of the remaining capacity of m lithium batteries corresponding to the C-th charge-discharge cycle

In the method for predicting the remaining capacity of a lithium battery based on a mechanism-data-driven model of the present invention, in the step (2), the remaining capacity degradation mechanism model of the lithium battery is established: q _C =k ₁ Ck ₂ e ^αC +k ₃ Q ₁ , specifically Proceed as follows:

The rate of change of battery capacity is a function of its charge-discharge cycle C and the remaining capacity q _C of the battery, so its degradation rate can be expressed as:

For the nonlinear function f(q _C , C) containing two independent variables q _C , C, use the Taylor series expansion of the multivariate function, according to the linearization approximation of the nonlinear differential equation, omit its high power item, you can get:

where a ₁ is the degradation factor and a ₂ is the fatigue damage accumulation factor. As the number of cycles increases, its capacity shows a decreasing trend, so there is

a ₁ q _C +a ₂ C<0 (3)

And q _C >0, C>0

this season

u＝a ₁ q _C +a ₂ C (4)

It can be known from formula (3):

u<0 (5)

Differentiate both sides of (4) to get:

du=a ₁ dq _C +a ₂ dC (6)

Bring (6) into (2) to get:

At this time, the solution of the differential equation has the following two situations:

1) When a ₁ u+a ₂ >0, the solution of equation (7):

2) When a ₁ u+a ₂ <0, the solution of equation (7):

Formula (10) and formula (13) can be written in the following form:

q _C ＝k ₁ Ck ₂ e ^αC +k ₃ Q ₁ (14)

Equation (14) is the model expression of the degradation mechanism of the battery remaining capacity q _C with the charge-discharge cycle C, where (α, k ₁ , k ₂ , k ₃ ) are unknown parameters of the model, and Q ₁ is the initial capacity of the lithium battery, namely The corresponding remaining capacity value of the lithium battery under the first charge and discharge cycle.

According to the lithium battery remaining capacity prediction method based on the mechanism-data-driven model of the present invention, in the step (4), a modeling error prediction model based on LSSVM is established, and the specific steps are as follows:

(4.1) Take another n=500 lithium batteries of the same type as those in step (1) and place them under n different non-standard working conditions to carry out cycle charge and discharge tests at the same time. Inside, put n pieces of lithium batteries under n random non-standard operating conditions (T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ) and cycle them simultaneously Charge and discharge test; the specific steps are: under the conditions of the external environment temperature of T _{jc and} relative humidity of W _jc , first charge the battery with I _in.jc current until the battery is fully charged, that is, the voltage reaches V _in.jc , and then charge with I The current of _out.jc is discharged to the cut-off voltage V _out.jc , and so on;

(4.2) Select the Gaussian kernel as the kernel function of LSSVM, and select the optimal kernel parameters as γ=20, σ ² =110; the multivariate operating condition information (C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ) and the prediction error Δq _jC of the degradation mechanism model are used as training samples for LSSVM regression fitting, and regression modeling is performed to obtain a modeling error prediction based on LSSVM model, denoted as Δq _jC = f(C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ); the LSSVM algorithm is an existing mature method, which is no longer repeat.

In the method for predicting the remaining capacity of a lithium battery based on a mechanism-data-driven model of the present invention, in the step (5), the current predicted value of the remaining capacity of the lithium battery to be tested is obtained Specific steps are as follows:

(5.1) Monitor the initial capacity value q _x.1 of the lithium battery to be tested and bring it into the general model of the remaining capacity degradation mechanism of this type of lithium battery in step (3) to obtain the value of the lithium battery to be tested converted to the standard working condition Predicted remaining capacity

(5.2) Monitor the multivariate information sequence (C, T _xc , W _xc , I _in.xc , V _in.xc , I _out.xc , V _out.xc ) of the current external working environment of the lithium battery to be tested and Bring it into the modeling error prediction model established in step (4), and obtain the prediction error of the degradation mechanism model Δq _xC = f(C, T _xc , W _xc , I _in.xc , V _in.xc , I _out.xc , V _out.xc );

(5.3) Obtain the predicted value of the remaining capacity of the current lithium battery to be tested

Description of drawings

Figure 1 is a flow chart of a method for predicting the remaining capacity of a lithium battery based on a mechanism-data-driven model.

Detailed ways

The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings.

The present invention provides a method for predicting the remaining capacity of a lithium battery based on a mechanism-data-driven model. The general idea is as follows: firstly, a degradation mechanism model of the remaining capacity of the lithium battery and its charge-discharge cycle is constructed; convex optimization is performed on the original remaining capacity data of the lithium battery Noise reduction processing, based on the highly reliable data obtained by preprocessing, uses the least square method to identify the unknown parameters in the mechanism model, thereby obtaining an accurate prediction model expression, and realizing the prediction of the remaining capacity of the lithium battery based on the mechanism model. Then establish a modeling error data-driven model based on LSSVM, and feed back the modeling error estimated by LSSVM to the prediction result of the mechanism model, so as to realize the high-precision prediction of the remaining capacity of the lithium battery. The method of the invention is applicable to the prediction of the remaining capacity of the lithium battery under different working conditions. In the prediction process, the external environment information of the battery and the influence of the working condition of the circuit on its life degradation are comprehensively considered, which is more consistent with the actual situation.

As shown in Figure 1, the method for predicting the remaining life of a lithium battery based on the capacity degradation mechanism model of the present invention, the specific implementation includes the following steps (1) to (5):

(1) The working condition information of the lithium battery mainly includes the external environment information and the circuit working information, wherein the environmental information includes the ambient temperature T and the relative humidity W, and the circuit working information includes the charging current I _in , the charging voltage V _in , Discharge current I _out , discharge cut-off voltage V _out ; set the ambient temperature T as T′, the relative humidity W as W′, the charging current I _in as I′ _in , and the charging voltage V _in as V′ _in , the discharge current I _out is set as I′ _out , and the discharge cut-off voltage V _out is set as V′ _out under the standard working conditions, take m pieces of the same lithium battery and carry out cycle charge and discharge experiments at the same time, monitor and record each lithium battery The battery charge-discharge cycle C and the corresponding remaining capacity q _ac of the lithium battery under the C-th charge-discharge cycle, so as to calculate the average value of the remaining capacity of m lithium batteries corresponding to the C-th charge-discharge cycle Among them, a is the number of the lithium battery under test, and a=1,2,3...m, C=1,2,3...t; the specific implementation is:

Carry out cyclic charging and discharging experiment to lithium battery under the standard working condition in described step (1), refer to be T ' at ambient temperature, under the condition that relative humidity is W ', first charge with the constant current of I ' _in to voltage reaches V' _in , and then discharged with a constant current of I' _out to the cut-off voltage of V' _out , and so on. The remaining capacity refers to the electric energy stored in the lithium battery when it is fully charged under different charge and discharge cycles. The mean value of the remaining capacity of m lithium batteries corresponding to the C-th charge-discharge cycle

(2) Establish a lithium battery remaining capacity degradation mechanism model: q _C = k ₁ Ck ₂ e ^αC + k ₃ Q ₁ , where (α, k ₁ , k ₂ , k ₃ ) are unknown parameters of the model, and Q ₁ is the lithium battery Initial capacity, q _C is the remaining capacity of the lithium battery corresponding to the Cth charge-discharge cycle; the mean value of the remaining capacity under different charge-discharge cycles obtained in step (1) The average initial capacity of m lithium batteries can be obtained Then the model expression of the remaining capacity degradation mechanism of the lithium battery under standard working conditions is obtained: The specific implementation is:

In the step (2), a lithium battery residual capacity degradation mechanism model is established: q _C =k ₁ Ck ₂ e ^αC +k ₃ Q ₁ , the specific steps are as follows:

The rate of change of battery capacity is a function of its charge-discharge cycle C and the actual capacity q _C of the battery, so its degradation rate can be expressed as:

For the nonlinear function f(q _C , C) containing two independent variables q _C , C, use the Taylor series expansion of the multivariate function, according to the linearization approximation of the nonlinear differential equation, omit its high power item, you can get:

where a ₁ is the degradation factor and a ₂ is the fatigue damage accumulation factor. As the number of cycles increases, its capacity shows a decreasing trend, so there is

a ₁ q _C +a ₂ C<0 (3)

And q _C >0, C>0

this season

u＝a ₁ q _C +a ₂ C (4)

It can be known from formula (3):

u<0 (5)

Differentiate both sides of (4) to get:

du=a ₁ dq _C +a ₂ dC (6)

Bring (6) into (2) to get:

At this time, the solution of the differential equation has the following two situations:

1) When a ₁ u+a ₂ >0, the solution of equation (7):

2) When a ₁ u+a ₂ <0, the solution of equation (7):

Formula (10) and formula (13) can be written in the following form:

q _C ＝k ₁ Ck ₂ e ^αC +k ₃ Q ₁ (14)

Equation (14) is the model expression of the degradation mechanism of the actual capacity q _C of the battery with the charge-discharge cycle C, where (α, k ₁ , k ₂ , k ₃ ) are unknown parameters of the model, and Q ₁ is the initial capacity of the lithium battery, namely The corresponding remaining capacity value of the lithium battery under the first charge and discharge cycle.

(3) For the remaining capacity data of the lithium battery obtained in step (1) Carry out convex optimization noise reduction processing, the processed data As the basis for the identification of model parameters; use the least squares method to identify the unknown parameters (α, k ₁ , k ₂ , k ₃ ) of the model in step (2), so as to obtain the specific expression of the degradation model: Where (α ^* , k ₁ ^* , k ₂ ^* , k ₃ ^* ) is the parameter identification result; thus, the specific expression of the general model for the remaining capacity degradation mechanism of this type of lithium battery under standard working conditions is obtained as:

(4) Another n=500 lithium batteries of the same type as those in step (1) were placed in n different non-standard working conditions and carried out the cycle charge and discharge test simultaneously, that is, each lithium battery was tested in different charge and discharge cycles. The ambient temperature, relative humidity, charging current, charging voltage, discharging current, and discharge cut-off voltage in the cycle are T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc , monitor and record the initial capacity value q _j.1 of each lithium battery, the charge-discharge cycle C, and the remaining capacity value q _jC under the C-th charge-discharge cycle; bring the initial capacity value q _j.1 into the step (3) In the general model of the degradation mechanism of the remaining capacity of the lithium battery, the predicted value of the remaining capacity of the lithium battery based on the degradation mechanism model is obtained to get the model prediction error Based on the Δq _jC obtained from the test and the multivariate information of the corresponding operating conditions (C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ), the modeling based on LSSVM is established The error prediction model, denoted as Δq _jC =f(C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ), can be based on the current operation of the lithium battery Δq _jC is obtained by multivariate information of working conditions; among them, j is the serial number of the lithium battery tested in this step, j=1,2,3...n; the specific realization is as follows:

(4.1) Take another n=500 lithium batteries of the same type as those in step (1) and place them under n different non-standard working conditions to carry out cycle charge and discharge tests at the same time. Inside, put n pieces of lithium batteries under n random non-standard operating conditions (T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ) and cycle them simultaneously Charge and discharge test; the specific steps are: under the conditions of the external environment temperature of T _{jc and} relative humidity of W _jc , first charge the battery with I _in.jc current until the battery is fully charged, that is, the voltage reaches V _in.jc , and then charge with I The current of _out.jc is discharged to the cut-off voltage V _out.jc , and so on;

(4.2) Select the Gaussian kernel as the kernel function of LSSVM, and select the optimal kernel parameters as γ=20, σ ² =110; the multivariate operating condition information (C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ) and the prediction error Δq _jC of the degradation mechanism model are used as training samples for LSSVM regression fitting, and regression modeling is performed to obtain a modeling error prediction based on LSSVM model, denoted as Δq _jC = f(C, T _jc , W _jc , I _in.jc , V _in.jc , I _out.jc , V _out.jc ); the LSSVM algorithm is an existing mature method, which is no longer repeat.

(5) Monitor the initial capacity value q _x.1 of the lithium battery to be tested and the current operating conditions (C, T _xc , W _xc , I _in.xc , V _in.xc , I _out.xc , V _out.xc ), to obtain the predicted value of the remaining capacity of the lithium battery under test converted to the standard working condition And the model prediction error Δq _xC , and then obtain the predicted value of the remaining capacity of the current lithium battery to be tested Where x is the serial number of the lithium battery to be tested. The specific implementation is: (5.1) monitor the initial capacity value q _x.1 of the lithium battery to be tested and bring it into step (3) the general model of the remaining capacity degradation mechanism of the lithium battery of this type, and obtain Measuring the predicted value of remaining capacity of lithium battery

(5.2) Monitor the multivariate information sequence (C, T _xc , W _xc , I _in.xc , V _in.xc , I _out.xc , V _out.xc ) of the current external working environment of the lithium battery to be tested and into the modeling error prediction model established in step (4), to obtain the prediction error of the degradation mechanism model Δq _xc = f(C, T _xc , W _xc , I _in.xc , V _in.xc , I _out.xc , V _out.xc );

(5.3) Obtain the predicted value of the remaining capacity of the current lithium battery to be tested

The above embodiments are only to illustrate the technical ideas of the present invention, and can not limit the protection scope of the present invention with this. All technical ideas proposed in accordance with the present invention, any changes made on the basis of technical solutions, all fall within the protection scope of the present invention. Inside.

Claims (5)

1. a kind of lithium battery residual capacity prediction technique based on mechanism-data-driven model, which is characterized in that including following step Suddenly：

(1) residing work information includes mainly external environment information and circuit job information when lithium battery is run, wherein extraneous Environmental information includes environment temperature T and relative humidity W, and circuit job information includes charging current I_in, charging voltage V_in, electric discharge Electric current I_out, discharge cut-off voltage V_out；It is set as T ' in environment temperature T, relative humidity W is set as W ', charging current I_inSetting For I '_in, charging voltage V_inIt is set as V '_in, discharge current I_outIt is set as I '_out, discharge cut-off voltage V_outIt is set as V '_out's Under standard condition, takes the identical lithium battery of m blocks to be carried out at the same time cycle charge-discharge experiment, monitor and record the charge and discharge of every piece of lithium battery The electric cycle period C and corresponding lithium battery residual capacity q under the C charge and discharge cycles periods_a.c, to calculate m blocks lithium electricity Pond corresponding residual capacity mean value under the C charge and discharge cycles periodsWherein, a is the lithium battery number tested, and a =1,2,3...m, C=1,2,3...t；

(2) lithium battery residual capacity degradation mechanism model is established：q_C=k₁C-k₂e^aC+k₃Q₁, wherein (α, k₁, k₂, k₃) be model not Know parameter, Q₁For lithium battery initial capacity, q_CFor corresponding lithium battery residual capacity under the C charge and discharge cycles periods；By step (1) the residual capacity mean value under the different charge and discharge cycles periods obtained inThe initial capacity that can obtain m block lithium batteries is equal ValueAnd then obtain this kind of lithium battery residual capacity degradation mechanism model expression under standard condition：

(3) the lithium battery residual capacity data that step (1) is obtainedCarry out convex optimization noise reduction process, data that treatedMake For the foundation of identification of Model Parameters；Using least square method to unknown-model parameter (α, k in step (2)₁, k₂, k₃) distinguished Know, is to obtain specific degradation model expression formula：Wherein (α^*, k₁ ^*, k₂ ^*, k₃ ^*) be Parameter identification result；Thus the model lithium battery residual capacity degradation mechanism universal model expression under standard condition is obtained For：

(4) another that n=500 blocks is taken to be respectively placed in n different nonstandard designs similarly hereinafter from the lithium battery of model of the same race in step (1) It is Shi Jinhang cycle charge discharge electric tests, i.e. every piece of lithium battery environment temperature residing under the different charge and discharge cycles periods, relatively wet Degree, charging current, charging voltage, discharge current, discharge cut-off voltage are respectively T_j.c, W_j.c, I_in.j.c, V_in.j.c, I_out.j.c, V_out.j.cWhen, monitor and record the initial capacity value q of every piece of lithium battery_j.1, charge and discharge cycles period C and followed in C charge and discharge Remaining capacity value q under the ring period_j.C；By initial capacity value q_j.1It is general to bring step (3) lithium battery residual capacity degradation mechanism into In model, the lithium battery residual capacity predicted value based on degradation mechanism model is obtainedTo Obtain model predictive errorThe Δ q obtained according to experiment_j.CAnd corresponding operating condition multiple information (C, T_j.c, W_j.c, I_in.j.c, V_in.j.c, I_out.j.c, V_out.j.c), the modeling error prediction model based on LSSVM is established, Δ q is denoted as_j.C= F (C, T_j.c, W_j.c, I_in.j.c, V_in.j.c, I_out.j.c, V_out.j.c), you can the polynary letter of operating condition being presently according to lithium battery It ceases to acquire Δ q_j.C；Wherein, j is that the lithium battery of this step experiment is numbered, j=1,2,3...n；

(5) the initial capacity value q of lithium battery to be measured is monitored_x.1And operating condition information (C, the T being presently in_x.c, W_x.c, I_in.x.c, V_in.x.c, I_out.x.c, V_out.x.c), conversion is obtained to the lithium battery residual capacity predicted value to be measured under standard conditionWith And model predictive error Δ q_x.C, and then acquire current lithium battery residual capacity predicted value to be measuredWherein x It is numbered for lithium battery to be measured.

2. the lithium battery residual capacity prediction technique based on mechanism-data-driven model, feature exist as described in claim 1 In, under step (1) the Plays operating mode to lithium battery carry out cycle charge-discharge experiment, refer to environment temperature be T ', relatively Under conditions of humidity is W ', first with I '_inConstant current charge reach V ' to battery up to full power state, that is, voltage_in, then with I '_out Constant current be discharged to blanking voltage V '_out, so cycle progress；Residual capacity refers to lithium under the different charge and discharge cycles periods Battery electric energy stored in the case of fully charged；M blocks lithium battery corresponding residual capacity under the C charge and discharge cycles periods Mean value

3. the lithium battery residual capacity prediction technique based on mechanism-data-driven model, feature exist as described in claim 1 In establishing lithium battery residual capacity degradation mechanism model in the step (2)：q_C=k₁C-k₂e^αC+k₃Q₁, it is as follows：

The change rate of battery capacity is its charge and discharge cycles period C and battery remaining power q_CFunction, therefore its catagen speed can To be expressed as：

For containing, there are two independent variable q_C, the nonlinear function f (q of C_C, C), it is utilized to the Taylor series expansion of the function of many variables, It is handled according to the linearization approximate of nonlinear differential equation, omits its high math power item, can obtain：

Wherein a₁For degradation factor, a₂For the fatigue damage accumulation factor.This differential equation is solved, following two solutions can be obtained：

Formula (3) can uniformly be written as following form with formula (4)：

q_C=k₁C-k₂e^αC+k₃Q₁ (5)

Formula (5) is battery remaining power q_CWith the degradation mechanism model expression of charge and discharge cycles period C, wherein (α, k₁, k₂, k₃) For unknown-model parameter, Q₁For lithium battery initial capacity, i.e. lithium battery corresponding remaining appearance under the first charge and discharge cycles period Magnitude.

4. the lithium battery residual capacity prediction technique based on mechanism-data-driven model, feature exist as described in claim 1 In modeling error prediction model of the foundation based on LSSVM, is as follows in the step (4)：

(4.1) another that n=500 blocks is taken to be respectively placed under n different nonstandard designs from the lithium battery of model of the same race in step (1) It refers to setting n blocks lithium battery respectively in the condition and range that lithium battery is capable of normal operation to be carried out at the same time cycle charge discharge electric test In n random non-standard operating condition (T_j.c, W_j.c, I_in.j.c, V_in.j.c, I_out.j.c, V_out.j.c) under be carried out at the same time cycle charge discharge Electric test；The specific steps are：It is T in ambient temperature_j.c, relative humidity W_j.cUnder conditions of, first with I_in.j.cElectric current charges Reach V to battery up to full power state, that is, voltage_in.j.c, then with I_out.j.cCurrent discharge to blanking voltage V_out.j.c, so recycle It carries out；

(4.2) kernel function of the Gaussian kernel as LSSVM is selected, and selects optimal nuclear parameter for γ=20, σ²=110；By step (4) operating condition multiple information (C, the T obtained in_j.c, W_j.c, I_in.j.c, V_in.j.c, I_out.j.c, V_out.j.c) and degradation mechanism model Predict error delta q_j.CAs the training sample of LSSVM regression fits, regression modeling is carried out, the modeling error based on LSSVM is obtained Prediction model is denoted as Δ q_j.C=f (C, T_j.c, W_j.c, I_in.j.c, V_in.j.c, I_out.j.c, V_out.j.c)。

5. the lithium battery residual capacity prediction technique based on mechanism-data-driven model, feature exist as described in claim 1 In acquiring current lithium battery residual capacity predicted value to be measured in the step (5) It is as follows：

(5.1) the initial capacity value q of lithium battery to be measured is monitored_x.1And carry it into the residual capacity of step (3) the model lithium battery In degradation mechanism universal model, the residual capacity predicted value converted to lithium battery to be measured under standard condition is obtained

(5.2) extraneous operating condition multiple information sequence (C, T that lithium battery to be measured is presently in are monitored_x.c, W_x.c, I_in.x.c, V_in.x.c, I_out.x.c, V_out.x.c) and carry it into the modeling error prediction model established in step (4), obtain degradation mechanism mould Type predicts error delta q_x.C=f (C, T_x.c, W_x.c, I_in.x.c, V_in.x.c, I_out.x.c, V_out.x.c)；

(5.3) current lithium battery residual capacity predicted value to be measured is acquired