CN109738811B - External short-circuit fault diagnosis method for lithium-ion battery pack based on two-stage model prediction - Google Patents

Abstract

The invention provides a method for diagnosing an external short circuit fault of a lithium ion battery pack based on two-stage model prediction, and relates to the technical field of lithium ion power battery safety. First, an external short-circuit experiment is carried out on the lithium-ion battery pack, and an external short-circuit two-stage equivalent circuit model of the battery pack is constructed, and the battery model parameters are identified offline optimally by using the measured experimental data; then, the battery is judged according to the battery measurement data during operation. The status of the batteries in the group, when it is found that the voltage of some batteries is abnormal, the adjacent battery cells that produce the abnormality are marked as a whole, recorded as an abnormal battery group, and the first-level battery model is activated. If the error of the first-level battery model is less than the critical threshold, then The second-level battery model is triggered, and the model error is obtained by calculation; finally, the fault diagnosis of abnormal batteries is carried out through the agreement between the measured data and the two-level model. The method has simple steps, is easy to implement online, has high reliability, and is suitable for online fault diagnosis and safety management of electric vehicle power batteries.

Description

External short-circuit fault diagnosis method for lithium-ion battery pack based on two-stage model prediction

technical field

The invention relates to the technical field of lithium-ion power battery safety, in particular to a method for diagnosing external short-circuit faults of lithium-ion battery packs based on two-stage model prediction.

Background technique

In recent years, electric vehicles have developed rapidly. The development of electric vehicles is regarded as an effective way to solve environmental pollution, reduce fuel consumption, and build green and environmentally friendly urban transportation. However, safety accidents such as fire and explosion often occur during the application of battery vehicles. , the root cause is often thermal runaway due to battery failure. External short circuit is one of the most common and serious faults in battery faults. When an external short circuit fault occurs, the battery pack generates a large current, which is easy to cause high temperature and high heat of the battery. The duration of an external short circuit fault is often only tens of seconds. It is a very important technical problem to carry out effective, accurate and rapid online fault diagnosis.

At present, most of the existing battery management systems are aimed at battery state estimation and life prediction, etc., but the methods for battery safety and fault diagnosis are still immature, especially the external short-circuit fault diagnosis technology for high-power battery packs is relatively lacking.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a method for diagnosing an external short-circuit fault of a lithium ion battery pack based on two-stage model prediction, aiming at the shortcomings of the above-mentioned prior art; the method has simple steps, is easy to implement online, has high reliability, and is suitable for On-line fault diagnosis and safety management of electric vehicle power battery.

In order to solve the above-mentioned technical problems, the technical scheme adopted by the present invention is:

The present invention provides a method for diagnosing an external short-circuit fault of a lithium-ion battery pack based on two-stage model prediction, comprising the following steps:

Step 1: Carry out the external short-circuit experiment of the battery pack, and record the experimental data, including the measured data of current I _c =[I _c1 ,I _c2 ,...,I _cN ] ^T and the measured data of terminal voltage U _c =[U _c1 ,U _c2 ,…,U _cN ] ^T , where N is the number of data samples, the value of N depends on the current duration and sampling step size in the external short-circuit test, and T represents the transposition of the matrix;

Step 2: establish a dual-level battery model for external short-circuit faults, and perform offline optimal parameter identification on the dual-level battery model based on the experimental data obtained in step 1;

The first-level battery model is an improved equivalent circuit model, and the improved method is as follows: the battery state of charge SOC in the traditional equivalent circuit model is improved to the discharge depth ξ _E during the short-circuit process, and the open-circuit voltage is regarded as Polynomial function of discharge depth ξ _E ;

The specific mathematical expression of the first-stage battery model is:

Among them, k represents the current sampling time, τ=R _p C _p , U _t , U _p , and U _oc represent the terminal voltage, polarization voltage, and open-circuit voltage of the battery pack, respectively; R _p , and R ₀ represent the polar The internal resistance and the ohmic internal resistance, C _p represents the polarized capacitance, i _L represents the battery pack current, _ip represents the current flowing on R _p , Δt is the sampling step size, and ξ _E represents the depth of discharge in the external short-circuit fault.

The second-stage battery model is a half-cell model, and the specific mathematical expression is:

代表恒定电压源；in

represents a constant voltage source;

Step 3: Use the battery management system to monitor the voltage of each cell in the battery pack in real time. When the voltage of some battery cells is lower than the critical threshold Vn, go to Step 4;

Step 4: Trigger the first-level battery model, regard the adjacent abnormal battery cells as an abnormal battery pack, use the battery pack current as the model input, and calculate the predicted voltage output by the model in real time;

Step 5: Calculate the coincidence degree σ between the predicted voltage of the first-stage battery model and the measured voltage, and the duration is T ₁ . If the coincidence degree σ < critical threshold χ ₁ , then eliminate the possibility of external short-circuit fault, and go to step 8 , otherwise, it is initially defined as an external short-circuit fault, triggering the second-level battery model, and entering step 6;

Step 6: Take the battery pack current as the input of the second-stage battery model and calculate the predicted voltage output by the model in real time, and calculate the degree of agreement between the predicted voltage of the _second -stage battery model and the measured voltage σ, and the duration is T2. If the degree σ>critical threshold χ ₂ , confirm that the abnormality is caused by an external short-circuit fault, locate the position of the abnormal battery cell and go to step 8; otherwise, increase the diagnosis duration to T ₃ and go to step 7;

Step 7: Use the second-level battery model to repeatedly judge the degree of fit, if the fit degree σ < critical threshold χ ₂ , the possibility of external short-circuit fault is excluded, if the fit degree σ > critical threshold χ ₂ , it is confirmed as an external short-circuit fault;

Step 8: Store and output the diagnosis result, return to Step 3, and wait for the next operation.

与电池内阻R₀、短路电阻R_S连为回路；在模型2中，有一个恒定电压源

与RC环节连并产生端电压Ut，RC环节由一个电容C与极化内阻Rp并联组成；整个电池的开路电压为可变电压源

与恒定电压源

之和：The two-stage battery circuit model in the step 2 is divided into two levels, wherein the first-level battery model is an overall battery model, the second-level battery model is a half-cell model; the second-level battery model is a half-cell model, and the second-level battery model is a half-cell model. The second-level modeling method is to regard the battery as a two-part equivalent circuit model, including model 1 and model 2, that is, the sum of model 1 and model 2 is the overall model of the battery, and the second-level battery model refers specifically to model 2; Model 1, consists of a variable voltage source

It is connected with the internal resistance R ₀ of the battery and the short-circuit resistance R _S as a loop; in Model 2, there is a constant voltage source

It is connected with the RC link and generates the terminal voltage Ut. The RC link is composed of a capacitor C and the polarization internal resistance Rp in parallel; the open circuit voltage of the entire battery is a variable voltage source

with constant voltage source

Sum:

The offline optimality parameter identification in the step 2 is to use the experimental current measurement value I _c as the model input, the terminal voltage output U=[U ₁ , U ₂ , . . . , U _N ] ^T as the model output, using a global optimization algorithm The offline optimality identification is performed on the model parameters in step 2, and the model identification process needs to identify the two-level model parameters separately, and the parameters of the two-level models are independent of each other.

The definition of the degree of fit σ in the step 5 is: the reciprocal of the root mean square error between the model prediction result and the actual test result within a certain duration, namely:

Among them, ρ is the sampling times in the duration T, U _t,m is the model prediction result of the terminal voltage, U _t is the online measurement data of the terminal voltage, and θ _n represents the model parameter matrix.

The critical threshold value χ ₁ in the step 6 is the critical threshold value of the fit degree of the first-stage battery model, and the value of the critical threshold value needs to be slightly lower than the model fit degree calculation result during the experiment;

The critical threshold value χ ₂ in the step 7 is the critical threshold value of the fit degree of the second-stage battery model, and the value of the critical threshold value needs to be slightly lower than the model fit degree calculation result in the experiment.

The beneficial effects of adopting the above technical solutions are as follows: the method for diagnosing external short-circuit faults of lithium-ion battery packs based on two-stage model prediction provided by the present invention adopts a two-stage optimized equivalent circuit model, wherein the first-stage model is The overall model of the battery contains more parameters to be identified, and the model has good adaptability but slightly lower accuracy. The second-level model is a half-cell model, which contains fewer parameters to be identified and has high model accuracy; external short-circuiting of the battery pack is used. The data identifies the parameters of the two-stage model, and the online fault diagnosis of the external short circuit is carried out through the agreement between the measured data of the battery pack and the model prediction. The method has simple steps, is easy to implement online, has high reliability, and is suitable for online fault diagnosis and safety management of electric vehicle power batteries.

Description of drawings

1 is a two-stage equivalent circuit model for external short-circuit fault diagnosis provided by an embodiment of the present invention, wherein a is the first-stage battery model; b is the model 1 in the second-stage battery model; c is the second-stage battery model model 2;

FIG. 2 is a flowchart of an external short-circuit online diagnosis and estimation method based on two-stage model prediction provided by an embodiment of the present invention;

3 is a diagram of an identification error analysis result of an external short-circuit two-stage model provided by an embodiment of the present invention, wherein a is an identification error analysis result diagram, and b is an error schematic diagram of the identification error analysis result;

Fig. 4 is a diagnostic diagram of an external short circuit of a battery pack provided by an embodiment of the present invention, wherein a-1 is a voltage diagnosis diagram, a-2 is a partial enlarged diagram of H in Fig. a-1, and b-1 is a two-stage model error diagram , b-2 is a partial enlarged view of Z in Figure b-1;

5 is a diagram showing the results of an external short circuit experiment provided by an embodiment of the present invention, wherein a-1 is a diagram showing the results of a voltage experiment, a-2 is a partial enlarged diagram of AB in FIG. a-1, and b is a diagram showing the results of a current experiment.

Detailed ways

The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following examples are intended to illustrate the present invention, but not to limit the scope of the present invention.

In the example, the 18650NMC cylindrical lithium-ion power battery is taken as an example, its rated voltage is 3.6V, its nominal capacity is 2.4Ah, and a battery pack is composed of 6 battery cells with SOH value > 0.96; the experimental equipment adopts: NEU_ESCTEST02 test bench With Haixiang Instrument GD-2045D temperature control box,

The method of this embodiment is as follows.

The present invention provides a method for diagnosing an external short-circuit fault of a lithium-ion battery pack based on two-stage model prediction, as shown in FIG. 2 , including the following steps:

Step 1: Carry out the external short-circuit experiment of the battery pack, and record the experimental data, including the measured data of current I _c =[I _c1 ,I _c2 ,...,I _cN ] ^T and the measured data of terminal voltage U _c =[U _c1 ,U _c2 ,…,U _cN ] ^T , where N is the number of data samples, the value of N depends on the current duration and sampling step size in the external short-circuit test, and T represents the transposition of the matrix;

Step 2: Establish a dual-level battery model for external short-circuit faults, and perform offline optimal parameter identification on the dual-level battery model through the experimental data obtained in step 1. The offline optimal parameter identification is the experimental current measurement value I. _c is used as the model input, the terminal voltage output U=[U ₁ , U ₂ ,..., U _N ] ^T is used as the model output, the global optimization algorithm is used to identify the model parameters in step 2 for offline optimality, and the model identification process The parameters of the two-level model need to be identified separately, and the parameters of the two-level model are independent of each other.

The first-level battery model is an improved equivalent circuit model, as shown in Figure a in Figure 1, the improved method is: the battery state of charge SOC in the traditional equivalent circuit model is improved to the short-circuit process. the depth of discharge ξ _E , and consider the open circuit voltage as a polynomial function of the depth of discharge ξ _E ;

The specific mathematical expression of the first-stage battery model is:

Among them, k represents the current sampling time, τ=R _p C _p , U _t , U _p , and U _oc represent the terminal voltage, polarization voltage, and open-circuit voltage of the battery pack, respectively; R _p , and R ₀ represent the polar The internal resistance and the ohmic internal resistance, C _p represents the polarized capacitance, i _L represents the battery pack current, _ip represents the current flowing on R _p , Δt is the sampling step size, and ξ _E represents the depth of discharge in the external short-circuit fault.

The open circuit voltage is expressed as a polynomial as follows

In the formula, N _p is the degree of the polynomial, α _i represents the polynomial coefficient, and ξ _E represents the discharge depth in the external short-circuit fault. The specific calculation method is as follows:

where _QR is the nominal capacity.

与电池内阻R₀、短路电阻R_S连为回路；在模型2中，如图1所示中c图所示，有一个恒定电压源

与RC环节连并产生端电压Ut，RC环节由一个电容C与极化内阻Rp并联组成；整个电池的开路电压为可变电压源

与恒定电压源

之和：The second-stage battery model is a half-cell model. The second-stage modeling method is to treat the battery as a two-part equivalent circuit model, including model 1 and model 2, that is, the sum of model 1 and model 2 is the overall model of the battery. The secondary battery model refers specifically to model 2; in model 1, as shown in Figure b in Figure 1, a variable voltage source

It is connected with the internal resistance R ₀ of the battery and the short-circuit resistance R _S as a loop; in Model 2, as shown in Figure c in Figure 1, there is a constant voltage source

It is connected with the RC link and generates the terminal voltage Ut. The RC link is composed of a capacitor C and the polarization internal resistance Rp in parallel; the open circuit voltage of the entire battery is a variable voltage source

with constant voltage source

Sum:

The second-stage battery model is a half-cell model, and the specific mathematical expression is:

代表恒定电压源；in

represents a constant voltage source;

For the first-stage battery model, the parameters to be identified θ ₁ =[α ₁ ,α ₂ ,...,α ₁₀ ,τ,R _p ,R ₀ ] have a total of 13 parameters, and for the second-stage battery model, with the identification parameter θ ₂ =[U ₀ , τ, R _p ] has three parameters in total. The two-stage battery model is used to identify the offline optimality parameters respectively. The identification method can adopt the global optimization method. In this embodiment, the genetic algorithm is used to identify the parameters. The selection method does not limit the present invention. The identification error is shown in Figure 3. It can be seen that the model constructed in this way has a very high prediction accuracy for the second-level battery model. After the identification is completed, record the identification results as shown in Table 1-Table 2:

Table 1 Parameter identification results of the first-stage battery model

Table 2 Parameter identification results of the second-stage battery model

parameter Identification result U₀(mV) 589.7 τ(s) 6.9 R_p(mΩ)) 6.5

Step 3: Use the battery management system to monitor the voltage of each cell of the battery pack in real time. If the voltage of some battery cells is lower than the critical threshold Vn, go to Step 4;

In this embodiment, the cell voltage critical threshold Vn=2.0V is set, and the critical threshold is set slightly lower than the normal discharge cut-off voltage of the battery; the battery management system is a battery management system for a new energy vehicle, and mainly has the function of current, voltage and temperature acquisition. , battery state estimation, overvoltage protection and safety management system;

Step 4: Trigger the first-level battery model, regard the adjacent abnormal battery cells as an abnormal battery pack, use the battery pack current as the model input, and calculate the predicted voltage output by the model in real time;

Step 5: Calculate the coincidence degree σ between the predicted voltage of the first-stage battery model and the measured voltage, and the duration is T ₁ . If the coincidence degree σ < critical threshold χ ₁ , then eliminate the possibility of external short-circuit fault, and go to step 8 , otherwise, it is initially defined as an external short-circuit fault, triggering the second-level battery model, and entering step 6;

The definition of fit σ is: the inverse of the root mean square error between the model prediction result and the actual test result within a certain duration, namely:

Among them, ρ is the sampling times in the duration T, U _t,m is the model prediction result of the terminal voltage, U _t is the online measurement data of the terminal voltage, and θ _n represents the model parameter matrix.

In this embodiment, the durations T ₁ =1.0s, T ₂ =3.0s, and T ₃ =10.0s are set; the critical threshold χ ₁ =3.5, and the critical threshold χ ₂ =30.

The critical threshold χ ₁ is the critical threshold of the fit of the first-stage battery model, and the value of the critical threshold needs to be slightly lower than the model fit calculation result in the experiment;

The critical threshold χ ₂ is the critical threshold of the fit of the second-stage battery model, and the value of the critical threshold should be slightly lower than the result of the model fit in the experiment.

The critical threshold of model fit is determined according to the experimental results, and the value of the critical threshold should be slightly lower than the model fit calculation result in the experiment, so as to ensure that there will be no missed judgment in the diagnosis process; according to the experimental results of external short circuit, the model The calculation results of the goodness of fit are shown in Table 3:

Table 3 Model fit

number of experiments First-level model fit Second-level model fit 1 3.9 33.8 2 4.1 32.4 3 4.4 31.7 4 3.7 35.5 5 4.6 37.1

Therefore, the critical threshold χ ₁ =3.5 and the critical threshold χ ₂ =30 are set.

Step 6: Take the battery pack current as the input of the second-stage battery model and calculate the predicted voltage output by the model in real time, and calculate the degree of agreement between the predicted voltage of the _second -stage battery model and the measured voltage σ, and the duration is T2. If the degree σ>critical threshold χ ₂ , confirm that the abnormality is caused by an external short-circuit fault, locate the position of the abnormal battery cell and go to step 8; otherwise, increase the diagnosis duration to T ₃ and go to step 7;

Step 7: Use the second-level battery model to repeatedly judge the degree of fit, if the fit degree σ < critical threshold χ ₂ , the possibility of external short-circuit fault is excluded, if the fit degree σ > critical threshold χ ₂ , it is confirmed as an external short-circuit fault;

Step 8: Store and output the diagnosis result, return to Step 3, and wait for the next operation.

Online operation, the battery management system is used to monitor the voltage of each cell of the battery pack in real time. In this embodiment, the battery pack composed of 6 batteries is short-circuited, and the battery pack voltage rapidly drops below 0.5V, which is lower than the critical threshold. Therefore, the first-level battery model is triggered, the abnormal battery cells are formed into abnormal battery packs according to adjacent individuals, the battery pack current is used as the model input, and the predicted voltage output by the model is calculated in real time, as shown in the solid line of a-1 in Figure 4. As shown, the measurement results of the terminal voltage of the battery pack are obtained online at the same time, as shown by the dotted line in a-1 in Figure 4, and b-1 in Figure 4 is a two-stage model error map of 3-7 seconds.

Calculate the fit σ between the model predicted voltage and the measured voltage, compare the fit with the critical threshold, and perform fault diagnosis according to the logical judgment process described in the Summary of the Invention. In this embodiment, the first-level battery model The coincidence degree σ≈6.34, triggering the second-level battery model, during the operation of the second-level battery model, the error between the model prediction results and the measured data is less than 20mV, after calculation, the second-level battery model coincidence degree σ≈93.7> critical threshold χ ₂ , according to the judgment criterion, it is confirmed as an external short-circuit fault, the online diagnosis process is completed, and the diagnosis result is stored and output, as shown in FIG. 5 .

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that it can still be The technical solutions described in the foregoing embodiments are modified, or some or all of the technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions depart from the scope defined by the claims of the present invention.

Claims (5)

1. A lithium ion battery pack external short circuit fault diagnosis method based on two-stage model prediction is characterized by comprising the following steps: the method comprises the following steps:

step 1, carrying out a short circuit experiment outside the battery pack, and recording experimental data including current measurement data I_c＝[I_c1,I_c2,…,I_cN]^TTerminal voltage measurement data U_c＝[U_c1,U_c2,…,U_cN]^TN is the number of data samples, the value of N depends on the duration time of current and sampling step length in an external short circuit test, and T represents the transposition of a matrix;

step 2, establishing a double-stage battery model of the external short circuit fault, and respectively carrying out off-line optimality parameter identification on the double-stage battery model through the experimental data obtained in the step 1;

the first-stage model is an improved equivalent circuit model, and the improved method comprises the following steps: the battery state of charge SOC in the traditional equivalent circuit model is improved into the depth of discharge xi in the short circuit process_EAnd the open circuit voltage is regarded as the depth of discharge xi_EA polynomial function of (a);

the specific mathematical expression form of the first-stage battery model is as follows:

where k denotes the current sampling instant, τ ═ R_pC_p，U_t、U_pAnd U_ocRespectively representing the terminal voltage, the polarization voltage and the open-circuit voltage of the battery pack; r_pAnd R₀Then respectively represent the polarization internal resistance and the ohm internal resistance, C_pRepresenting the polarization capacitance, i_LRepresenting the battery current, i_pRepresents R_pThe current flowing upwards, delta t is the sampling step length, xi_EIndicating a depth of discharge in an external short-circuit fault;

the second-stage battery model is a half-battery model, and the specific mathematical expression form is as follows:

in the formula

Represents a constant voltage source;

and step 3: monitoring the voltage of each single battery in real time by using a battery management system, and when the voltage of partial single batteries is lower than a critical threshold value V_nEntering step 4;

and 4, step 4: triggering a first-stage battery model, regarding adjacent abnormal single batteries as an abnormal battery pack, inputting the current of the battery pack as a model, and calculating the predicted voltage output by the model in real time;

and 5: calculating the goodness of fit sigma between the predicted voltage and the actually measured voltage of the first-stage battery model, and the duration T₁At the moment, if the goodness of fit sigma is less than the critical threshold value chi₁If not, preliminarily defining the external short-circuit fault, triggering a second-stage battery model, and entering the step 6;

step 6: taking the current of the battery pack as the input of a second-stage battery model, calculating the predicted voltage output by the model in real time, and calculating the goodness of fit sigma between the predicted voltage and the actually measured voltage of the second-stage battery model and the duration T₂At the moment, if the goodness of fit sigma is larger than the critical threshold value chi₂If the abnormal condition is caused by the external short circuit fault, the position of the abnormal battery monomer is positioned and the step 8 is carried out; otherwise, increase the diagnostic duration to T₃And go to step 7;

and 7: repeatedly judging the goodness of fit by adopting a second-stage battery model, and if the goodness of fit sigma is less than a critical threshold value chi₂The possibility of external short-circuit faults is eliminated if the goodness of fit sigma > the critical threshold chi₂If yes, confirming as an external short-circuit fault;

and 8: and storing and outputting the diagnosis result, returning to the step 3, and waiting for the next operation.

2. The lithium ion battery pack external short-circuit fault diagnosis method based on the two-stage model prediction as claimed in claim 1, wherein: what is needed isThe two-stage battery models in the step 2 are divided into two stages in total, wherein the first-stage battery model is a battery integral model, and the second-stage battery model is a half-battery model; the second-stage modeling method is to regard the battery as a two-part equivalent circuit model, and comprises a model 1 and a model 2, namely the sum of the model 1 and the model 2 is a battery integral model, and the second-stage battery model refers in particular to the model 2; in model 1, a variable voltage source is used

And internal resistance R of battery₀Short-circuit resistor R_SAre connected into a loop; in model 2, there is a constant voltage source

Connected with RC link and generating terminal voltage U_tThe RC link consists of a capacitor C and a polarization internal resistance R_pAre connected in parallel; the open-circuit voltage of the whole battery is a variable voltage source

And a constant voltage source

And (3) the sum:

3. the lithium ion battery pack external short-circuit fault diagnosis method based on the two-stage model prediction as claimed in claim 1, wherein: the off-line optimality parameter in step 2 is identified as the experimental current measurement value I_cAs model input, terminal voltage output U ═ U₁,U₂,…,U_N]^TAs model output, performing offline optimality identification on the model parameters in the step 2 by using a global optimization algorithm, wherein the identification process of the model needs to respectively identify the parameters of the two-stage battery model, and the two-stage battery is powered onThe parameters of the pool model are independent of each other.

4. The lithium ion battery pack external short-circuit fault diagnosis method based on the two-stage model prediction as claimed in claim 1, wherein: the definition of the goodness of fit σ in the step 5 is as follows: the inverse of the root mean square error of the model predicted results and the actual test results over a certain duration, namely:

where ρ is the number of samples in the duration T, U_tM is the model prediction result of terminal voltage, U_tFor on-line measurement of terminal voltage, theta_nRepresenting a model parameter matrix.

5. The lithium ion battery pack external short-circuit fault diagnosis method based on the two-stage model prediction as claimed in claim 1, wherein: the critical threshold value χ in the step 5₁The value of the critical threshold value is slightly lower than the calculation result of the goodness of fit of the model during the experiment;

the critical threshold value χ in the step 7₂And the value of the critical threshold value is a critical threshold value of the goodness of fit of the second-stage battery model, and the value of the critical threshold value needs to be slightly lower than a model goodness of fit calculation result in an experiment.

Images