CN117807942B - Two-stage battery model parameter identification method - Google Patents

Abstract

The invention discloses a two-stage battery model parameter identification method, which comprises the following steps: s1, obtaining a noisy open-circuit voltage sequence, namely OCVN sequence, based on a battery n-order RC equivalent circuit model and sampling current; s2, establishing a mapping relation between OCVN and the real SOC; s3, fitting OCVN-SOC mapping points by adopting an n-order polynomial, and calculating a fitting error function; s4, obtaining final battery model parameters. The two-stage battery model parameter identification method provided by the invention avoids dependence on known OCV-SOC curves and specific working conditions, has low parameter identification difficulty and is not easy to sink into local optimum. Meanwhile, the invention is beneficial to battery parameter updating and health state estimation under the full life cycle.

Description

A two-stage battery model parameter identification method

Technical Field

The present invention relates to the field of lithium-ion battery model parameter identification, and in particular to a two-stage battery model parameter identification method for a battery dynamic discharge condition when an OCV-SOC curve is unknown.

Background Art

There are many existing battery model parameter identification methods, such as Kalman filter (KF) based on filtering and particle swarm optimization (PSO) based on optimization. However, most of these methods rely on the premise that the battery OCV-SOC curve is known for parameter identification, and also require specific working conditions, such as the hybrid power pulse characteristic (HPPC) test condition. The identification principle is shown in Figure 1.

Although the recursive least squares method (RLS) does not rely on the known battery OCV-SOC curve, the parameters identified in dynamic conditions fluctuate greatly and usually require secondary processing. In addition, some scholars define several OCV points and identify them together with the battery model parameters, but this will lead to a high-dimensional optimization space, which greatly increases the amount of calculation for optimization and is more likely to fall into local optimality. The identification principle of this type of method is shown in Figure 2. Therefore, how to identify battery model parameters for ordinary battery dynamic discharge conditions when the battery OCV-SOC curve is unknown is still a difficult problem.

Summary of the invention

In order to solve the problems of existing battery model parameter identification methods, such as dependence on known OCV-SOC curves and specific working conditions, large amount of calculation, high optimization difficulty, and easy falling into local optimum, the present invention proposes a two-stage battery model parameter identification method, comprising the following steps:

S1. Based on the battery n-order RC equivalent circuit model and the sampling current, the battery internal resistance voltage drop sequence and polarization voltage sequence are calculated using the current battery model parameters combined with the current sequence, and the battery internal resistance voltage drop sequence and polarization voltage sequence are subtracted from the battery terminal voltage sequence to obtain the noisy open circuit voltage sequence, i.e., the OCVN sequence;

S2, combining the OCVN sequence in S1 with the real SOC sequence from the Coulomb counting method to establish a mapping relationship between OCVN and the real SOC;

S3, fitting the OCVN-SOC mapping points using an n-th order polynomial and calculating the fitting error function;

S4. Based on the current battery model parameters and the fitting error function, an optimization algorithm is used to obtain the best battery model parameters.

Preferably, the calculation formula of the OCVN sequence is as follows:

U _ocvn,s =U _ocv,s +v _s =U _t,s -U _0,s -(U _1,s +U _2,s +...+U _n,s );

Among them, U _ocvn,s represents the OCVN sequence, U _ocv,s is the real OCV sequence of the battery, _vs is the noise voltage sequence, U _t,s is the battery terminal voltage sequence, U _0,s is the battery internal resistance voltage drop sequence, and _Un,s is the polarization voltage sequence on the nth RC structure.

Preferably, the calculation method of the battery internal resistance voltage drop sequence and the polarization voltage sequence on the nth RC structure is:

The battery internal resistance voltage drop at time k is R ₀ I _k , where R ₀ and I _k are the ohmic internal resistance and the operating current respectively; the polarization voltage on the nth RC structure at time k is Where _Rn and _Cn are the polarization internal resistance and polarization capacitance of the nth RC structure, respectively, △t is the time step, and _Un,k-1 is the polarization voltage on the nth RC structure at time k-1.

Preferably, the method for fitting the OCVN-SOC mapping points in S3 is:

The monotonic shape of the real OCV-SOC curve of the battery is used as prior information, and the OCVN-SOC mapping points are fitted based on the preset n-order polynomial shape function of the OCV-SOC curve.

Preferably, the method of S4 is:

The battery model parameters are taken as optimization variables, the fitting error function is taken as the objective function, and the particle swarm optimization algorithm is used for cyclic iteration. The set of parameters obtained in the last cycle is the optimal battery model parameters.

Preferably, the battery model parameter identification is realized using the battery dynamic operating condition data in the intermediate SOC region of 90%-20%.

Beneficial effects of the present invention:

(1) The present invention proposes a two-stage battery model parameter identification method for a dynamic discharge condition of a battery when the OCV-SOC curve is unknown. Compared with the existing battery model parameter identification method, the two-stage battery model parameter identification method proposed in the present invention reduces the dimension of optimization variables, which not only reduces the amount of calculation, but also reduces the difficulty of optimization and is not easy to fall into local optimality.

(2) The battery model parameter identification method proposed in the present invention can realize model parameter identification by using the dynamic discharge condition of the battery when the OCV-SOC curve is unknown, which can ensure that the parameter identification method can still work well offline under different degrees of battery aging, and is helpful for battery parameter update and health status estimation throughout the life cycle.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a battery model parameter identification method of an existing known OCV-SOC curve;

FIG2 is a schematic diagram of a battery model parameter identification method for an existing unknown OCV-SOC curve;

FIG3 is a schematic diagram of a two-stage battery model parameter identification method according to an embodiment of the present invention;

FIG4 is a schematic diagram of the first stage of a two-stage battery model parameter identification method according to an embodiment of the present invention;

FIG5 is a schematic diagram of the second stage of the two-stage battery model parameter identification method according to an embodiment of the present invention;

6 is a schematic diagram showing the comparison results between the actual terminal voltage and the model terminal voltage under the DST condition of 25° C. according to an embodiment of the present invention;

FIG7 is a schematic diagram showing the comparison results between the actual terminal voltage and the model terminal voltage under FUDS conditions at 25° C. according to an embodiment of the present invention;

FIG. 8 is a schematic diagram showing a comparison result between the actual terminal voltage and the model terminal voltage under FUDS conditions at 25° C. according to an embodiment of the present invention.

DETAILED DESCRIPTION

In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and examples.

The present application discloses a two-stage battery model parameter identification method. In one embodiment, the implementation principle thereof is shown in FIG3 and includes the following steps:

S1, based on the battery n-order RC equivalent circuit model and the sampling current, as shown in FIG4, in the first stage, the battery internal resistance voltage drop sequence and the polarization voltage sequence are calculated by combining the current battery model parameters with the current sequence, and the battery internal resistance voltage drop sequence and the polarization voltage sequence are subtracted from the battery terminal voltage sequence to obtain the noisy open circuit voltage sequence, i.e., the OCVN sequence;

The calculation formula of OCVN sequence is as follows:

U _ocvn,s =U _ocv,s +v _s =U _t,s -U _0,s -(U _1,s +U _2,s +...+U _n,s );

Among them, U _ocvn,s represents the OCVN sequence, U _ocv,s is the real OCV sequence of the battery, _vs is the noise voltage sequence, U _t,s is the battery terminal voltage sequence, U _0,s is the battery internal resistance voltage drop sequence, and _Un,s is the polarization voltage sequence on the nth RC structure.

The battery internal resistance voltage drop at time k is R ₀ I _k , where R ₀ and I _k are the ohmic internal resistance and the operating current respectively; the polarization voltage on the nth RC structure at time k is Where _Rn and _Cn are the polarization internal resistance and polarization capacitance of the nth RC structure, respectively, △t is the time step, and _Un,k-1 is the polarization voltage on the nth RC structure at time k-1.

S2. In the second stage, as shown in FIG5 , the OCVN sequence in S1 is combined with the real SOC sequence from the Coulomb counting method to establish a mapping relationship between OCVN and the real SOC;

S3, introducing the monotonic shape of the real OCV-SOC curve of the battery as prior information, fitting the OCVN-SOC mapping points based on the preset n-order polynomial shape function of the OCV-SOC curve, and calculating the fitting error function;

In one embodiment, based on a preset sixth-order polynomial morphology function of the OCV-SOC curve, the OCVN-SOC mapping points within the SOC range of 90%-20% are fitted using the least squares method, and the fitting error function is calculated using the mean square error (MSE) formula, which is as follows:

Wherein, N is the total number of samples of OCVN in the middle SOC region of 90%-20%; OCVN _i is the OCVN at a certain SOC, and OCV _i is the OCV at the corresponding SOC on the fitting curve.

S4. Using the battery model parameters as optimization variables and the fitting error function as the objective function, the particle swarm optimization (PSO) algorithm is used to iterate cyclically to find the best set of battery model parameters. The set of parameters obtained in the last cycle will be saved as the best battery model parameters.

The theoretical basis of this embodiment is that the more accurate the battery model parameters are, the closer the output OCVN is to the actual OCV of the battery, and the smaller the error of the polynomial morphological function fitting is.

Through experiments, it is found that the OCVN fluctuation calculated by the battery model parameters identified by the battery dynamic operating condition data in the middle SOC area is smaller than that obtained by using the high SOC and low SOC areas, and because in actual battery applications, it is ensured as much as possible that it does not perform deep charging and discharging. Therefore, the battery dynamic operating condition data range selected in the second stage of this embodiment is the middle SOC area of 90%-20%, and similarly, the fitting range is also the middle SOC area of 90%-20%.

In a specific embodiment, in order to verify the feasibility of the scheme, the present invention was experimentally verified using a public data set of lithium iron phosphate batteries from the University of Maryland. The public data set contains battery data collected at seven different temperatures, namely 0°C, 10°C, 20°C, 25°C, 30°C, 40°C and 50°C, based on three different vehicle driving conditions, namely, dynamic stress test (DST), US06 driving chart and Federal City Driving Chart (FUDS), and the data sampling time is 1 second.

In the experiment, based on the first-order RC battery equivalent circuit model, the invented two-stage battery model parameter identification method was used to perform parameter identification on the DST data set at different temperatures, and the battery model parameters at different temperatures were obtained as shown in Table 1.

Table 1: Results of parameter identification using DST operating data at different temperatures

Temperature(℃) Ohm internal resistance (me) Polarization internal resistance (mΩ) Polarization capacitance(kF) 0 0.198 0.057 702 10 0.171 0.042 840 20 0.159 0.031 967 25 0.158 0.027 996 30 0.154 0.024 1061 40 0.153 0.019 1203 50 0.153 0.016 1267

It can be seen from Table 1 that as the temperature decreases, the ohmic internal resistance and polarization internal resistance of the battery gradually increase, and the polarization capacitance gradually decreases, which is consistent with the actual working characteristics of the battery.

Based on the battery model parameters identified at each temperature, combined with the DST, US06 and FUDS operating currents, the battery equivalent circuit model terminal voltage at different operating currents at each temperature can be calculated. The comparison between the model terminal voltage and the real terminal voltage at 25°C is shown in Figures 6, 7 and 8. It can be seen from the figure that the difference between the model terminal voltage and the real terminal voltage is small, especially in the middle SOC area of 90%-20%.

Table 2 summarizes the model terminal voltage errors under DST, US06 and FUDS conditions at different temperatures using root mean square error (RMSE) and maximum error (MAXE) functions. The specific formula is as follows:

Where M is the total number of samples; Vi _is the measured actual terminal voltage of the battery, It is the terminal voltage of the battery equivalent circuit model calculated based on the identification parameters.

Table 2: Model terminal voltage error under DST, US06 and FUDS conditions at different temperatures

It can be seen from Table 2 that in the intermediate SOC region of 90%-20%, the RMSE of the battery model terminal voltage error at different temperatures is at the mV level, with a maximum of only 7.3mV and a minimum of 1.7mV, which can meet the requirements of actual battery applications for identification accuracy. Even in the full range of SOC, the RMSE of the battery model terminal voltage error at different temperatures is as high as 23.2mV and as low as 5.2mV. The verification results fully demonstrate that the identified battery model parameters have high accuracy, especially in the commonly used intermediate SOC region of 90%-20%. The experimental results prove the feasibility and effectiveness of the two-stage battery model parameter identification method of the present invention.

The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims (3)

1. A two-stage battery model parameter identification method, characterized in that it includes the following steps: S1. Based on the battery n-order RC equivalent circuit model and the sampling current, the battery internal resistance voltage drop sequence and polarization voltage sequence are calculated using the current battery model parameters combined with the current sequence, and the battery internal resistance voltage drop sequence and polarization voltage sequence are subtracted from the battery terminal voltage sequence to obtain the noisy open circuit voltage sequence, i.e., the OCVN sequence; The calculation formula of OCVN sequence is as follows: U _ocvn,s =U _ocv,s +v _s =U _t,s -U _0,s -(U _1,s +U _2,s +...+U _n,s ); Among them, U _ocvn,s represents the OCVN sequence, U _ocv,s is the real OCV sequence of the battery, _vs is the noise voltage sequence, U _t,s is the battery terminal voltage sequence, U _0,s is the battery internal resistance voltage drop sequence, and U _n,s is the polarization voltage sequence on the nth RC structure; The battery internal resistance voltage drop at time k is R ₀ I _k , where R ₀ and I _k are the ohmic internal resistance and the operating current respectively; the polarization voltage on the nth RC structure at time k is Where R _n and C _n are the polarization internal resistance and polarization capacitance of the nth RC structure, respectively, △t is the time step, and _Un,k-1 is the polarization voltage on the nth RC structure at time k-1; S2, combining the OCVN sequence in S1 with the real SOC sequence from the Coulomb counting method to establish a mapping relationship between OCVN and the real SOC; S3. Fit the OCVN-SOC mapping points using an n-order polynomial and calculate the fitting error function; the OCVN-SOC mapping point fitting method is: The monotonic shape of the real OCV-SOC curve of the battery is used as prior information, and the OCVN-SOC mapping points are fitted based on the preset n-order polynomial shape function of the OCV-SOC curve; S4. Based on the current battery model parameters and the fitting error function, an optimization algorithm is used to obtain the best battery model parameters. 2. A two-stage battery model parameter identification method according to claim 1, characterized in that the method of S4 is: The battery model parameters are taken as optimization variables, the fitting error function is taken as the objective function, and the particle swarm optimization algorithm is used for cyclic iteration. The set of parameters obtained in the last cycle is the optimal battery model parameters. 3. A two-stage battery model parameter identification method according to claim 2, characterized in that the battery model parameter identification is realized by using battery dynamic operating condition data in the intermediate SOC region of 90%-20%.