Battery parameter identification method based on VFFRLS algorithm and considering environment temperature influence

Abstract

The invention discloses a battery parameter identification method based on a VFFLS algorithm and considering environment temperature influence, and belongs to the technical field of battery parameter identification. Comprising the following steps: establishing a second-order RC model of a battery; the influence of the environment temperature is considered, an OCV-SOC-T three-dimensional relation is established, and OCV is obtained; and constructing a VFFRLS identification model to identify battery parameters. According to the method, the error of parameter identification caused by environment temperature change is reduced, the temperature self-adaptive capability is achieved, multiple information matrixes are applied, and the identification precision, real-time performance and stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery parameter identification, and in particular to a battery parameter identification method based on a VFFRLS algorithm and taking into account the influence of ambient temperature. Background Art

[0002] Lithium batteries are widely used in electric vehicles, household appliances, energy storage systems and other fields due to their high energy density, high output voltage and easy management. With the widespread application of batteries, the capacity, safety, state of charge (SOC), state of health (SOH) and endurance of batteries have become the focus of attention, but these cannot be directly characterized or measured in batteries. In order to solve this problem, it is often necessary to model the battery and identify the battery model parameters. Then the battery management system estimates the battery state and obtains information such as SOC, SOH and remaining service life. The accuracy of the model parameters is directly related to the estimation of the battery state. Accurate parameters help the battery to achieve early fault detection, predict the degradation process and give risk warnings, so as to perform battery maintenance and replacement. At the same time, it is of great value to improve system reliability and reduce operating costs, so it is very important to improve the accuracy of battery model parameters.

[0003] Battery model parameters often change with ambient temperature, SOC, and service life. How to accurately identify battery parameters under different ambient temperatures and working conditions and update them in real time is an urgent problem that needs to be solved.

[0004] At present, there are two main methods for battery parameter identification: offline identification and online identification.

[0005] The offline parameter identification method is to conduct a series of tests before the battery leaves the factory to obtain the battery model parameters. Two types of battery tests need to be performed, one is the OCV-SOC curve acquisition experiment, and the other is the model parameter identification experiment. The purpose of the OCV-SOC curve acquisition experiment is to provide OCV for model parameter identification. The specific operation is to discharge the battery with equal capacity, measure the OCV values ​​at different SOC points, and then fit the points to obtain the entire OCV-SOC curve. The model parameter identification experiment uses the pulse experiment method to perform pulse discharge on the battery at equal capacity intervals, record the changes in the terminal voltage, and then fit the terminal voltage curve. Combined with the OCV value obtained previously, the battery model parameters are obtained. In general, different SOCs at equal capacity intervals correspond to different parameters, and eventually a set of model parameters for the entire SOC cycle will be obtained.

[0006] The online identification method is based on the voltage and current data at the current moment for identification, and the parameters are updated online as the voltage and current change in real time. The online identification methods are mainly divided into three categories: the recursive least squares algorithm (RLS) and some of its improved methods, the Kalman filter algorithm, and the intelligent algorithm. Among them, the recursive least squares method is the most classic and relatively stable online identification algorithm. It uses the estimated parameters and storage vectors of the previous sampling moment and the system input and output values ​​of the current moment to calculate, obtain the value of the estimated parameters at the current moment, and then recursively push the estimated parameters and storage vectors at the current moment to the next moment, and continue to identify the parameters at the next moment. However, the recursive least squares method is an algorithm with infinite memory length. For the battery system, the least squares method has more and more old data in the recursive operation process, which will cause the recursive result to not reflect the characteristics of the new data well. In order to avoid data redundancy, Liu Houde, Kang Longyun and others proposed "a least squares lithium battery model parameter identification method with forgetting factor", introduced the forgetting factor, alleviated the problem of old data superposition in the recursive operation process, and overcame the "data saturation" phenomenon.

[0007] In the above existing technologies, the offline identification method has poor real-time performance and cannot update parameters as the ambient temperature and battery life change; the online identification method has low accuracy and unstable identification results. Among them, the forgetting factor of the recursive least squares algorithm with forgetting factor (FFRLS) is fixed, and its optimal value is often determined by trial and error, which cannot be applied to all working conditions.

[0008] Therefore, a battery parameter identification method based on the VFFRLS algorithm and considering the influence of ambient temperature is needed to improve the accuracy, real-time performance and temperature adaptability of battery parameter identification. Summary of the invention

[0009] The purpose of the present invention is to propose a battery parameter identification method based on a VFFRLS algorithm and taking into account the influence of ambient temperature, comprising the following steps:

[0010] Establish a second-order RC model of the battery;

[0011] Considering the influence of ambient temperature, the three-dimensional relationship of OCV-SOC-T is established to obtain OCV;

[0012] A VFFRLS identification model is constructed to identify battery parameters.

[0013] The second-order RC model is:

[0014]

[0015] U=U OC -IR0-U1-U2

[0016] In the formula, Uoc represents the open circuit voltage of the battery; U represents the terminal voltage of the battery; I represents the current, which is negative for charging and positive for discharging; R0, R1, and R2 represent the ohmic internal resistance, electrochemical polarization internal resistance, and concentration polarization internal resistance of the battery, respectively; C1 and C2 represent the electrochemical polarization capacitance and concentration polarization capacitance, respectively; U1 and U2 represent the electrochemical polarization capacitance voltage and the concentration polarization capacitance voltage, respectively.

[0017] The VFFRLS identification model is:

[0018]

[0019] λ(k)=λ min +(1-λ min )·2 L(k)

[0020]

[0021] Where λ(k) is the variable forgetting factor; min is the minimum value of the forgetting factor; e(k) is the difference between the estimated y(k) and the actual y(k), also called the innovation; ρ is the sensitivity factor.

[0022] The beneficial effects of the present invention are:

[0023] 1. The present invention establishes an OCV-SOC-T relationship so that the OCV-SOC curve can change adaptively under different ambient temperatures, thereby reducing the error caused by ambient temperature changes in parameter identification and having temperature adaptive capability.

[0024] 2. The forgetting factor in the VFFRLS algorithm proposed in the present invention is adjusted online with the system new information, and a multi-new information matrix is ​​used to improve the accuracy, real-time performance and stability of recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of a battery parameter identification method based on the VFFRLS algorithm and taking into account the influence of ambient temperature of the present invention;

[0026] Figure 2 It is the second-order RC model diagram of the battery;

[0027] Figure 3(a) is the OCV-SOC curve at -10°C;

[0028] Figure 3(b) is the OCV-SOC curve at 50°C;

[0029] Figure 4 It is the three-dimensional relationship diagram of OCV-SOC-T. DETAILED DESCRIPTION

[0030] The present invention proposes a battery parameter identification method based on a VFFRLS algorithm and taking into account the influence of ambient temperature. The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0031] Figure 1 The flowchart of the battery parameter identification method based on the VFFRLS algorithm and considering the influence of ambient temperature of the present invention is as follows:

[0032] (1) Battery Modeling

[0033] When selecting a battery model, high accuracy and low computational complexity need to be considered at the same time. The Rint model and the first-order RC model are simple and direct, with low computational complexity, but the accuracy is not enough. Although the high-order RC has high accuracy and can well display the dynamic and static characteristics of the battery, as the order increases, the identification complexity of the system will also increase, and the computational complexity will increase. After comprehensive consideration, the second-order RC model is selected, such as Figure 2 .

[0034] In this model, U oc Represents the open circuit voltage of the battery; U represents the terminal voltage of the battery; I represents the current, which is negative for charging and positive for discharging; R0, R1, and R2 represent the ohmic internal resistance, electrochemical polarization internal resistance, and concentration polarization internal resistance of the battery, respectively; C1 and C2 represent the electrochemical polarization capacitor and concentration polarization capacitor, respectively; U1 and U2 represent the electrochemical polarization capacitor voltage and concentration polarization capacitor voltage, respectively. Its mathematical model can be described as:

[0035]

[0036] The battery model parameters are R0, R1, R2, C1, and C2. The purpose of identification is to obtain the accurate values ​​of these five parameters and to be able to update them in real time.

[0037] In order to obtain the above model parameters, the battery mathematical model is first converted into a discrete model. First, it is convenient for computer processing. Second, the current and voltage data are not continuous but obtained by interval sampling, so they must be converted into discrete form. First, through Laplace transformation, the expression of the second-order RC model in the frequency domain is obtained:

[0038]

[0039] The transfer function is:

[0040]

[0041] Then use bilinear transformation to discretize:

[0042] make but

[0043]

[0044] The difference equation is:

[0045] U OC (k)-U(k)=y(k)=a1y(k-1)+a2y(k-2)+a3I(k)+a4I(k-1)+a5I(k-2) (5)

[0046] That is the system equation, which is also the equation to be identified. Definition θ is the direct identification parameter. There is a relationship between θ and the battery model parameters. When θ is obtained, the model parameters are obtained. The relationship between θ and the model parameters is derived as follows:

[0047] make Perform an inverse bilinear transformation on equation 2 to obtain the transfer function:

[0048]

[0049] Let a=τ1τ2, b=τ1+τ2, c=R0+R1+R2, d=R0τ1+R0τ2+R1τ2+R2τ1

[0050] Comparing with formula (3), we get:

[0051]

[0052] make

[0053] have to

[0054]

[0055] R2=c-R0-R1

[0056] C1=τ1R1

[0057] C2=τ2R2

[0058] Therefore, the model parameters R0, R1, R2, C1, and C2 can be solved by θ = [a1 a2 a3 a4 a5].

[0059] In the equation to be identified U OC (k)-U(k)=y(k)=a1y(k-1)+a2y(k-2)+a3I(k)+a4I(k-1)+a5I(k-2), It is known data, so now we only need to know the open circuit voltage Uoc and then identify the direct parameter θ=[a1 a2 a3 a4 a5] to get the battery model parameters.

[0060] (2) Considering the influence of ambient temperature, establish the three-dimensional relationship between OCV-SOC-T

[0061] First, the OCV-SOC curves at multiple temperatures are obtained through battery experiments. The wider the temperature range, the better. This embodiment takes the battery data of the public data set A123 of the University of Maryland as an example. OCV-SOC tests are performed at 8 temperatures including -10°C, 0°C, 10°C, 20°C, 25°C, 30°C, 40°C, and 50°C. The OCV-SOC curves at -10°C and 50°C are shown in Figures 3(a) and 3(b).

[0062] As can be seen from the figure, the OCV-SOC relationship curves at different temperatures are similar but not the same; the OCV-SOC relationship of charging and discharging at the same temperature is also different. Due to the hysteresis effect, the lower the temperature, the more obvious the difference between charging and discharging. The average of charging and discharging is used as the OCV-SOC relationship at that temperature, and 8 relationship curves at eight temperatures are obtained. Then, with temperature and SOC as independent variables, the OCV-SOC-T three-dimensional surface is drawn, as shown in Figure 4 .

[0063] In this way, the eight temperature points are expanded into a wide temperature range, and the OCV-SOC relationship at other temperatures in this temperature range can be obtained by interpolation fitting. In actual battery operation, the most realistic OCV value can be adaptively obtained as the ambient temperature and SOC change. In this way, the influence of temperature on the OCV-SOC relationship is used as the starting point to eliminate the influence of temperature on parameter identification, so that parameter identification has the ability to adapt to temperature, thereby improving the accuracy and real-time performance of parameter identification.

[0064] (3) Recursive least squares algorithm with variable forgetting factor

[0065] The purpose of the variable forgetting factor recursive least squares algorithm (VFFRLS) is to identify the direct parameters θ = [a1 a2 a3 a4 a5] in the system equation. The existing least squares algorithm with fixed forgetting factor (FFRLS) is as follows:

[0066]

[0067] In the formula, λ is the forgetting factor, which is a constant value between 0 and 1. Its function is to alleviate the problem of old data superposition in the recursive operation process and prevent data saturation. In the parameter identification of lithium-ion power battery model, the optimal value of the forgetting factor is often affected by the following factors: the working state of the power battery, the sampling interval of the power battery data, the accuracy of the sensor, etc. Among these factors, the working state of the battery has the greatest influence on the selection of the optimal value of the forgetting factor. However, in practical applications, the working state of the battery is time-varying, so the forgetting factor should also change adaptively. Therefore, in order to improve the adaptability and accuracy of the algorithm, a VFFRLS algorithm based on adaptive adjustment of the forgetting factor based on model error is proposed.

[0068] Referring to JD Park in 1991, a method of variable forgetting factor was proposed:

[0069] λ(k)=λ min +(1-λ min )·2 L(k)

[0070] L(k)=-NINT(ρe(k) 2 )

[0071] Among them, λ(k) is the variable forgetting factor; λ min is the minimum value of the forgetting factor; e(k) is the difference between the estimated y(k) and the actual y(k), also called the innovation; ρ is the sensitivity factor; NINT is the rounding function.

[0072] Taking into account the actual situation of battery operation, based on the above method, it is necessary to:

[0073] a. Remove the rounding function. When identifying parameters, e(k) is usually very small, and ρe(k) will appear. 2 When λ(k) is less than 0.5, L(k) becomes 0 and λ(k) becomes 1, and there is no forgetting function.

[0074] b. Select the multi-innovation matrix E(k). If the current data is disturbed, a large error will be generated and the algorithm will be unstable. The multi-innovation theory (MILS) can make full use of the data at each moment, accelerate the convergence speed of the identification algorithm, and improve the identification accuracy.

[0075]

[0076] Thus, the VFFRLS identification model is obtained. It can adaptively change the size of the forgetting factor based on the system's new information, find the optimal value without manual trial and error, and improve the accuracy and real-time performance of the algorithm; and adopt a multi-news matrix to consider the influence of previous data when adjusting the size of the forgetting factor, which can improve the stability of the algorithm.

[0077] Combined with the OCV value obtained above, the battery model parameters can be identified. The present invention is based on two points: first, the OCV-SOC-T relationship is established to obtain an accurate OCV value; second, the VFFRLS algorithm is proposed to enable parameter identification to have temperature adaptive capability, while improving the accuracy, real-time performance and stability of identification.

Claims

1. A battery parameter identification method based on VFFRLS algorithm and considering the influence of ambient temperature, characterized in that: The following steps are involved: Establish a second-order RC model of the battery; Considering the influence of ambient temperature, the three-dimensional relationship of OCV-SOC-T is established to obtain OCV; A VFFRLS identification model is constructed to identify battery parameters.

2. The battery parameter identification method based on the VFFRLS algorithm and considering the influence of ambient temperature according to claim 1, characterized in that: The second-order RC model is: U=U OC -IR0-U1-U2 In the formula, Uoc represents the open circuit voltage of the battery; U represents the terminal voltage of the battery; I represents current, which is negative for charging and positive for discharging; R0, R1, and R2 represent the ohmic internal resistance, electrochemical polarization internal resistance, and concentration polarization internal resistance of the battery, respectively; C1 and C2 represent the electrochemical polarization capacitance and concentration polarization capacitance, respectively; U1 and U2 represent the electrochemical polarization capacitance voltage and concentration polarization capacitance voltage, respectively.

3. The battery parameter identification method based on the VFFRLS algorithm and considering the influence of ambient temperature according to claim 1, characterized in that: The VFFRLS identification model is: λ(k)=λ min +(1-l min )·2 L(k) Where λ(k) is the variable forgetting factor; min is the minimum value of the forgetting factor; e(k) is the difference between the estimated y(k) and the actual y(k), also called the new information; ρ is the sensitivity factor.

Images