CN108519557A - A power battery parameter identification method suitable for sparse data - Google Patents

Abstract

The invention provides a power battery parameter identification method suitable for sparse data. Based on the set member identification algorithm, when the statistical distribution characteristics of the actual system noise are difficult to determine, it is not necessary to assume the statistical distribution characteristics of the system noise, and only need to know Bounds on system noise, errors introduced by measurements via sensors, machine number rounding errors, and modeling errors can all be considered as bounded errors. At the same time, the algorithm has the ability to identify redundant data. When the sampling interval of the battery management system increases, the identification accuracy can also be guaranteed. Therefore, it is especially suitable for sparse data and has many beneficial effects such as significantly improving the reliability of the power battery management system. .

Description

A power battery parameter identification method suitable for sparse data

technical field

The invention relates to the field of power battery systems, in particular to a parameter identification technology for vehicle power batteries.

Background technique

As an important part of new energy vehicles, the power battery can estimate the remaining capacity of the battery and evaluate the aging degree of the battery through the battery management system, so as to keep the SOC and SOH of the battery within the normal working level and prevent the damage caused by the battery. Overcharge, overdischarge, and overcurrent cause permanent damage to the battery, thereby increasing the service life of the battery and reducing the cost of electric vehicles. Accurate identification and estimation of power battery parameters is the basis for ensuring accurate online estimation of the internal state.

Online parameter identification can usually directly use current, terminal voltage and other information to obtain the corresponding parameter value under the SOC in real time, which is suitable for complex and changeable real vehicle systems. In this case, a large error will be generated, and even divergence will occur, and with the increase of the sampling interval of the vehicle sensor, the accuracy of the identification result will decrease. The commonly used recursive least squares identification method based on forgetting factor, its random error must obey the normal distribution of zero mean and zero covariance, which is obviously difficult to be satisfied in practical application. Therefore, there is still a need in the art for a power battery parameter identification method based on sparse data that is not limited to noise distribution and can reduce the amount of required data to a certain extent.

Contents of the invention

In view of the above-mentioned technical problems in this field, the present invention provides a power battery parameter identification method suitable for sparse data, which specifically includes the following steps:

Step 1. Obtain and store the current and terminal voltage information of the power battery during operation online in real time;

Step 2, establishing a state space model for the power battery;

Step 3, performing online parameter identification and updating of the state space model based on the set member identification algorithm;

Step 4: Using the estimated value of OCV at the previous moment and the estimated value of the parameter vector at the current moment obtained by the online identification to obtain the estimated result of OCV at the current moment, and at the same time obtain the change of OCV by interpolating according to the corresponding relationship between OCV-SOC curve;

Step 5: comparing the OCV online identification result obtained in step 3 with the OCV change curve obtained by interpolation in step 4, and verifying the online identification result.

Further, in the step 2, establishing a state space model for the power battery specifically includes: based on the Thevenin equivalent circuit model and having the following input and output relationships:

Among them, U _t , i represent the terminal voltage and current respectively, k represents the time, Φ(k) represents the observation vector composed of the known input and output, θ(k) represents the unknown parameter vector to be identified, e( k) represents the model interference or noise sequence, U _oc (k) represents the OCV value at time k, and a ₁ , a ₂ , a ₃ are the components of the parameter vector to be identified.

Further, in the step 3, the online parameter identification and update of the state space model is performed based on the member identification algorithm, specifically including:

3.1. For the observation vector Φ(k) in the state space model, the unknown parameter vector θ(k), the covariance matrix P(k) in the identification algorithm, the boundary γ of the system noise and the radius vector σ of the ellipsoid ² (k) is initialized to obtain the initial value of the observation vector Φ(0), the initial value of the unknown parameter vector θ(0), the initial value of the covariance matrix P(0) and the radius vector σ ² (0) of the ellipsoid. Usually, according to the convergence of the algorithm and the parameter characteristics of the battery, Φ(0), θ(0) can be set to 0, and P(0) can be set to μ is a small positive number, usually μ=10 ^-4 , I represents the n-dimensional unit matrix, and the radius of the ellipsoid is greater than zero, so that σ ² (0)=1, according to the actual measurement equipment and construction Factors such as mode error determine the system noise threshold γ.

3.2. Use the following formula to recursively update and calculate the parameters at the current k moment, where k∈(1,2,…):

Among them, δ(k)=y(k)-Φ(k)θ(k), G(k)=Φ(k) ^T P(k-1)Φ(k);

The parameters α(k) and β(k) are solved by ellipsoid optimization criterion.

Furthermore, the selection of the above parameters α(k), β(k) is related to the accuracy of the final identification result, so the minimum volume ellipsoid criterion min det(P(k)), the minimum trace ellipsoid criterion min tr (P(k)) and optimization criteria such as minimizing the parameter σ ² (k) to calculate.

Further, using the OCV estimated value at the previous moment and the parameter vector estimated value at the current moment obtained by the online identification in the step 4 to obtain the OCV estimation result at the current moment specifically includes:

According to the online identification result of the unknown parameter vector θ(k)=[U _oc (k)-a ₁ U _oc (k-1)a ₁ a ₂ a ₃ ], the OCV estimated value at time k-1 is compared with the Describe the component a ₁ in the parameter vector to be identified, and obtain the OCV estimation result at time k:

U _oc (k)＝θ ₁ (k)+a ₁ (k)U _oc (k-1)

Wherein, θ ₁ (k) represents the first component of the unknown parameter vector θ(k).

Further, the interpolation according to the corresponding relationship of OCV-SOC described in the step 4 to obtain the change curve of OCV specifically includes:

4.1. Carry out the open circuit voltage test to obtain multiple OCV values under different states of charge;

4.2. Using the current i(k) collected in step 1, since the power battery has a relatively stable temperature and aging state, there is a one-to-one mapping relationship between SOC and OCV, so it can be obtained based on the ampere-hour integration method SOC value at the corresponding moment;

4.3. Based on the multiple OCV values and the SOC value, an OCV-SOC relationship curve is obtained through linear interpolation.

The above-mentioned method for online parameter identification of power batteries based on the set member identification algorithm does not need to make assumptions about the statistical distribution characteristics of the system noise when the statistical distribution characteristics of the actual system noise are difficult to determine. Errors introduced by measurement, machine number rounding errors, and modeling errors can all be seen as bounded errors. At the same time, the algorithm has the ability to identify redundant data. When the sampling interval of the battery management system increases, the identification accuracy can also be guaranteed. Therefore, it is especially suitable for sparse data and has many beneficial effects such as significantly improving the reliability of the power battery management system. .

Description of drawings

Fig. 1 is a schematic flow chart of the method provided according to the present invention

Figure 2 is a schematic diagram of the Thevenin equivalent circuit model

Figure 3 is a comparison of OCV identification results at 1-second sampling intervals under UDDS conditions

Figure 4 is a comparison of OCV identification results at 10-second sampling intervals under UDDS conditions

Figure 5 is a comparison of OCV identification results at 1-second sampling intervals under DST conditions

Figure 6 is a comparison of OCV identification results at 10-second sampling intervals under DST conditions

Detailed ways

A power battery parameter identification method suitable for sparse data provided by the present invention will be further explained in detail below in conjunction with the accompanying drawings.

A power battery parameter identification method suitable for sparse data provided by the present invention, as shown in Figure 1, specifically includes the following steps:

Step 1. Obtain and store the current and terminal voltage information of the power battery during operation online in real time;

Step 2, establishing a state space model for the power battery;

Step 3, performing online parameter identification and updating of the state space model based on the set member identification algorithm;

Step 4: Using the estimated value of OCV at the previous moment and the estimated value of the parameter vector at the current moment obtained by the online identification to obtain the estimated result of OCV at the current moment, and at the same time obtain the change of OCV by interpolating according to the corresponding relationship between OCV-SOC curve;

Step 5: comparing the online identification result obtained in step 3 with the OCV change curve obtained by interpolation in step 4, and verifying the online identification result.

In a preferred embodiment of the present application, the nickel-cobalt-manganese ternary battery NMC is selected as the research object, with a rated capacity of 25Ah, charge and discharge cut-off voltages of 4.2V and 2.5V, and a rated current of 7.5A. The test conditions are Dynamic Stress Condition (DST) and Urban Road Cycle Condition (UDDS). The terminal voltage value measured by the battery test system is used as a reference value and compared with the terminal voltage estimated value of the algorithm as an error to verify the stability of the algorithm in the case of sparse data, and the OCV result obtained by interpolation of the OCV-SOC curve It is used as a reference value to compare the error of the estimated value to verify the reliability of the algorithm in the case of sparse data. At the same time, the algorithm proposed by the present invention is compared with the recursive least squares identification algorithm based on forgetting factor in the case of sparse data to illustrate the applicability and stability of the algorithm in sparse data.

The establishment of a state space model for the power battery in the second step is based on the Thevenin equivalent circuit model shown in FIG. 2 .

Fig. 3 and Fig. 4 are under UDDS working condition respectively, the recursive least squares method based on forgetting factor and the OCV identification result comparison of the algorithm proposed in the present invention when the sampling interval is 1s and 10s, Fig. 5 and Fig. 6 have shown DST working condition Compared with the OCV identification results under the above conditions, it can be seen from the accompanying drawings that the parameter identification method proposed by the present invention is still suitable for sparse data when the sampling interval is large. Compared with the traditional recursive least squares method based on forgetting factor, the In the low SOC stage of the battery, the OCV identification results can still maintain convergence, strong stability, and no large shock peaks.

Under UDDS working conditions, the root mean square statistical characteristics of OCV identification results at different sampling intervals are shown in Table 1:

Table 1 Root mean square statistical characteristics of OCV identification results at different sampling intervals

It can be seen from Table 1 that the power battery parameter identification method for sparse data proposed by the present invention has high precision, and the root mean square estimation error can be controlled within 5%.

Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications and substitutions can be made to these embodiments without departing from the principle and spirit of the present invention. and modifications, the scope of the invention is defined by the appended claims and their equivalents.

Claims (6)

1. A power battery parameter identification method suitable for sparse data, characterized in that: specifically comprising the following steps: Step 1. Obtain and store the current and terminal voltage information of the power battery during operation online in real time; Step 2, establishing a state space model for the power battery; Step 3, performing online parameter identification and updating of the state space model based on the set member identification algorithm; Step 4: Using the estimated value of OCV at the previous moment and the estimated value of the parameter vector at the current moment obtained by the online identification to obtain the estimated result of OCV at the current moment, and at the same time obtain the change of OCV by interpolating according to the corresponding relationship between OCV-SOC curve; Step 5: comparing the OCV online identification result obtained in step 3 with the OCV change curve obtained by interpolation in step 4, and verifying the online identification result. 2. The method according to claim 1, characterized in that: in said step 2, a state-space model is established for said power battery, specifically comprising: based on Thevenin equivalent circuit model and having the following input and output relations: Among them, U _t , i represent the terminal voltage and current respectively, k represents the time, Φ(k) represents the observation vector composed of the known input and output, θ(k) represents the unknown parameter vector to be identified, e( k) represents the model interference or noise sequence, U _oc (k) represents the OCV value at time k, and a ₁ , a ₂ , a ₃ are the components of the parameter vector to be identified. 3. The method according to claim 2, characterized in that: in said step 3, online parameter identification and updating of said state-space model is performed based on a member identification algorithm, specifically comprising: 3.1. For the observation vector Φ(k) in the state space model, the unknown parameter vector θ(k), the covariance matrix P(k) in the identification algorithm, the boundary γ of the system noise and the radius vector σ of the ellipsoid ² (k) is initialized to obtain the initial value of the observation vector Φ(0), the initial value of the unknown parameter vector θ(0), the initial value of the covariance matrix P(0) and the radius vector σ ² (0) of the ellipsoid. 3.2. Use the following formula to recursively update and calculate the parameters at the current k moment, where k∈(1,2,…): Among them, δ(k)=y(k)-Φ(k)θ(k), G(k)=Φ(k) ^T P(k-1)Φ(k); The parameters α(k) and β(k) are solved by ellipsoid optimization criterion. 4. The method according to claim 3, characterized in that: said ellipsoid optimization criterion adopts minimum volume ellipsoid criterion min det (P (k)), minimum trace ellipsoid criterion min tr (P (k)) and Minimize the parameter σ ² (k) and so on. 5. The method as claimed in claim 2, characterized in that: the OCV estimated value at the previous moment obtained by the online identification and the parameter vector estimated value at the current moment described in the step 4 are used to obtain the current The OCV estimation results at each moment, including: According to the online identification result of the unknown parameter vector θ(k)=[U _oc (k)-a ₁ U _oc (k-1) a ₁ a ₂ a ₃ ], the OCV estimated value at time k-1 is compared with the Describe the component a ₁ in the parameter vector to be identified, and obtain the OCV estimation result at time k: U _oc (k)＝θ ₁ (k)+a ₁ (k)U _oc (k-1) Wherein, θ ₁ (k) represents the first component of the unknown parameter vector θ(k). 6. The method according to claim 1, characterized in that: the interpolation according to the corresponding relationship of OCV-SOC described in the step 4 obtains the variation curve of OCV, specifically comprising: 4.1. Carry out the open circuit voltage test to obtain multiple OCV values under different states of charge; 4.2. Using the current i(k) collected in step 1, since the power battery has a relatively stable temperature and aging state, there is a one-to-one mapping relationship between SOC and OCV, so it can be obtained based on the ampere-hour integration method SOC value at the corresponding moment; 4.3. Based on the multiple OCV values and the SOC value, an OCV-SOC relationship curve is obtained through linear interpolation.