WO2018188321A1

WO2018188321A1 - Method for enhancing battery state estimation robustness

Info

Publication number: WO2018188321A1
Application number: PCT/CN2017/108286
Authority: WO
Inventors: 冯代伟; 黄平江; 李永强
Original assignee: 绵阳世睿科技有限公司
Priority date: 2017-04-13
Filing date: 2017-10-30
Publication date: 2018-10-18
Also published as: CN107064816A

Abstract

A method for enhancing the battery state estimation robustness, comprising the following steps: step 1, establishing a battery mathematical model according to the battery characteristics; step 2, adjusting the battery mathematical model by adding a redundant state variable; and step 3, using a system state estimation method to perform online estimation on all of the state variables in the adjusted battery mathematical model. In the event that there are model deviation and system noise, by adding one or more system state variables with or without physical meaning to the battery model, and accordingly changing the battery mathematical model equation, the method can effectively improve the battery state estimation accuracy, accelerate the convergence rate of the estimation deviation, and enhance the parameter adaptability of the estimation algorithm. The method achieves a good estimation effect, and the effectiveness of the method is verified by testing.

Description

Method for enhancing battery state estimation robustness

Technical field

The invention belongs to the technical field of batteries, and in particular relates to a method for enhancing the robustness of battery state estimation.

Background technique

The battery management system collects the temperature, voltage, current and other parameters of the battery in real time, and based on the preset algorithm, avoids overcharging and overdischarging of the battery. The state of the battery is estimated to be at the heart of the battery management system. Its accurate estimation on the one hand improves the ease of use of the system at the user level, the driver can drive and maintain the car body within a reasonable time range; on the other hand, on the internal level of the battery, it can avoid possible permanent damage to the battery. Extends the service life of lithium-ion battery.

The battery state estimation is originally expected to be performed by measuring the voltage of the battery, because the state of charge of the battery has a monotonous correspondence with its open circuit voltage. However, due to the complex characteristics of battery polarization dynamics, the accuracy of estimating the state of the battery by voltage is very high. Low, it is difficult to meet the practical needs. For this reason, estimating the battery state by modeling the battery and combining the voltage and current signals of the battery is currently the main method for improving the accuracy of BMS estimation.

The method of battery state estimation can be divided into model-based and non-model-based methods from a model-based perspective. A typical non-model-based approach is the chrono-integration method. The types of models in the model-based approach include electrochemical models, neural network models, empirical formula models, and circuit models. Among them, the circuit model has received a lot of attention because it can reflect the polarization dynamic characteristics of the battery to a certain extent, and the related battery state estimation methods are basically concentrated. In the expansion of the Kalman filter method and its deformation method.

Given the high complexity of battery characteristics, it is difficult for any battery model to accurately characterize the battery because model-based battery state estimation methods have estimated bias due to model bias. When the model deviation is large due to the influence of temperature and charge and discharge depth, the accuracy of the battery state is significantly reduced. Moreover, the noise of the actual working conditions is also difficult to accurately model, and the battery state estimation will also have obvious adverse effects. In view of this, it is necessary to find a strong robust battery state estimation method that has higher estimation accuracy even in the presence of model deviation and noise.

Summary of the invention

In order to overcome the above disadvantages of the prior art, the present invention proposes a method for enhancing the robustness of battery state estimation, aiming to overcome the shortcomings of existing methods for model deviation and system noise, even in the presence of model deviation and system noise. In this case, the battery state can be estimated with high accuracy, and the robustness of the estimation can be improved.

The technical solution adopted by the present invention to solve the technical problem thereof is: a method for enhancing the robustness of battery state estimation, comprising the following steps:

Step 1: Establish a mathematical model of the battery according to the characteristics of the battery;

Step 2: Adjust the battery mathematical model by adding redundant state variables;

Step 3: Perform online estimation of all state variables in the adjusted battery mathematical model by using the system state estimation method.

Compared with the prior art, the positive effect of the present invention is that the present invention adds one or several system state variables having physical meaning or no physical meaning to the battery model. The mathematical model equation of the battery should be changed, and the accuracy of the battery state estimation, the convergence rate of the estimated deviation, and the parameter adaptability of the estimation algorithm can be effectively improved in the presence of model deviation and system noise. The present invention achieves a good estimation effect, and the effectiveness of the method of the present invention is verified by tests.

DRAWINGS

The invention will be illustrated by way of example and with reference to the accompanying drawings in which:

Figure 1 shows the battery equivalent circuit model.

Figure 2 shows the robustness of the robust enhancement to model bias.

Figure 3 shows the fault tolerance of robust enhancements to estimate state deviations.

Figure 4 shows the effect of robust enhancement on the parameter fit of the estimation method.

detailed description

A method for enhancing battery state estimation robustness includes the following steps:

Step 1. Establish a mathematical model of the battery:

Based on the battery characteristics, establish the appropriate model state variable x, input variable u, process noise w, measure noise and model parameters θ, and establish the battery system equation:

y=g(t,x,u,θ,v)

And then converted to discrete forms:

x _k+1 =f(x _k ,u _k ,θ)+w _k

y _k =g(x _k ,u _k ,θ)+v _k

Step 2: Add redundant state variables to the battery mathematical model and adjust the battery mathematical model accordingly:

Augment the state variable x:

The s in the equation is the augmented state variable, and the battery model is adjusted accordingly as follows:

Step 3: Perform online estimation of all state variables in the adjusted battery mathematical model by using the system state estimation method:

Where m _k is the measured value of the battery output at the sampling instant k, and L _k is the feedback gain of the estimation method used.

It should be noted:

The battery model of the present invention is not limited to any form, and may be a battery equivalent circuit model, an electrochemical model, or the like.

The redundant state variable added to the battery model of the present invention does not limit the number of redundant state variables, and may be one or more, and does not limit the physical meaning of the redundant variable, and may be a variable having a specific physical meaning, It can be a variable without specific physical meaning.

The system state estimation method adopted by the present invention does not limit any type of estimation method, and may be Kalman estimation, H _∞ estimation, sliding mode estimation, and the like.

The method of the invention is also not limited to what type of battery.

In order to explain in detail the technical content, algorithm features, and the objects and effects of the present invention, the system operation flow will be described in detail below in conjunction with the specific embodiments.

A method for enhancing battery state estimation robustness established by the present invention in conjunction with FIG. 1 includes the following steps:

Step 1. Model the battery using an equivalent circuit model and establish a mathematical model of the battery;

Figure 1 shows a typical battery equivalent circuit model, including the Open Circuit Voltage (OCV) controlled by the State of Charge (SoC), which is OCV (SoC); describes the battery equivalent The internal resistance R ₀ , which describes the second-order RC model of the electrochemical polarization and concentration polarization of the battery, namely R ₁ C ₁ and R ₂ C ₂ . Establish a mathematical model of the battery discrete system corresponding to the circuit model, the state of which is:

The state space equations and output equations of the mathematical model are:

x _k+1 =Ax _k +BI _k +Fw _k (2)

Where w _k is the process noise (ie current measurement noise), v _k is the voltage measurement noise, and y _k is the system output, which is the port voltage of the circuit shown in Figure 1 calculated using the system state. Combining the circuit model in Figure 1, the variation range of the SoC is taken as 0-100 (in percent), and the matrix in equation (2) is as follows:

Where Cap is the battery capacity, τ ₁ = R ₁ C ₁ , τ ₂ = R ₂ C ₂ , Δt is the system sampling period.

For a certain battery, the parameters and values required for the model are shown in Table 1:

Table 1 model resistance capacity parameters

Step 2, increase the redundant state variables of the mathematical model, and adjust the battery mathematical model accordingly, as follows:

The battery SoC is the integral of the current I, and the measured value of I contains the bias current I ^b which affects the accuracy of the SoC estimation, and is augmented as the state of the system, and is estimated online, the circuit model The system state variable is taken as:

among them

with

Corresponding to the voltage on the capacitor (or resistor) of the resistor-capacitor network. Taking II ^b as the actual current of the input battery, the value of which is positive indicates charging, and the state space equation and output equation of the discrete system can be written as:

Adjust the battery model accordingly. A ^a , B ^a and F ^{a in} equations (8) and (9) are:

Step 3. At each sampling instant, use the H _∞ observer to estimate the battery status, as follows:

First set the H _∞ observer parameters P ₀ , Q, W and V,

For the unit matrix, combined with the sampled value m _{k of the} current voltage of the battery, the following calculation is performed at each sampling instant k:

The test results are verified by the method, and the test results are shown in Figure 2-4. It can be seen from Fig. 2 that when there is obvious model deviation, the fault tolerance of the model deviation after enhancement is obviously better than that before the enhancement, indicating that the method of enhancing robustness has strong fault tolerance for model deviation. As can be seen from Fig. 3, when there is an initial battery state deviation, the speed at which the estimated value converges to the actual value is made faster after the enhancement. If the estimated effect of the estimator is sensitive to its parameter value, it means that the estimator is prone to debugging difficulties or work instability in practical applications, so the estimator of the engineerable estimator has a wider range of parameters, the better. It is the problem of parameter adaptability. It can be seen from Fig. 4 that, when there is no robust enhancement, whether the value of a certain parameter W is appropriate or not seriously affects the estimation accuracy of the battery state; and after robust enhancement, the battery state is in a wide range of values of W. The estimation accuracy is almost unaffected, indicating that the robust enhancement method will significantly increase the range of estimator parameters, and thus effectively increase the estimated parameter adaptability.

Claims

A method for enhancing robustness of battery state estimation, comprising the steps of:

Step 1: Establish a mathematical model of the battery according to the characteristics of the battery;

Step 2: Adjust the battery mathematical model by adding redundant state variables;

Step 3: Perform online estimation of all state variables in the adjusted battery mathematical model by using the system state estimation method.
A method for enhancing battery state estimation robustness according to claim 1, wherein said battery model comprises a battery equivalent circuit model and an electrochemical model.
A method for enhancing battery state estimation robustness according to claim 2, wherein the method for establishing a battery mathematical model according to battery characteristics in step 1 is:

(1) Establish a mathematical model of the battery discrete system corresponding to the equivalent circuit model. The state variables are:

Where: SoC represents the state of charge of the battery, and R 1 C 1 and R 2 C 2 represent a second-order RC model of electrochemical polarization and concentration polarization of the battery;

(2) Establish the state space equation and output equation of the mathematical model:

X k+1 =Ax k +BI k +Fw k

Where: w k is the process noise, v k is the voltage measurement noise, y k is the system output, R 0 is the battery equivalent internal resistance, and OCV (SoC) is the battery open circuit voltage OCV controlled by the battery state of charge SoC, and:

Where: Cap is the battery capacity, τ 1 = R 1 C 1 , τ 2 = R 2 C 2 , Δt is the system sampling period.
The method for enhancing battery state estimation robustness according to claim 3, wherein the method for adjusting the battery mathematical model by adding redundant state variables according to step 2 is:

Taking the bias current I b included in the measured value of current I that affects the accuracy of SoC estimation as a redundant state variable, the system state variable of the circuit model is taken as:

among them:
with
For the voltage corresponding to the capacitor or resistor of the resistor-capacitor network, II b is taken as the actual current of the input battery, and the value is positive for charging. The state space equation and output equation of the discrete system are:

Among them, A a , B a and F a are:
A method for enhancing robustness of battery state estimation according to claim 4, wherein said system state estimation method comprises a Kalman estimation method, an H ∞ estimation method, and a sliding mode estimation method.
The method for enhancing the robustness of battery state estimation according to claim 5, wherein the method for estimating the state variables of the adjusted battery mathematical model by using the system state estimation method in step 3 is : At each sampling instant, the H ∞ observer is used to estimate the battery status:

First, set the H ∞ observer parameters P 0 , Q, W and V, II is the unit matrix, and combined with the sample value m k of the current voltage of the battery, the following calculation is performed at each sampling time k:
The method for enhancing battery state estimation robustness according to claim 1, wherein the redundant state variable is one or more, and is a variable having a specific physical meaning or a variable having no specific physical meaning. .