CN106443496A - Battery charge state estimation method with improved noise estimator - Google Patents

Abstract

The invention discloses a method for estimating the state of charge of a battery with an improved noise estimator. The method is as follows: establish a battery space state equation according to the battery equivalent circuit model, and use the improved noise estimator to obtain an estimated noise value at time k Then use this noise estimate as the noise statistical information of the adaptive unscented Kalman filter method, combine the battery space state equation with the adaptive unscented Kalman filter method to estimate the battery state of charge, and obtain the intermediate state quantity at time k g( _xi,k ), y _k ), and as the input of the improved noise estimator at time k+1, the estimated value of the battery state of charge is obtained by cyclic recursion. Compared with the standard unscented Kalman filter method, the method for estimating the state of charge of a battery with an improved noise estimator is higher in estimation accuracy and better in robustness.

Description

A battery state of charge estimation method with an improved noise estimator

technical field

The invention belongs to the technical field of design and control of a MW-level battery energy storage system in a smart grid, and relates to a method for estimating a state of charge of a battery.

Background technique

With the rapid development of wind power and photovoltaic power generation, battery energy storage systems have also been greatly developed. The battery is the main carrier of energy storage and release in the battery energy storage system. Accurately determining the power of the battery directly determines whether the battery energy storage system can be effectively operated and controlled. However, the charging and discharging process of the battery is a complex electrochemical reaction process, and the power contained in it is difficult to obtain directly. Usually, the battery state of charge (State of Charge, SOC) is used to characterize the battery power.

At present, the commonly used SOC estimation methods mainly include: ampere-hour method, open circuit voltage method, impedance method, neural network method, fuzzy logic method, extended Kalman filter (EKF) and standard unscented Kalman filter (UKF) and so on. The shortcomings of the current method: (1) The accuracy of the ampere-hour method is not high due to the shortcomings of accumulation of errors and obtaining a clear initial value of SOC; (2) The open-circuit voltage method is not suitable for online estimation and is time-consuming; (3) Impedance method has the disadvantages of complex algorithm and inconvenient actual operation; (4) Neural network and fuzzy logic method need to obtain a large amount of experimental data, which is difficult to obtain during the actual operation of the battery, and its actual accuracy is not high; (5) ) Extended Kalman filter method (EKF) has the disadvantages of calculating the Jacobian matrix and ignoring high-order terms, and its estimation accuracy is not high. (6) The standard unscented Kalman filter (UKF) has the advantages of not needing to calculate the Jacobian matrix, and the calculation amount is small, but in the actual application process, the statistical information in the standard unscented Kalman filter (UKF) (such as system noise, measurement noise, etc.) are not constant, or even unknown or unclear, resulting in low estimation accuracy and poor robustness. For example, a power battery charge estimation method disclosed in document (CN103675706A). In the process of realizing the present invention, the inventors found that the existing methods at least have problems such as low precision, poor real-time performance, large amount of calculation, and poor robustness.

Contents of the invention

The purpose of the present invention is to address the above problems and propose a battery state of charge estimation method with an improved noise estimator to achieve the advantages of high precision, good real-time performance, less time-consuming, and strong robustness. The standard unscented Kalman filter method (UKF) without noise estimator is compared, and the accuracy and high robustness of a battery state of charge estimation method with improved noise estimator of the present invention are further explained.

The object of the invention is to realize through the following technical solutions:

The present invention provides a battery state of charge estimation method with an improved noise estimator, said method is as follows: the first step is to determine the equivalent circuit model (1) of the known battery; the second step is to establish the battery space state equation ( 2); the 3rd step, utilize the improved noise estimator (3) to obtain the noise estimation value (4) of k moment; The 4th step, replace the adaptive unscented Kalman filtering method with the noise estimation value (4) of k moment The noise statistical information of AUKF(5) uses the battery state of charge SOC in the battery space state equation (2) and the terminal voltage of two RC parallel circuits as the state variables of the adaptive unscented Kalman filter method AUKF(5), The input state space equation and the output voltage state space equation of the battery space state equation (2) are respectively used as the nonlinear state equation f(·) and measurement equation g(·) of the adaptive unscented Kalman filter method AUKF(5), The fifth step is to use the adaptive unscented Kalman filter method AUKF (5) to obtain the intermediate state quantity (6) at time k, and output the estimated value of SOC _{b, k} of the battery system state of charge at time k at the same time; the sixth step is to The intermediate state quantity (6) is used as the input quantity of the improved noise estimator (3) at time k+1, and the estimated value of the SOC of the battery state of charge at different times is obtained by recursion in this way.

The battery equivalent circuit model (1) is a second-order equivalent circuit model, and the main circuit of the model is composed of two RC parallel circuits, a controlled voltage source U ₀ (SOC), and the internal resistance R of the battery. According to the circuit structure of the battery model and its charging and discharging characteristics, the mathematical expression of the equivalent circuit model is:

In the formula, a ₀ ~ a ₅ , b ₀ ~ b ₅ , c ₀ ~ c ₂ , d ₀ ~ d ₂ , e ₀ ~ e ₂ , f ₀ ~ f ₂ are all model coefficients, which can be obtained from the battery experimental data through least squares It is obtained by multiplication fitting; Q ₀ is the rated power of the battery; SOC ₀ is the initial value of SOC, generally a constant between 0 and 1; R _s and R _l respectively represent the resistance of two RC parallel circuits in the battery model; C _s , C ₁ respectively represents the capacitance of two RC parallel circuits in the battery model; U ₀ and R represent the open circuit voltage and internal resistance of the battery respectively; U and I represent the terminal voltage and current of the battery respectively.

The establishment of the battery space state equation (2) is as follows: 1), the terminal voltage U _s (t) and U ₁ (t) of the battery SOC and two RCs are used as the system state variable x _k , and U and I are respectively used as The system measurement variable y _k and the system input variable, according to the equivalent circuit model, the battery space state equation is established as

In the formula, U _s and U ₁ are the terminal voltages of two RC parallel circuits, τ ₁ and τ ₂ are time constants, ω _k is the system noise, Δt is the sampling period, and k is a natural number greater than 1; 2), according to Keele Hough's voltage law, combined with the battery equivalent circuit model, the battery output measurement equation can be obtained as: [U _k ]=U _0,k -R _k I _b,k -U _s,k -U _l,k +υ _k =g _k (x _k )+υ _k =y _k , where, υ _k is the system measurement noise, and k is a natural number greater than 1.

The main steps of the adaptive unscented Kalman filter method AUKF (5) are: 1) initializing x mean value E ( ) variance and noise statistics: 2) Calculate sampling points x _i,k and corresponding weight ω _i :

In the formula, λ=α ² (n+h)-n, n is the dimension of the state variable; ω ^m and ω ^c represent the weight of the variance and the mean value respectively, and the operator is the Cholesky decomposition of a symmetric matrix, and α, β, and h are all constants; 3) Time update of state estimation and mean square error: state estimation time is updated as In the formula, q _k is the noise mean value of the state equation; the mean square error time is updated as Q _k is the noise variance of the state equation; the system output time is updated as In the formula, g _k-1 (·) is the measurement equation, r _k is the noise mean value of the measurement equation; 4) Calculate the gain matrix: In the formula, P _y,k is the autocovariance, P _xy,k is the autocovariance, R _k is the noise variance of the state equation; 5) The state estimation and the measurement update of the mean square error: the state estimation measurement update is The mean squared error measure is updated to

Described improved noise estimator (3) is:

In the formula, diag( ) is a diagonal matrix, y _k is the actual measured value of the system output, Respectively represent the estimated value of state equation noise mean value, state equation noise variance estimate value, measurement equation noise mean value estimate value, and measurement equation noise variance estimate value at time k, that is, the noise estimate value at time k (4).

The intermediate state quantity (6) of described k moment has: P _k , x _i,k , f( _xi,k ), g( _xi,k ), y _k , where y _k is the actual measured value output by the system.

The battery type is one of lithium-ion battery or lead-acid battery.

Compared with the standard unscented Kalman filter method (UKF) without noise estimator for battery SOC estimation, the present invention has the following beneficial technical effects: one, the whole discharge process, the band improved noise estimation adopted by the present invention The battery state of charge estimation method (AUKF) with a noise estimator has higher accuracy than the standard unscented Kalman filter (UKF) without a noise estimator; the second is that the AUKF with an improved noise estimator is more accurate than the UKF Faster convergence and better robustness

Description of drawings

Fig. 1 is a kind of flow chart of battery state of charge estimation method with improved noise estimator;

Figure 2 is a battery equivalent circuit model diagram containing two RC parallel circuits;

Figure 3 shows the SOC variation under the battery constant current discharge condition when SOC ₀ =0.8;

Fig. 4 shows the variation of the SOC error under the constant current discharge condition of the battery when SOC ₀ =0.8.

detailed description

The present invention will be further described in detail below in conjunction with specific examples, which are for explanation of the present invention rather than limitation.

According to an embodiment of the present invention, as shown in FIG. 1, FIG. 2, FIG. 3 and FIG. 4, a battery state of charge estimation method with an improved noise estimator is provided. The flow chart of the embodiment is shown in FIG. 1, It mainly includes the following steps:

1. Determine the equivalent circuit model of the known battery

The battery equivalent circuit model (1) is a second-order equivalent circuit model. The main circuit of the model is composed of two RC parallel circuits, a controlled voltage source U ₀ (SOC) and the internal resistance R of the battery, as shown in Figure 2. The specific battery equivalent circuit model is as follows:

In the formula, the values of a ₀ ~ a ₅ are -0.915, 40.867, 3.632, 0.537, 0.499, 0.522 respectively, the values of b ₀ ~ b ₅ are 0.1463, 30.27, 0.1037, 0.0584, 0.1747, 0.1288, c ₀ ~ c The values of ₂ are 0.1063, 62.49, 0.0437 respectively, the values of d ₀ ~ d ₂ are -200, 138, 300 respectively, the values of e ₀ ~ e ₂ are 0.0712, 61.4, 0.0288 respectively, and the values of f ₀ ~ f ₂ are respectively 3083, 180, 5088.

2. Establish the battery space state equation

1) Take the battery SOC and the terminal voltages U _s (t) and U _l (t) of the two RCs as the system state variable x _k , take U and I as the system measurement variable y _k and the system input variable respectively, according to The efficient circuit model establishes the battery space state equation as

In the formula, U _s and U _l are the terminal voltages of two RC parallel circuits, τ ₁ and τ ₂ are time constants, ω _k is the system noise, Δt is the sampling period, and k is a natural number greater than 1.

2), according to Kirchhoff's voltage law, combined with the battery equivalent circuit model, the battery output measurement equation can be obtained as: [U _k ]=U _0,k -R _k I _k -U _s,k -U _{l, k} +υ _k , where k is a natural number greater than 1.

3. Use the improved noise estimator (3) to obtain the noise estimate at time k (4)

Using the improved noise estimator combined with the intermediate state quantity at the previous moment to obtain the noise estimate at time k (4), that is

4. The noise estimate at time k (4) respectively as the statistical information values (q _k , Q _k , r _k , R _k ) of the adaptive unscented Kalman filter method AUKF(5), namely

Taking the battery state of charge SOC in the battery space state equation (2) and the terminal voltage of two RC parallel circuits as the state variables of the adaptive unscented Kalman filter method AUKF (5), that is

The input state space equation and the output voltage state space equation of the battery system space state equation (2) are respectively used as the nonlinear state equation f( ) and measurement equation g( ) of the adaptive unscented Kalman filter method AUKF (5) ,Right now

g _k (x _k )=U _0,k -R _k I _b,k -U _bs,k -U _b1,k .

5. Use the adaptive unscented Kalman filter method AUKF (5) to estimate the battery SOC, and obtain the intermediate state quantity (6) at time k, namely P _k , x _i,k , f(x _i,k ), g(x _i,k ), y _k .

1) Initialize the state variable x mean E() and noise information: 2) Calculate the sampling points x _i,k and the corresponding weight ω:

In the formula, λ=α ² (n+h)-n, n=3, the value of α is 1, the value of β is 2, and the value of h is 0; 3) Time update of state estimation and mean square error: state Estimated time updated to The mean square error time is updated as The system output time is updated as 4) Calculate the gain matrix: 5) The measurement update of state estimation and mean square error: the state estimation measurement is updated as The mean squared error measure is updated to

At the same time, the state variables are estimated The output of the first element of is the estimated value of SOC _k of the battery system at time k.

6. The intermediate state quantity (6) at time k+1 is used as the input quantity of the improved noise estimator (3) at the next time moment, and the estimated value of the SOC of the battery state of charge is obtained by recursively recursively.

System simulation results and effect comparison

According to a battery state of charge estimation method with an improved noise estimator of the present invention, the SOC is estimated for the lithium-ion battery, and the standard UKF without the noise estimator is used to estimate the SOC of the battery, and the simulation results and experimental data are compared. To verify that a battery state of charge estimation method with an improved noise estimator in the present invention has the advantages of fast convergence speed, strong robustness and high precision. The main constant current condition of the simulation test is that the battery supplies power to the outside in a constant current mode (0.8A). Figure 3 shows the change of SOC under the constant current discharge condition of the battery when SOC ₀ =0.8. It can be seen from Figure 3 that the AUKF with improved noise estimator (marked as AUKF in the figure, the same below) is the same as the one without noise estimator When two standard UKF algorithms are used for SOC estimation, AUKF can track the experimental data faster than UKF, and the convergence time of the two is 100s and 200s respectively, which proves that the method proposed in the present invention has faster convergence speed and better performance. robustness. Fig. 4 shows the variation of the SOC error under the constant current discharge condition of the battery when SOC ₀ =0.8. It can be seen from Fig. 4 that during the whole discharge process, the error of the AUKF with the improved noise estimator is obviously smaller than that of the standard UKF without the noise estimator, which proves that the accuracy of the method proposed by the present invention is higher.

Claims (4)

1. The present invention discloses a method for estimating the state of charge of a battery with an improved noise estimator, the method comprising the following steps: The first step is to determine the equivalent circuit model of the known battery (1); the second step is to establish the battery space state equation (2); the third step is to use the improved noise estimator (3) to obtain the noise estimate at time k ( 4); the fourth step, replace the noise statistical information of the adaptive unscented Kalman filter method AUKF (5) with the noise estimate value (4) of the k moment, and use the battery state of charge SOC in the battery space state equation (2) , The terminal voltage of two RC parallel circuits is used as the state variable of the adaptive unscented Kalman filter method AUKF (5), and the input state space equation and the output voltage state space equation of the battery space state equation (2) are respectively used as the adaptive unscented Kalman filter method AUKF (5). Non-linear state equation f(·) and measurement equation g(·) of the traced Kalman filter method AUKF(5), the fifth step is to use the adaptive unscented Kalman filter method AUKF(5) to obtain the intermediate state at time k Quantity (6), while outputting the SOC _{b of the battery system at time k,} the estimated value of k; the sixth step, using the intermediate state quantity (6) as the input quantity of the improved noise estimator (3) at time k+1, In this way, the estimated value of the SOC of the battery state of charge at different times is obtained by recursively recursively. 2. a kind of battery state of charge estimation method with improved noise estimator according to claim 1, is characterized in that described improved noise estimator (3) is: In the formula, diag( ) is a diagonal matrix, Respectively represent the estimated value of state equation noise mean value, state equation noise variance estimate value, measurement equation noise mean value estimate value, and measurement equation noise variance estimate value at time k, that is, the noise estimate value at time k (4). 3. a kind of battery state of charge estimation method with improved noise estimator according to claim 1, it is characterized in that the intermediate state quantity (6) of described k moment has: P _k , x _i,k , f( _xi,k ), g( _xi,k ), y _k , where y _k is the actual measured value output by the system. 4. A kind of battery state of charge estimation method with improved noise estimator according to claim 1, is characterized in that described battery type is a kind of in lithium-ion battery or lead-acid battery.