Method suitable for energy storage system to estimate battery SOC

Abstract

The invention provides a battery SOC estimation method suitable for an energy storage system, and the method comprises the steps: constructing a second-order equivalent circuit of a battery and a corresponding equivalent state equation, and collecting the circuit parameters of the battery at different moments; based on the circuit parameters, performing parameter optimization on an EKF algorithm by using a multi-target flower pollination algorithm to obtain an optimal Q matrix and an optimal R matrix; and based on the optimal Q matrix, the optimal R matrix and the equivalent state equation, estimating the SOC of the battery by using an EKF algorithm to obtain the SOC estimation performance of the battery. The method solves the problem that in the prior art, when an extended Kalman filtering algorithm is used for estimating the state of charge (SOC) of the battery in an energy storage system, due to the uncertainty and randomness of system noise, accurate measurement is difficult to conduct on the SOC, and the accuracy of the estimation result of the SOC of the battery is not high.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle power batteries, and in particular to a method for estimating battery SOC applicable to an energy storage system. Background Art

[0002] Energy storage systems play a vital role in integrated energy systems. They can not only improve energy efficiency, but also enhance system flexibility and reliability. By storing excess renewable energy, energy storage systems can provide necessary energy during peak demand, help reduce dependence on fossil fuels, and promote energy transformation and the achievement of carbon reduction goals. Lithium-ion batteries have received widespread attention in the energy storage industry because they can store a large amount of energy in the same volume.

[0003] However, in the battery energy storage system, there are many groups of batteries connected in series and parallel (up to 1500V or more), a large number of batteries, a harsh electromagnetic environment, serious interference, huge data, and complex control. It puts forward extremely high requirements on wiring design, anti-interference EMC design, data processing capabilities, response speed, etc. Therefore, when the extended Kalman filter algorithm is used in the energy storage system to estimate the battery state of charge (SOC), it is difficult to measure it accurately due to the uncertainty and randomness of the system noise. It can only be estimated manually, resulting in low accuracy of the battery SOC estimation result. The battery energy storage system cannot fully utilize the energy capacity of the battery, resulting in performance degradation. At the same time, it is also easy for the battery energy storage system to misjudge the actual state of charge of the battery, increasing the risk of battery overheating, overcharging or over-discharging, which may cause safety problems such as battery overheating and explosion. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for estimating battery SOC in an energy storage system, which solves the problem in the prior art that when an extended Kalman filter algorithm is used to estimate the battery state of charge (SOC) in an energy storage system, it is difficult to accurately measure it due to the uncertainty and randomness of system noise, resulting in low accuracy of the estimation result of the battery SOC.

[0005] According to an embodiment of the present invention, a method for estimating battery SOC in an energy storage system includes:

[0006] Construct the second-order equivalent circuit of the battery and the corresponding equivalent state equation, and collect the circuit parameters of the battery at different times;

[0007] Based on the circuit parameters, the multi-objective flower pollination algorithm is used to optimize the parameters of the EKF algorithm to obtain the optimal Q matrix and the optimal R matrix;

[0008] Based on the optimal Q matrix, optimal R matrix and equivalent state equation, the battery SOC is estimated using the EKF algorithm to obtain the battery SOC estimation performance.

[0009] Preferably, the second-order equivalent circuit of the battery includes an ideal voltage source Uoc, an internal resistance R0, a first RC branch and a second RC branch connected in series in sequence, the first RC branch includes a first resistor R1 and a first capacitor C1 connected in parallel, and the second RC branch includes a second resistor R2 and a second capacitor C2 connected in parallel.

[0010] Preferably, the circuit parameters include a load current, a first transient voltage and a second transient voltage, the first transient voltage is a voltage of a first RC branch, and the second transient voltage is a voltage of a second RC branch.

[0011] Preferably, the method of using the multi-objective flower pollination algorithm to optimize the parameters of the EKF algorithm to obtain the optimal parameters includes:

[0012] S1: Setting the maximum number of iterations and the population, wherein the population includes multiple objects, each object includes a Q matrix and an R matrix;

[0013] S2: Calculate the individual fitness corresponding to all objects according to the Q matrix and the R matrix and determine the current optimal solution of the population and the search mechanism of each object;

[0014] S3: Update the Q matrix and R matrix of all objects according to the search mechanism;

[0015] S4: Repeat steps S3-S4 until the number of repetitions reaches the maximum number of iterations, and perform weighted fusion on the Q matrix and R matrix of all objects to obtain the optimal Q matrix and the optimal R matrix.

[0016] Preferably, the search mechanism includes global search and local search, setting a search threshold for the population and assigning a random probability P to each object in the population, wherein the value range of P is 0 to 1. If the P value is less than the search threshold, the corresponding object uses global search; if the P value is greater than or equal to the search threshold, the corresponding object uses local search.

[0017] Preferably, the global search formula is as follows:

[0018]

[0019] The formula for local search is as follows:

[0020]

[0021] in, and are the solutions of the tth iteration and t+1th iteration respectively, g* is the optimal solution in the current population, λ is a constant 3 / 2, Γ(λ) is the standard gamma function, and is randomly selected from the population. The solution of ,∈ is a random number that obeys the (0, 1) uniform distribution.

[0022] Preferably, the calculation method of individual fitness is as follows:

[0023]

[0024] Where N is the number of samples, RSOC is the SOC obtained by the ampere-hour integration method as a reference, and SOC EKF Indicates the SOC value obtained from the EKF algorithm; Voltage mea It is the actual measured voltage. EKF Represents the voltage obtained from the EKF algorithm.

[0025] Preferably, the steps of weighted fusion are as follows:

[0026] f=w i (SOC MAE )+w i (Voltage MAE )

[0027] Among them, w i is a non-negative weight coefficient.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The actual battery circuit is virtualized into a second-order equivalent circuit to reduce the computational complexity. The circuit parameters of the battery collected at different times are then used as part of the input of the multi-objective flower pollination algorithm to optimize the parameters of the EKF algorithm and find the optimal Q matrix and R matrix that best suits the battery circuit. The EKF algorithm is used to estimate the battery SOC under the premise of the optimal Q matrix and R matrix to reduce the impact of noise on the EKF algorithm and improve the accuracy of the EKF algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 FIG. 4 is a flow chart of estimating battery SOC according to an embodiment of the present invention.

[0031] Figure 2 4 is a second-order equivalent circuit diagram of an embodiment of the present invention.

[0032] Figure 3 It is a comparison diagram of the actual SOC value of the battery under constant current discharge conditions and the EKF estimated battery SOC curve after parameter optimization in an embodiment of the present invention.

[0033] Figure 4 This is an error diagram of the EKF estimated battery SOC after parameter optimization under constant current discharge conditions in an embodiment of the present invention.

[0034] Figure 5 This is an EKF estimated battery SOC curve diagram for different SOC initial values ​​after parameter optimization under the dynamic pressure test condition of an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0036] like Figure 1 As shown, an embodiment of the present invention proposes a method for estimating battery SOC in an energy storage system, including:

[0037] Construct the second-order equivalent circuit of the battery and the corresponding equivalent state equation, and collect the circuit parameters of the battery at different times;

[0038] The second-order equivalent circuit of the battery is as follows Figure 2 As shown, the second-order equivalent circuit of the battery includes an ideal voltage source Uoc, an internal resistance R0, a first RC branch and a second RC branch connected in series in sequence. The first RC branch includes a first resistor R1 and a first capacitor C1 connected in parallel, and the second RC branch includes a second resistor R2 and a second capacitor C2 connected in parallel.

[0039] Among them, UOC represents an ideal voltage source with a nonlinear relationship with SOC, UT is the terminal voltage of the battery, and I is the load current; R0 represents the ohmic internal resistance of the battery, and the two RC branches represent the polarization characteristics of the battery, one of which describes the electrochemical polarization and the other describes the concentration polarization. The transient voltages of the two RC branches are represented by U1 and U2 respectively. The direction of the discharge current is defined as positive and the charging current is negative. According to Kirchhoff's law, the characteristics of the battery can be expressed by the following formula:

[0040]

[0041] According to the Coulomb counting method, the SOC at time t is expressed as SOC t , can be calculated by the following formula:

[0042]

[0043] Among them, SOC t0 is the initial value of SOC, η is the Coulomb efficiency, which is 1 in the present invention, and Q n is the verified capacity of the battery. By combining the above two equations, the discrete state space equivalent state equation of the battery can be obtained:

[0044]

[0045] Among them, T s is the sampling time, the subscripts k and k-1 represent the time step, τ1 and τ2 are time constants, where τ1 = R1*C1, τ2 = R2*C2, and R1, R2, C1, and C2 are identified using the fitting toolbox in Matlab software.

[0046] Based on the circuit parameters, the multi-objective flower pollination algorithm is used to optimize the parameters of the EKF algorithm to obtain the optimal Q matrix and the optimal R matrix;

[0047] EKF is a state estimation technique based on Kalman filter, which is an algorithm for estimating system state. It combines system model and sensor measurements to provide state estimation in an optimal way. Compared with the standard Kalman filter, the extended Kalman filter allows nonlinear system models and nonlinear observation models. This means that it can be applied to a wider range of system and sensor models, making it more flexible in practical applications. The state space model of a nonlinear discrete system is expressed as follows:

[0048]

[0049] Among them, x k+1 represents the state of the system at time k+1, u k represents the control variable, y k represents the measurement vector, ω k and k denote process noise and measurement noise respectively, ω k and k are uncorrelated and follow a Gaussian distribution. Let Q k is the variance matrix of the system noise array; R k is the variance matrix of the measurement noise array.

[0050] Select x k =[SOC k U 1,k U 2,k ] T , U T As observable variables, the linear system state equation and observation equation of EKF are expressed as follows (7):

[0051]

[0052] Combined with the discrete state space equivalent state equation of the battery,

[0053]

[0054] Ck =[-1 -1 0]

[0055] D=-R0

[0056] The EKF algorithm process is as follows:

[0057] initialization:

[0058]

[0059] Status prediction:

[0060]

[0061] Error covariance:

[0062]

[0063] Calculate the Kalman gain:

[0064]

[0065] Status Update:

[0066]

[0067] Error covariance update:

[0068] P k / k =(IK k C k ) k / k-1

[0069] Since the measurement of the process noise covariance matrix Q and the measurement error covariance matrix R in the EKF algorithm is difficult and has large errors, the final SOC is easily estimated inaccurately, so it is necessary to tune the parameters of the Q matrix and the R matrix.

[0070] Step 1: Initialize algorithm related parameters.

[0071] Set the current number of iterations to t=0; the maximum number of iterations is tmax=200. Multiple iterations can ensure the convergence of the algorithm while ensuring the accuracy of the algorithm.

[0072] The population size is set to 20, that is, there are 20 objects in the population, each object contains a Q matrix and an R matrix, both of which are the process noise covariance matrix and the measurement error covariance matrix in the EKF algorithm, and the initial value of each element in the Q matrix and the R matrix in each object is a random value.

[0073] Step 2: Calculate the individual fitness of all objects. Take the Q matrix and R matrix of each object as input, load the voltage and current data collected in the energy storage system, and calculate its SOC through the above EKF. Finally, calculate the individual fitness. The fitness calculation formula is as follows:

[0074]

[0075] Where N is the number of samples, RSOC is the SOC obtained by the ampere-hour integration method as a reference, and SOCEKF represents the SOC value obtained from the EKF algorithm; Voltage mea It is the actual measured voltage. EKF Represents the voltage obtained from the EKF algorithm.

[0076] Therefore, the goal of the multi-objective flower pollination algorithm is SOC MAE and Voltage MAE , and confirm the current optimal matrix Q and matrix R based on the current individual fitness.

[0077] Step 3: Determine the search mechanism.

[0078] The search mechanism includes global search and local search. A search threshold is set for the population and a random probability P is assigned to each object in the population. The value range of P is 0 to 1. If the P value is less than the search threshold, the corresponding object uses global search. If the P value is greater than or equal to the search threshold, the corresponding object uses local search.

[0079] Global search:

[0080] The cross-pollination of organisms can be considered as a global pollination process, and pollinators cross-pollinate in a manner that obeys Levy flight, the formula of which is:

[0081]

[0082] in, and are the solutions of the tth iteration and t+1th iteration respectively, g* is the optimal solution in the current population, L is a parameter related to pollination intensity, which is essentially a step length parameter based on the Levy flight method, and its calculation formula is:

[0083]

[0084] Where λ is a constant 3 / 2, Γ(λ) is the standard gamma function, and the final output is That is the SOC corresponding to the object MAE and Voltage MAE .

[0085] Local Search:

[0086] Abiotic self-pollination is considered a localized pollination process. Localized pollination and the regularity of flowers can be expressed by the following equation:

[0087]

[0088] in and is randomly selected from the population. The solution of ,∈ is a random number that obeys the (0, 1) uniform distribution.

[0089] Step 4: Repeat steps 1-3, using the About SOC MAE and Voltage MAE Perform update iterations, and the repeated stopping condition is: whether the number of iterations reaches the maximum number of iterations tmax. If it reaches it, the iteration of the multi-objective flower pollination algorithm ends, the algorithm converges, and the individual fitness of all objects is combined using the following weighted formula to obtain the optimal solution for the population:

[0090]

[0091] Where m is the number of targets, w i (i=1,2,...,m) is a non-negative weight coefficient, which is artificially designed based on historical experience, and the final optimal Q matrix and optimal R matrix are determined based on f.

[0092] Based on the optimal Q matrix, optimal R matrix and equivalent state equation, the EKF algorithm is used to estimate the battery SOC performance under various working conditions, such as Figure 3-Figure 5 .

[0093] In order to quantify the performance of the algorithm, the following four evaluation indicators are used to evaluate the SOC estimation performance of EKF under two working conditions before and after optimization. The four evaluation indicators include: maximum error (Error(Max)), mean square error (MSE), mean absolute error (MAE), and root mean square error (RMSE).

[0094] According to the four evaluation indicators, the corresponding Max, MSE, MAE, and RMSE comparison table of the optimized EKF algorithm is established, and the following Table 1 is obtained:

[0095] Table 1: Evaluation index comparison table

[0096]

[0097] from Figure 3-Figure 5 As can be seen from the above table, after using the optimized optimal Q matrix and optimal R matrix (Voltage(V)), compared with before use (SOC), the four evaluation indicators have all decreased significantly, the error has been significantly reduced, and the accuracy has been significantly improved.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A battery SOC performance estimation method, characterized in that: include: Construct the second-order equivalent circuit of the battery and the corresponding equivalent state equation, and collect the circuit parameters of the battery at different times; Based on the circuit parameters, the multi-objective flower pollination algorithm is used to optimize the parameters of the EKF algorithm to obtain the optimal Q matrix and the optimal R matrix; Based on the optimal Q matrix, optimal R matrix and equivalent state equation, the battery SOC is estimated using the EKF algorithm to obtain the battery SOC estimation performance.

2. A method for estimating battery SOC in an energy storage system according to claim 1, characterized in that: The second-order equivalent circuit of the battery includes an ideal voltage source Uoc, an internal resistance R0, a first RC branch and a second RC branch connected in series in sequence, the first RC branch includes a first resistor R1 and a first capacitor C1 connected in parallel, and the second RC branch includes a second resistor R2 and a second capacitor C2 connected in parallel.

3. A method for estimating battery SOC in an energy storage system according to claim 1, characterized in that: The circuit parameters include a load current, a first transient voltage, and a second transient voltage, wherein the first transient voltage is a voltage of a first RC branch, and the second transient voltage is a voltage of a second RC branch.

4. A method for estimating battery SOC in an energy storage system according to claim 1, characterized in that: The method of using the multi-objective flower pollination algorithm to optimize the parameters of the EKF algorithm and obtain the optimal parameters includes: S1: Setting the maximum number of iterations and the population, wherein the population includes multiple objects, each object includes a Q matrix and an R matrix; S2: Calculate the individual fitness corresponding to all objects according to the Q matrix and the R matrix and determine the current optimal solution of the population and the search mechanism of each object; S3: Update the Q matrix and R matrix of all objects according to the search mechanism; S4: Repeat steps S3-S4 until the number of repetitions reaches the maximum number of iterations, and perform weighted fusion on the Q matrix and R matrix of all objects to obtain the optimal Q matrix and the optimal R matrix.

5. A method for estimating battery SOC in an energy storage system according to claim 4, characterized in that: The search mechanism includes global search and local search, setting a search threshold for the population and assigning a random probability P to each object in the population, wherein the value range of P is 0 to 1. If the P value is less than the search threshold, the corresponding object uses global search; if the P value is greater than or equal to the search threshold, the corresponding object uses local search.

6. A method for estimating battery SOC in an energy storage system according to claim 5, characterized in that: The formula for the global search is as follows: The formula for local search is as follows: in, and are the solutions of the tth iteration and t+1th iteration respectively, g* is the optimal solution in the current population, λ is a constant 3 / 2, Γ(λ) is the standard gamma function, and is randomly selected from the population. The solution of ,∈ is a random number that obeys the (0, 1) uniform distribution.

7. A method for estimating battery SOC in an energy storage system according to claim 4, characterized in that: The calculation method of individual fitness is as follows: Where N is the number of samples, RSOC is the SOC obtained by the ampere-hour integration method as a reference, and SOC EKF Indicates the SOC value obtained from the EKF algorithm; Voltage mea It is the actual measured voltage. EKF Represents the voltage obtained from the EKF algorithm.

8. A method for estimating battery SOC in an energy storage system according to claim 4, characterized in that: The steps of weighted fusion are as follows: f=w i (SOC MAE )+w i (Voltage MAE ) Among them, w i is a non-negative weight coefficient.

Images