Sodium ion energy storage battery parameter online identification method based on maximum correlation entropy

Abstract

The invention belongs to the technical field of parameter identification, and discloses a sodium ion energy storage battery parameter online identification method based on maximum correlation entropy, which comprises the following steps: establishing a first-order Thevenin model of a sodium ion energy storage battery; based on the first-order Thevenin model, obtaining a discrete time domain model of the sodium ion energy storage battery; and constructing a cost function by adopting a maximum correlation entropy criterion based on a Gaussian kernel function so as to estimate parameters of the discrete time domain model, thereby obtaining sodium ion energy storage battery parameters and completing sodium ion energy storage battery parameter identification. According to the identification method, the pulse noise under the energy storage working condition can be inhibited, accurate and robust parameter estimation is realized, and meanwhile, the convergence speed of a parameter estimator is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of parameter identification, and more specifically, relates to an online identification method for sodium ion energy storage battery parameters based on maximum correlation entropy. Background Art

[0002] In recent years, with the increasing prominence of resource and environmental issues, it is urgent to accelerate energy transformation. The access of a high proportion of new energy has brought severe challenges to the power-energy balance of the power system at multiple time scales. The large-scale application of energy storage technology is an effective means to solve the problem of power supply and demand balance. Among them, electrochemical energy storage technology has the advantages of flexible configuration, rapid response, and efficient energy conversion, and has now taken the lead in entering the scale. At present, due to the advantages of high energy density, long cycle life, low self-discharge rate, and no memory effect, lithium-ion batteries have an absolute dominant position in the fields of power batteries and energy storage batteries. However, safety performance and production cost have always been two key issues restricting the large-scale application of lithium-ion batteries in the field of energy storage. Sodium-ion energy storage batteries are highly safe, low-cost, and have a long life, and are expected to become an important supplement to the lithium-ion battery technology route. Therefore, online identification of the parameters of sodium-ion energy storage batteries is of great research significance.

[0003] At present, the research on modeling and parameter identification of sodium-ion energy storage batteries has just started, and in essence, it still follows the research methods in the field of lithium-ion batteries. The performance verification of battery models is mostly carried out under constant current conditions, standard dynamic conditions and actual vehicle driving conditions. However, changes in battery systems and application scenarios indicate different optimization requirements for models and algorithms. The operating conditions of energy storage batteries are completely different from those of power batteries, and the current size and direction will frequently change suddenly, which puts higher requirements on the convergence and robustness of parameter identification algorithms.

[0004] At present, the online parameter identification methods for power battery working conditions include the least square method, Kalman filter method and other algorithms, all of which are based on the minimum mean square error criterion. The common feature of these filters is that they have strict requirements on the statistical distribution of noise. They can achieve optimal filtering performance under Gaussian noise conditions, but serious performance degradation will occur under pulse noise conditions. Under the working conditions of energy storage batteries, due to the frequent mutations in the magnitude and direction of the current, the statistical distribution of the noise is difficult to meet the Gaussian noise distribution. In addition, the current spikes in the energy storage frequency modulation working conditions will form pulse noise in the voltage prediction error, resulting in the accuracy and robustness of the estimation algorithm based on the minimum mean square error criterion cannot be guaranteed. That is, directly using the estimation algorithm based on the minimum mean square error criterion under the working conditions of sodium ion energy storage batteries will have two negative effects: first, the pulse noise generated under the working conditions of sodium ion energy storage batteries will disturb the converged model parameters, causing the parameter estimation values ​​to oscillate for a short time and a large amplitude, or even significantly deviate from the normal value; second, the pulse noise will cause the abnormal growth of the covariance matrix in the state estimator, weakening the stability of the recursive filter or observer. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a method for online identification of sodium ion energy storage battery parameters based on maximum correlated entropy, which aims to improve the accuracy and identification stability of online identification of sodium ion energy storage battery parameters.

[0006] To achieve the above object, according to a first aspect of the present invention, a method for online identification of sodium ion energy storage battery parameters based on maximum correlation entropy is provided, comprising:

[0007] S1. Establish the first-order Thevenin model of sodium-ion energy storage battery: including the voltage source U in series OC , ohmic internal resistance R0 and R1C1 circuit consisting of resistor C1 and capacitor C1 in parallel; where U OC represents the open circuit voltage of the sodium ion energy storage battery, and R0 represents the resistance of the electrodes, electrolyte, and diaphragm to Na + The R1C1 circuit represents the polarization effect of the sodium ion energy storage battery;

[0008] S3. Based on the first-order Thevenin model, a discrete time domain model of the sodium ion energy storage battery is obtained:

[0009] U t,k -U OC,k =a0I k +a1I k-1 +b1(U t,k-1 -U OC,k-1 ), k ≥ 2

[0010] Among them, U t,k, U OC,k ,I k Corresponding to the terminal voltage U of the sodium ion energy storage battery t , open circuit voltage U OC , the kth sampling value of the current I; and the parameters a0, a1, b1 satisfy:

[0011]

[0012] Where, T is the preset sampling period, τ1=R1C1;

[0013] S3. A cost function is constructed by adopting the maximum correlation entropy criterion based on the Gaussian kernel function to estimate the parameters a0, a1, and b1 of the discrete time domain model, thereby obtaining the parameters R0, R1, and C1 of the sodium ion energy storage battery, and completing the parameter identification of the sodium ion energy storage battery.

[0014] Furthermore, in S3, the cost function constructed using the maximum correlation entropy criterion based on the Gaussian kernel function is:

[0015]

[0016] In the formula, represents the correlation entropy based on the Gaussian kernel function, σ is the kernel length of the Gaussian kernel function, and λ represents the forgetting factor; Represents θ k The estimated value of θ k ,φ k ,y k The corresponding parameter matrix, input matrix and output matrix of the discrete time domain model are: θ k =[a 0,k a 1,k b 1,k ] T ,φ k =[I k I k-1 U t,k-1 -U OC,k-1 ] T ,y k =U t,k -U OC,k ; Among them, a 0,k 、a 1,k , b 1,k Correspondingly represents the estimated values ​​of parameters a0, a1, and b1 at the kth sampling time.

[0017] Furthermore, the cost function is used to estimate the parameters a0, a1, and b1 of the discrete time domain model, including:

[0018] The correlation entropy The maximum is the optimization target, and the following recursive formula is used to solve θ in real time k Estimates

[0019]

[0020] Among them, P k represents the covariance matrix of the kth sampling; K k is the gain matrix,

[0021] Further, in S2, based on the first-order Thevenin model, a discrete time domain model of the sodium ion energy storage battery is obtained, including:

[0022] Based on the first-order Thevenin model, the continuous time domain model of sodium-ion energy storage batteries is determined;

[0023] Performing Laplace transformation on the continuous time domain model to obtain a corresponding continuous transfer function H(s); and performing bilinear transformation on the continuous transfer function H(s) to obtain a corresponding discrete transfer function H(z);

[0024] The discrete time domain model is determined using the discrete transfer function H(z).

[0025] Furthermore, the input matrix φ k Middle,U OC,k-1 Determined by:

[0026] Obtain the OCV-SoC function relationship of the sodium ion energy storage battery; where OCV represents the open circuit voltage U OC , SoC represents the state of charge;

[0027] The open circuit voltage value U is obtained by the OCV-SoC function relationship OC,k-1 .

[0028] Furthermore, the OCV-SoC function relationship of the sodium-ion energy storage battery is obtained, including:

[0029] Conduct low current charge and discharge experiments on sodium ion energy storage batteries to obtain the relationship between the open circuit voltage and the state of charge of the sodium ion energy storage battery during discharge, and the relationship between the open circuit voltage and the state of charge of the sodium ion energy storage battery during charge;

[0030] The open circuit voltage during charging and the open circuit voltage during discharging under the same state of charge are averaged to obtain a relationship curve between the average open circuit voltage and the state of charge;

[0031] The relationship curve between the average open circuit voltage and the state of charge is fitted with a polynomial to obtain the OCV-SoC function relationship.

[0032] Furthermore, a polynomial fitting is performed on the relationship curve between the average open circuit voltage and the state of charge, including:

[0033] Between 10% and 95% of the SoC, with 35% SoC as the boundary, a piecewise second-order polynomial fitting is performed on the relationship curve between the average open circuit voltage and the state of charge.

[0034] According to a second aspect of the present invention, there is provided an electronic device comprising a computer-readable storage medium and a processor;

[0035] The computer-readable storage medium is used to store executable instructions;

[0036] The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the online identification method of sodium ion energy storage battery parameters described in any one of the first aspects.

[0037] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method for online identification of parameters of a sodium-ion energy storage battery as described in any one of the first aspects is implemented.

[0038] According to a fourth aspect of the present invention, there is provided a computer program product, comprising a computer program, which, when executed on a computer, enables the computer to execute the method for online identification of parameters of a sodium ion energy storage battery as described in any one of the first aspects.

[0039] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

[0040] (1) The method for online identification of sodium ion energy storage battery parameters based on maximum correlation entropy of the present invention, for sodium ion energy storage battery, the present invention first establishes a first-order Thevenin model of sodium ion energy storage battery, and converts the open circuit voltage U OC , the electrode, electrolyte, and diaphragm have a great influence on Na + The blocking effect of the sodium ion energy storage battery and the polarization effect of the sodium ion energy storage battery are characterized as a first-order equivalent circuit, and the discrete time domain model of the sodium ion energy storage battery is determined based on the first-order equivalent circuit. The cost function is constructed using the maximum correlation entropy criterion based on the Gaussian kernel function to perform online identification of the sodium ion energy storage battery parameters. When the impulse noise comes, that is, the input parameter [I k I k-1 U t,k-1 -U OC,k-1 ] TAfter the pulse noise caused by the current spike contained in is acted upon by the Gaussian kernel function (exponential function), the corresponding cost function increment tends to 0, so it will not have a substantial impact on parameter estimation. In this way, the cost function constructed by the maximum correlation entropy criterion based on the Gaussian kernel function is equivalent to having a suppressive effect on the pulse noise, thereby achieving accurate and robust parameter estimation, and improving the accuracy and identification stability of online identification of sodium-ion energy storage battery parameters.

[0041] (2) Furthermore, in the present invention, under the working condition of the sodium ion energy storage battery, the output measurement value y is measured by using the correlation entropy based on the Gaussian kernel function. k and output estimates The cost function of sodium ion energy storage battery parameter identification is defined based on the maximum correlation entropy criterion of the Gaussian kernel function. That is, the model error at the current moment is evaluated under the correlation entropy induced distance based on the Gaussian kernel function. In this way, when the impulse noise comes, Rapidly increases, after the exponential function, the corresponding cost function tends to 0. The contribution of can be ignored, thus achieving the suppression of impulse noise.

[0042] (3) Further, the present invention uses the correlation entropy The maximum is the optimization target, the corresponding recursive formula is designed, the formula is used for iterative solution, and the newly obtained data is used to correct the original estimated value, so that the real-time online update of the parameters can be realized, and there is no need to store a large amount of data, and the computer memory occupied is small, and the real-time high-precision robust parameter estimation of the sodium ion energy storage battery model can be realized under the energy storage frequency modulation condition. At the same time, the data selection mechanism established by the maximum related entropy criterion is spontaneous, and there is no need to introduce any new parameters or steps into the recursive formula. The algorithm complexity is the same as the traditional least squares algorithm, that is, the method of the present invention improves the accuracy and identification stability of the online identification of the sodium ion energy storage battery parameters, and does not affect the identification speed compared with the traditional method.

[0043] (4) As a preference, considering that the batteries in the battery energy storage system are used as frequency control backup, the SoC operating range is limited to between 10% and 95% of the SoC, and within this range, the OCV-SoC curve of the sodium-ion energy storage battery is close to linear, which is easy to fit; a piecewise second-order polynomial fitting is adopted, and the order of the polynomial fitting is lower, and the amount of calculation is smaller.

[0044] In summary, the online parameter identification method of the sodium ion energy storage battery of the present invention is based on the constructed first-order Thevenin model of the sodium ion energy storage battery and its discrete time domain model, and uses the maximum correlation entropy criterion to construct a cost function, so that under the battery energy storage condition, when a current spike occurs, the identification algorithm can suppress pulse noise, achieve accurate and robust parameter estimation, and ensure the convergence speed of the parameter estimator. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Schematic diagram of a method for online identification of parameters of a sodium ion energy storage battery based on maximum correlation entropy in an embodiment of the present invention.

[0046] Figure 2 The figure is a schematic diagram of a specific process of identifying unknown parameters of a sodium ion battery using the maximum correlation entropy algorithm in an embodiment of the present invention.

[0047] Figure 3 It is the OCV-SoC curve fitting result in the embodiment of the present invention.

[0048] Figure 4 Graphs showing changes in load current, battery voltage and SoC during the test process of an embodiment of the present invention; (a)-(h) in the graph correspond to curves showing changes in current over time and voltage over time on the 82nd, 318th, 76th and 336th days randomly selected from the annual reference current operating conditions of energy storage frequency regulation; (i) shows curves showing changes in SoC over time on the 82nd, 318th, 76th and 336th days.

[0049] Figure 5 Graphs showing the identification results of the ohmic internal resistance R0 obtained by using different identification methods in an embodiment of the present invention; (a)-(d) respectively represent the identification results of the ohmic internal resistance R0 on the 82nd day, the 318th day, the 76th day and the 336th day. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0051] Example 1

[0052] like Figure 1 As shown, an embodiment of the present invention provides a method for online identification of sodium ion energy storage battery parameters based on maximum correlation entropy, which mainly includes:

[0053] S1. Establish a first-order Thevenin model of sodium-ion energy storage battery. The model includes a voltage source U OC , an ohmic internal resistance R0 and an R1C1 circuit consisting of a resistor R1 and a capacitor C1 in parallel; wherein the voltage source U OC , an ohmic internal resistor R0 and R1C1 circuit connected in series; voltage source U OC It represents the open circuit voltage of the sodium ion energy storage battery. The ohmic internal resistance R0 represents the resistance of the electrodes, electrolyte, and diaphragm in the sodium ion energy storage battery to Na + The R1C1 network represents the polarization effect of sodium-ion energy storage batteries;

[0054] S2. Based on the first-order Thevenin model, the discrete time domain model of the sodium-ion energy storage battery is obtained:

[0055] U t,k -U OC,k =a0I k +a1I k-1 +b1(U t,k-1 -U OC,k-1 ), k ≥ 2

[0056] Among them, U t,k Represents the terminal voltage U of the sodium ion energy storage battery t The value of the k-th sampling, U OC,k Indicates the open circuit voltage U of the sodium ion energy storage battery OC The value of the k-th sampling, I k represents the value of the kth sampling of the current I of the sodium ion energy storage battery. Correspondingly, U t,k-1 , U OC,k-1 ,I k-1 Respectively represent the t , U OC , I performs the k-1th sampling value; parameters a0, a1, b1 satisfy:

[0057]

[0058] Wherein, T is the preset sampling period, which indicates the time interval between two adjacent samplings.

[0059] As a preferred implementation method, based on the first-order Thevenin model, a discrete time domain model of the sodium ion energy storage battery is obtained, including:

[0060] S21. Based on the first-order Thevenin model, the continuous time domain model of the sodium-ion energy storage battery is determined as:

[0061]

[0062] Among them, U1(t) is the polarization voltage of the sodium ion energy storage battery at time t, Ut (t) is the terminal voltage of the sodium ion energy storage battery at time t, U OC (t) is the open circuit voltage of the sodium ion energy storage battery at time t, and I(t) is the current of the sodium ion energy storage battery at time t.

[0063] S22. Perform Laplace transform on the continuous time domain model of the sodium ion energy storage battery to obtain the corresponding continuous transfer function H(s):

[0064]

[0065] Where U t (s), U OC (s), I(s) correspond to U t (t), U OC (t), U OC Laplace transform of (t); s is a complex variable, τ1 is a time constant, τ1 = R1C1.

[0066] S23, bilinearly transform the continuous transfer function H(s) to obtain the corresponding discrete transfer function H(z):

[0067]

[0068]

[0069] Where T is the preset sampling period, which indicates the time interval between two adjacent samplings; a0, a1, and b1 are the parameters to be identified.

[0070] S23. Determine the discrete time domain model of the sodium ion energy storage battery based on the discrete transfer function H(z):

[0071] U t,k -U OC,k =a0I k +a1I k-1 +b1(U t,k-1 -U OC,k-1 ), k ≥ 2

[0072] Among them, the parameter mapping relationship between the discrete time domain model and the continuous time domain model is:

[0073]

[0074] S3. The maximum correlation entropy criterion based on the Gaussian kernel function is used to construct a cost function to estimate the parameters a0, a1, and b1, and then the parameters R0, R1, and C1 of the sodium-ion energy storage battery are obtained to complete the parameter identification of the sodium-ion energy storage battery.

[0075] As a preferred implementation, Figure 2As shown, in the embodiment of the present invention, the maximum correlation entropy criterion based on the Gaussian kernel function is used to construct a cost function to estimate the parameters a0, a1, and b1, including:

[0076] S31. Write the discrete time domain model of the sodium-ion energy storage battery in matrix form:

[0077] System parameter matrix: θ k =[a 0,k a 1,k b 1,k ] T ;

[0078] System input matrix: φ k =[I k I k-1 U t,k-1 -U OC,k-1 ] T ;

[0079] System output matrix: y k =U t,k -U OC,k ;

[0080] System news:

[0081] Among them, a 0,k 、a 1,k 、b 1,k Correspondingly, it represents the estimated values ​​of a0, a1, and b1 at the kth iteration, that is, the parameters to be identified corresponding to the kth sampling. The kth iteration corresponds to the kth sampling. θ k-1 estimate.

[0082] Among them, the system input matrix φ k Middle,U OC,k-1 Determined by:

[0083] Obtain the OCV-SoC function relationship of the sodium ion energy storage battery; where OCV represents the open circuit voltage U OC , SoC represents the state of charge;

[0084] The open circuit voltage value U is obtained by the OCV-SoC function relationship OC,k-1 .

[0085] As a preferred implementation, a low current method is used to establish an OCV-SoC function relationship, including:

[0086] Carry out a small current charge and discharge experiment. The specific process is as follows:

[0087] Fully charge the sodium-ion energy storage battery using the standard charging method and leave it for a period of time tm After a period of 30 minutes (for example, 30 minutes), the battery is discharged at a small rate (for example, 0.05C) to a preset lower cutoff voltage V1 (for example, 2V), and the relationship between the open circuit voltage OCV and the state of charge SoC during discharge of the sodium ion energy storage battery is obtained; wherein the open circuit voltage OCV is the above-mentioned U OC ;

[0088] The sodium ion energy storage battery is left to stand for a period of time t m Then, the voltage is charged from the preset lower cut-off voltage V1 to the preset upper cut-off voltage V2 (for example, 3.95V) at the same rate to obtain the relationship between the open circuit voltage OCV and the state of charge SoC when the sodium ion energy storage battery is charged;

[0089] The average open circuit voltage OCV during charging and the open circuit voltage OCV during discharging under the same state of charge SoC are taken to obtain an average OCV-SoC curve;

[0090] The average OCV-SoC curve was fitted with a polynomial to obtain the desired OCV-SoC functional relationship.

[0091] Preferably, a polynomial fitting is performed on the average OCV-SoC curve to obtain the desired OCV-SoC functional relationship, including:

[0092] Since the batteries in the battery energy storage system are used as frequency control backup, the SoC operating range is limited to 10% to 95% of SoC, and the OCV-SoC curve of the sodium-ion energy storage battery is close to linear in this range, which is easy to fit. Therefore, when fitting the average OCV-SoC curve, the fitting range is 10%≤SoC≤95%;

[0093] Taking 35% SoC as the boundary, the average OCV-SoC curve is fitted with a piecewise second-order polynomial to obtain the desired OCV-SoC curve. In the embodiment of the present invention, it is experimentally measured that taking 35% SoC as the boundary, the average OCV-SoC curve is fitted with a piecewise second-order polynomial, and the order of the polynomial fitting is low, and the amount of calculation is small.

[0094] In the embodiment of the present invention, a piecewise second-order polynomial fitting is performed on the average OCV-SoC curve with 35% SoC as the boundary, that is:

[0095]

[0096] Among them, k1, k2, ..., k6 are fitting coefficients, OCV represents the open circuit voltage U OC (t).

[0097] The fitted OCV-SOC curve is as follows: Figure 3 shown.

[0098] S32. The cost function constructed using the maximum correlation entropy criterion based on the Gaussian kernel function is:

[0099]

[0100] In the formula, represents the correlation entropy based on the Gaussian kernel function, Represents θ k is an estimate of , σ is the kernel length of the Gaussian kernel function, and λ represents the forgetting factor.

[0101] S33, with correlation entropy The maximum optimization goal is:

[0102]

[0103] S34, using the following recursive formula, solve the corresponding parameter value that maximizes the correlation entropy, and obtain the estimated values ​​of a0, a1, and b1 at each sampling time

[0104]

[0105] Among them, P k represents the covariance matrix of the kth iteration, P k-1 It represents the covariance matrix of the k-1th iteration, and its initial value is assigned by experience, usually taking the unit matrix; the kth iteration corresponds to the kth sampling, that is, one iteration is performed for each sampling. When performing online parameter identification, the parameters can be updated in real time according to the iterative formula; The initial value of can be any value in theory, usually 0, or a value can be calculated by the existing conventional offline identification algorithm as the initial value, which can speed up the convergence of the model; K k is an intermediate variable, representing the gain matrix. ω represents the parameter related to the core length σ,

[0106] According to the above recursive formula, the estimated values ​​of a0, a1, and b1 for each sampling can be obtained: , and put it into the following formula to obtain the unknown parameters R0, R1 and C1 of the first-order equivalent circuit of the sodium ion energy storage battery:

[0107]

[0108] In the embodiment of the present invention, the correlation entropy of two random variables can be used to measure the closeness between the two random variables X and Y. It is defined as:

[0109]

[0110] In the formula, x and y are the values ​​of random variables X and Y; E[·] is the expected value; F X,Y (x,y) is the joint distribution function of X and Y; G σ (·) is a kernel function with a kernel length of σ. If the kernel function is a Gaussian kernel function, then:

[0111]

[0112] Based on this, under the working condition of sodium ion energy storage battery, the output measurement value y is measured by the correlation entropy based on Gaussian kernel function. k and output estimates The cost function of sodium ion energy storage battery parameter identification is defined based on the maximum correlation entropy criterion of the Gaussian kernel function. for:

[0113]

[0114] In this way, the embodiment of the present invention no longer uses the Euclidean distance to measure the model error at a certain moment. Instead of evaluating the model error at the current moment under the correlation entropy induced distance based on the Gaussian kernel function, the model error is evaluated at the current moment under the correlation entropy induced distance based on the Gaussian kernel function. When the impulse noise comes, increases sharply. After the exponential function is applied, the corresponding cost function increment tends to 0. The contribution of is negligible, so it will not have any substantial impact on parameter estimation, which is equivalent to automatically filtering out the noise data. That is, after constructing the cost function using the maximum correlation entropy criterion, the parameter estimator implicitly contains the ability of data selection. This feature is very important for real-time parameter estimation of battery models under energy storage frequency regulation conditions. The impulse noise suppression capability it provides is an important prerequisite for achieving accurate and robust parameter estimation.

[0115] The kernel length σ is a key parameter that determines the performance of the algorithm. When σ tends to infinity, the maximum correlation entropy algorithm will degenerate into a least squares algorithm and will no longer be able to filter noise. On the contrary, when σ tends to zero, all noise will be filtered indiscriminately and the real-time update of parameters will stop. By controlling the value of the kernel length σ, reducing the kernel length can bring stronger noise suppression capabilities, but it will also slow down the convergence speed of the parameter estimator. Therefore, the kernel length selection needs to make a trade-off between the real-time performance and stability of the algorithm according to actual needs.

[0116] The recognition effect of the method of the present invention is further illustrated below with specific examples.

[0117] From the annual reference current conditions of energy storage frequency regulation established in the existing technology, the current curves of 4 days were randomly selected and scaled to the single battery scale for testing. Figure 4The changes of load current, battery voltage and SoC during the test are shown. The recursive least squares algorithm (RLS) with forgetting factor and the recursive maximum correntropy algorithm (RMC) designed by the present invention are used to identify the model parameters online. Both methods are initialized according to Table 1. The core lengths are set to σ1 = 0.5912mV, σ2 = 0.9008mV, and σ3 = 1.5877mV, respectively. The identification results of the ohmic internal resistance R0 are shown in Figure 1. Figure 5 As shown in the figure, we can observe the excellent anti-interference performance of RMC. When using RLS for online parameter identification, the estimated value of R0 often jitters due to abnormal noise, and even deviates greatly from the normal range in severe cases. RMC can completely eliminate this jitter phenomenon, and the curve can remain stable.

[0118] Based on the identification results, voltage prediction is performed. The prediction results with mean absolute error (MAE) and root mean square error (RMSE) as indicators are shown in Table 2. The voltage prediction accuracy of the battery model when online parameter identification is performed using RLS and RMC is evaluated. Compared with RLS, under low noise suppression intensity (σ=σ3), RMC not only significantly improves the robustness of the parameter estimation algorithm, but also achieves further improvement in model accuracy. Therefore, for parameter estimation of battery models under energy storage frequency regulation conditions, RMC is a more reasonable and appropriate choice.

[0119] Table 1 Parameter identification initial value setting

[0120]

[0121] Table 2 Voltage prediction results

[0122]

[0123] The present invention takes into account that directly using the estimation algorithm based on the minimum mean square error criterion under the working condition of the sodium ion energy storage battery will cause serious performance degradation under impulse noise conditions. This is because the minimum mean square error criterion is based on the cost function designed by the second-order statistics, and it will inevitably explode at the moment when the current spike arrives, and forgetting is gradual, so this instantaneous change will have a lasting impact for a period of time afterwards, which will force the parameter estimator to re-enter the dynamic convergence process. Directly applying it to the identification of sodium ion energy storage battery parameters will affect the accuracy and identification stability of the online identification of sodium ion energy storage battery parameters.

[0124] Based on this, for sodium ion energy storage batteries, the present invention first establishes a first-order Thevenin model of sodium ion energy storage batteries, and converts the open circuit voltage U OC , the electrode, electrolyte, and diaphragm have a great influence on Na + The blocking effect of the sodium ion energy storage battery and the polarization effect of the sodium ion energy storage battery are characterized as a first-order equivalent circuit, and the discrete time domain model of the sodium ion energy storage battery is determined based on the first-order equivalent circuit. The cost function is constructed using the maximum correlation entropy criterion based on the Gaussian kernel function to perform online identification of the sodium ion energy storage battery parameters. When the impulse noise comes, that is, the input parameter [I k I k-1 U t,k-1 -U OC,k-1 ] T After the pulse noise caused by the current spike contained in is acted upon by the Gaussian kernel function (exponential function), the corresponding cost function increment tends to 0, so it will not have a substantial impact on parameter estimation. In this way, the cost function constructed by the maximum correlation entropy criterion based on the Gaussian kernel function is equivalent to having a suppressive effect on the pulse noise, thereby achieving accurate and robust parameter estimation, and improving the accuracy and identification stability of online identification of sodium-ion energy storage battery parameters.

[0125] The present invention uses the maximum correlation entropy criterion to identify the parameters of sodium-ion energy storage batteries. After the cost function is constructed using the maximum correlation entropy criterion, the parameter estimator contains the ability of data selection and has the ability to suppress impulse noise, which not only significantly improves the robustness of the parameter estimation algorithm, but also further improves the accuracy of the model. Real-time high-precision robust parameter estimation of the sodium-ion energy storage battery model can be achieved under energy storage frequency modulation conditions. At the same time, the data selection mechanism established by the maximum correlation entropy criterion is spontaneous, and there is no need to introduce any new parameters or steps into the recursive algorithm. The algorithm complexity is the same as the traditional least squares algorithm, that is, the method of the present invention improves the accuracy and identification stability of the online identification of sodium-ion energy storage battery parameters, while compared with the traditional method, it does not affect the identification speed.

[0126] Example 2

[0127] An embodiment of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the above-mentioned method for online identification of sodium ion energy storage battery parameters based on maximum correlation entropy is implemented.

[0128] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The memory may be used to store computer programs and / or modules, and the processor may run or execute computer programs and / or modules stored in the memory, and call data stored in the memory, to realize various functions of the electronic device.

[0129] The relevant technical solutions are the same as above and will not be described again here.

[0130] Example 3

[0131] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned method for online identification of sodium ion energy storage battery parameters based on maximum correlation entropy is implemented.

[0132] Specifically, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0133] The relevant technical solutions are the same as above and will not be described again here.

[0134] Example 4

[0135] The present application provides a computer program product, including a computer program. When the computer program is run on a computer, the computer executes the above-mentioned method for online identification of sodium ion energy storage battery parameters based on maximum correlation entropy.

[0136] The relevant technical solutions are the same as above and will not be described again here.

[0137] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for online identification of sodium ion energy storage battery parameters based on maximum correlation entropy, characterized in that: include: S1. Establish the first-order Thevenin model of sodium-ion energy storage battery: including the voltage source U in series OC , ohmic internal resistance R0 and R1C1 circuit consisting of resistor R1 and capacitor C1 in parallel; Among them, U OC represents the open circuit voltage of the sodium ion energy storage battery, and R0 represents the voltage of the electrodes, electrolyte, and diaphragm in the sodium ion energy storage battery to Na + The R1C1 circuit represents the polarization effect of the sodium ion energy storage battery; S2. Based on the first-order Thevenin model, a discrete time domain model of the sodium ion energy storage battery is obtained: IN t,k -IN OC,k =a0I k +a1I k-1 +b1(U t,k-1 -IN OC,k-1 ),k≥2 Among them, U t,k , U OC,k ,I k Corresponding to the terminal voltage U of the sodium ion energy storage battery t , open circuit voltage U OC , the kth sampling value of the current I; and the parameters a0, a1, b1 satisfy: Where, T is the preset sampling period, τ1=R1C1; S3. A cost function is constructed by adopting the maximum correlation entropy criterion based on the Gaussian kernel function to estimate the parameters a0, a1, and b1 of the discrete time domain model, thereby obtaining the parameters R0, R1, and C1 of the sodium ion energy storage battery, and completing the parameter identification of the sodium ion energy storage battery.

2. The method for online identification of sodium ion energy storage battery parameters according to claim 1, characterized in that: In S3, the cost function constructed using the maximum correlation entropy criterion based on the Gaussian kernel function is: In the formula, represents the correlation entropy based on the Gaussian kernel function, σ is the kernel length of the Gaussian kernel function, and λ represents the forgetting factor; Represents θ k The estimated value of θ k ,φ k ,y k The corresponding parameter matrix, input matrix and output matrix of the discrete time domain model are: θ k =[a 0,k a 1,k b 1,k ] T ,φ k =[I k I k-1 U t,k-1 -U OC,k-1 ] T ,y k =U t,k -U OC,k ; Among them, a 0,k 、a 1,k 、b 1,k Correspondingly represents the estimated values ​​of parameters a0, a1, and b1 at the kth sampling time.

3. The method for online identification of sodium ion energy storage battery parameters according to claim 2, characterized in that: The cost function is used to estimate the parameters a0, a1, and b1 of the discrete time domain model, including: The correlation entropy The maximum is the optimization target, and the following recursive formula is used to solve θ in real time k Estimates Among them, P k represents the covariance matrix of the kth sampling; K k is the gain matrix, 4. The method for online identification of sodium ion energy storage battery parameters according to any one of claims 1 to 3, characterized in that: In S2, based on the first-order Thevenin model, a discrete time domain model of the sodium ion energy storage battery is obtained, including: Based on the first-order Thevenin model, the continuous time domain model of sodium-ion energy storage batteries is determined; Performing Laplace transformation on the continuous time domain model to obtain a corresponding continuous transfer function H(s); and performing bilinear transformation on the continuous transfer function H(s) to obtain a corresponding discrete transfer function H(z); The discrete time domain model is determined using the discrete transfer function H(z).

5. The method for online identification of sodium ion energy storage battery parameters according to claim 2 or 3, characterized in that: Input matrix φ k Middle,U OC,k-1 Determined by: Obtain the OCV-SoC function relationship of the sodium ion energy storage battery; where OCV represents the open circuit voltage U OC , SoC represents the state of charge; The open circuit voltage value U is obtained by the OCV-SoC function relationship OC,k-1 .

6. The method for online identification of sodium ion energy storage battery parameters according to claim 5, characterized in that Obtain the OCV-SoC function relationship of sodium-ion energy storage batteries, including: Conduct low current charge and discharge experiments on sodium ion energy storage batteries to obtain the relationship between the open circuit voltage and the state of charge of the sodium ion energy storage battery during discharge, and the relationship between the open circuit voltage and the state of charge of the sodium ion energy storage battery during charge; The open circuit voltage during charging and the open circuit voltage during discharging under the same state of charge are averaged to obtain a relationship curve between the average open circuit voltage and the state of charge; The relationship curve between the average open circuit voltage and the state of charge is fitted with a polynomial to obtain the OCV-SoC function relationship.

7. The method for online identification of sodium ion energy storage battery parameters according to claim 6 is characterized in that: Performing polynomial fitting on the relationship curve between the average open circuit voltage and the state of charge includes: Between 10% and 95% of the SoC, with 35% SoC as the boundary, a piecewise second-order polynomial fitting is performed on the relationship curve between the average open circuit voltage and the state of charge.

8. An electronic device, characterized in that: comprising a computer readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the online identification method of sodium ion energy storage battery parameters according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the online identification method for parameters of a sodium ion energy storage battery as described in any one of claims 1 to 7 is implemented.

10. A computer program product, characterized in that It includes a computer program, which, when running on a computer, enables the computer to execute the method for online identification of sodium ion energy storage battery parameters according to any one of claims 1 to 7.

Images