Super capacitor SOC estimation method considering leakage current

Abstract

The invention discloses a super-capacitor SOC estimation method considering leakage current, and belongs to the technical field of power management. Comprising the steps of establishing a super-capacitor model based on a support vector machine; obtaining a first sample data set through a cell experiment; in a standing state, a second sample data set is obtained online, and the first sample data set and the second sample data set are used to complete training of the super capacitor model based on the support vector machine; and obtaining an environment variable of a to-be-tested super capacitor, obtaining a change value of the SOC based on the trained super capacitor model based on the support vector machine, and completing SOC estimation of the super capacitor. According to the invention, dynamic modeling and real-time correction are carried out on the leakage current according to the actual working state and the environmental condition of the super capacitor, the accuracy of SOC estimation is improved, the over-charging or over-discharging of the power supply can be effectively prevented, the energy management strategy is optimized, and the service life of the system is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of power management, and in particular to a supercapacitor SOC estimation method considering leakage current. Background Art

[0002] Supercapacitors have been widely used in electric vehicles, renewable energy storage systems, and industrial power systems in recent years due to their high power density, fast charge and discharge capabilities, and long cycle life. Compared with traditional lithium batteries, supercapacitors can better cope with short-term high power demands and have a longer service life. Therefore, supercapacitors become ideal energy storage devices in scenarios that require frequent energy conversion and storage.

[0003] One of the core functions of the power management system is to accurately estimate the SOC (state of charge) of the power system, that is, the estimation of the remaining capacity of the power supply. Accurate SOC estimation can effectively prevent overcharging or over-discharging and improve the overall efficiency and safety of the system. Currently, the commonly used SOC estimation methods include voltage integration method, coulomb counting method and model prediction method. The traditional coulomb counting method estimates the state of charge of the capacitor by measuring the integral of the current during the charging and discharging process, while the voltage integration method determines the remaining charge of the capacitor by monitoring the change in the voltage at the capacitor terminal. These methods can achieve relatively accurate estimation in the dynamic process of normal charging and discharging. However, supercapacitors have a leakage current phenomenon that cannot be ignored. Leakage current is the spontaneous charge loss inside the capacitor when the power supply is in a static or light load working state. If it is not taken into account, leakage current will cause a large error in SOC estimation, especially when the static time is long, this error may accumulate into a significant deviation. The model prediction method was proposed to compensate for the leakage current of the supercapacitor to a certain extent. For example, by estimating the charge loss during the static time and making corrections, or by establishing a simple leakage current model, the SOC estimation is roughly corrected. However, these leakage current compensation schemes are mostly based on fixed parameters and do not adequately consider the dynamic characteristics of the leakage current. Especially in complex application scenarios, the size and variation of the leakage current may vary depending on multiple factors such as ambient temperature and charge and discharge status. Summary of the invention

[0004] In view of the defects of the prior art, the present invention accurately estimates the change in SOC caused by the leakage current of the battery cell during the static process by performing SVM modeling on the dynamic characteristics of the leakage current, so as to solve the problem of inaccurate SOC estimation caused by the leakage current in the prior art.

[0005] In order to achieve the above object, the present invention provides a supercapacitor SOC estimation method considering leakage current, comprising:

[0006] (1) Establish a supercapacitor model based on support vector machine, with the input being the environmental variables of the supercapacitor and the output being the supercapacitor SOC change value;

[0007] (2) obtaining the environmental variables of the supercapacitor and the corresponding supercapacitor SOC change values ​​through the battery cell experiment as the first sample data set to train the supercapacitor model based on the support vector machine;

[0008] (3) in a static state, online obtaining the environmental variables when the supercapacitor voltage drops to 0V and the supercapacitor SOC change value as a second sample data set, and training the supercapacitor model based on the support vector machine;

[0009] (4) obtaining the environmental variables of the supercapacitor to be tested, inputting the environmental variables into the trained supercapacitor model based on the support vector machine, and obtaining the change value of SOC;

[0010] (5) Estimating the SOC of the supercapacitor based on the change in the SOC.

[0011] Furthermore, the environmental variables of the supercapacitor include current temperature, static start voltage, static duration, and current voltage; and the supercapacitor SOC change value is an SOC drop value caused by leakage current during the static duration.

[0012] Furthermore, the supercapacitor model fitting formula for the support vector machine is:

[0013]

[0014] Among them: a i is the weight of the support vector, b is the bias term, x i is an environment variable, K(x i , x) is the kernel function.

[0015] Furthermore, the training process of the supercapacitor model based on support vector machine is as follows:

[0016] (2.1) For each input sample, calculate the model output value error;

[0017] ξ c =|y c -f(x c )|

[0018]

[0019] (2.2) Classify the samples according to the model output value error;

[0020] The sample is divided into three subsets: the error support vector set E, the boundary support vector set S, and the remaining sample set R;

[0021] If ξ c = 0, then the sample belongs to the set R;

[0022] If 0 < ξ c < C, then the sample belongs to E;

[0023] If ξ c = C, then the sample belongs to S;

[0024] Where: C is a set threshold;

[0025] (2.3) If the sample belongs to the set S or the set E, it is used to update the model parameters. If the sample belongs to the set R, the model parameters are not updated;

[0026] (2.4) The objective function of the supercapacitor model based on the support vector machine is:

[0027]

[0028] Where: W is a vector composed of a i ; ξ is a predefined error variable; ξ i and are slack variables, both greater than 0;

[0029] By minimizing the objective function, the training of the supercapacitor model based on the support vector machine is completed.

[0030] Further, the specific step (5) is:

[0031] If the battery cell of the supercapacitor is not static, then the ampere-hour integration method is used to calculate the value of the SOC of the battery cell;

[0032] If the battery cell of the supercapacitor is static, then the SOC is corrected according to the change value of the SOC;

[0033] SOC(k) = SOC(k - 1) - △SOC.

[0034] Where: SOC(k) is the SOC value at the current moment, SOC(k - 1) is the SOC estimated value at the previous moment, and △SOC is the change value of the SOC under the environmental variables of the supercapacitor to be tested.

[0035] Advantages of the present invention:

[0036] 1. Dynamic leakage current compensation: According to the actual working state and environmental conditions of the capacitor, the present invention dynamically models and corrects the leakage current in real time, improving the accuracy of SOC estimation.

[0037] 2. Strong applicability: The present invention is suitable for supercapacitor SOC estimation under different working conditions, especially showing higher accuracy under long-term static and frequent charging and discharging conditions.

[0038] 3. Enhance system safety and reliability: The present invention can effectively prevent overcharging or over-discharging of power supply, optimize energy management strategy and extend system service life by improving the accuracy of SOC estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The figure is a flow chart of a method for estimating the SOC of a supercapacitor taking leakage current into consideration according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] like Figure 1 As shown, the present invention provides a supercapacitor SOC estimation method considering leakage current, comprising the following steps:

[0042] S101, establishing a supercapacitor model based on a support vector machine, with the input being the environmental variables of the supercapacitor and the output being the SOC change value of the supercapacitor.

[0043] The environmental variables of the supercapacitor include the current temperature, static starting voltage, static duration and current voltage, which are used as the input data of the support vector machine model, and the supercapacitor SOC change value (i.e., the SOC drop caused by leakage current) is used as the output data to build a supercapacitor model based on the support vector machine.

[0044] The support vector machine model is reduced to a type of nonlinear regression problem, which is used to fit the training sample input x i and output y i The relationship expression is:

[0045]

[0046] Among them: a i is the weight of the support vector, b is the bias term, x i is the input environment variable, K(x i , x) is the kernel function, Used to calculate the similarity between sample points and support vectors.

[0047] S102, obtaining environmental variables of the supercapacitor and corresponding supercapacitor SOC change values ​​through a battery cell experiment, taking them as a first sample data set, and training the supercapacitor model based on the support vector machine.

[0048] Through cell experiments, a data set {(x1,y1),(x2,y2),...(x l ,y l )} can be obtained as the first sample data set. Among them, y i is the SOC change value △SOC caused by leakage current, and x i is the environmental variable, including the current temperature, the starting voltage of standing, the duration of standing, and the current cell voltage.

[0049] The training process is as follows:

[0050] (1) For each input sample, calculate the error of the model output value.

[0051] ξ c =|y c -f(x c )|

[0052]

[0053] (2) Classify the samples according to the error of the model output value.

[0054] The samples are divided into 3 subsets: the error support vector set E, the boundary support vector set S, and the remaining sample set R.

[0055] If ξ c =0, the sample belongs to the set R.

[0056] If 0<ξ c <C, the sample belongs to E.

[0057] If ξ c =C, the sample belongs to S.

[0058] Among them: C is the set threshold.

[0059] (3) If the sample belongs to the set S or the set E, it is used to update the model parameters. If the sample belongs to the set R, the model parameters are not updated.

[0060] (4) The objective function of the supercapacitor model based on the support vector machine is:

[0061]

[0062] Among them: W is a vector composed of a i , ξ is a predefined error variable, and ξ i and are slack variables both greater than 0.

[0063] Minimize its objective function to complete the training of the supercapacitor model based on the support vector machine.

[0064] S103 . In a static state, online obtaining the environmental variables when the supercapacitor voltage drops to 0V and the supercapacitor SOC change value as a second sample data set to train the supercapacitor model based on the support vector machine.

[0065] In addition to using experimental data for model training, the embodiment of the present invention also trains the model according to the actual working state and environmental conditions of the supercapacitor.

[0066] The environmental variables when the supercapacitor voltage drops to 0V and the supercapacitor SOC change value are obtained online as a second sample data set, and the training of the supercapacitor model based on the support vector machine is completed with reference to the method of step S102.

[0067] S104, obtaining environmental variables of the supercapacitor to be tested, and inputting the environmental variables into the trained supercapacitor model based on the support vector machine to obtain a change value of the SOC.

[0068] Based on the trained model, the environmental variables of the supercapacitor to be tested are obtained to obtain the corresponding SOC change value.

[0069] S105 . Estimating the SOC of the supercapacitor based on the change in the SOC.

[0070] If the supercapacitor cell is not at rest, the ampere-hour integration method is used to calculate the SOC value of the cell.

[0071] If the supercapacitor cell is at rest, the SOC is corrected according to the SOC change value:

[0072] SOC(k)=SOC(k-1)-△SOC

[0073] Where: SOC(k) is the SOC value at the current moment, SOC(k-1) is the SOC value at the previous estimated moment, and △SOC is the change in SOC of the supercapacitor to be tested under the environmental variables.

[0074] 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 principles and spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A supercapacitor SOC estimation method considering leakage current, characterized in that: The steps include: (1) Establish a supercapacitor model based on support vector machine, with the input being the environmental variables of the supercapacitor and the output being the supercapacitor SOC change value; (2) obtaining the environmental variables of the supercapacitor and the corresponding supercapacitor SOC change values ​​through the battery cell experiment as the first sample data set to train the supercapacitor model based on the support vector machine; (3) in a static state, obtaining online the environmental variables when the supercapacitor voltage drops to 0V and the supercapacitor SOC change value as a second sample data set, and training the supercapacitor model based on the support vector machine; (4) obtaining the environmental variables of the supercapacitor to be tested, inputting the environmental variables into the trained supercapacitor model based on the support vector machine, and obtaining the change value of SOC; (5) Estimating the SOC of the supercapacitor based on the change in the SOC.

2. The supercapacitor SOC estimation method considering leakage current according to claim 1, characterized in that: The environmental variables of the supercapacitor include current temperature, static start voltage, static duration, and current voltage; the supercapacitor SOC change value is the SOC drop value caused by leakage current during the static duration.

3. The supercapacitor SOC estimation method considering leakage current according to claim 1, characterized in that: The supercapacitor model fitting formula for the support vector machine is: Among them: a i is the weight of the support vector, b is the bias term, x i is an environment variable, K(x i , x) is the kernel function.

4. The supercapacitor SOC estimation method considering leakage current according to claim 3, characterized in that: The training process of the supercapacitor model based on support vector machine is as follows: (2.1) For each input sample, calculate the model output value error; x c =|y c -f(x c )| (2.2) Classify the samples according to the model output value error; Divide the samples into three subsets: error support vector set E, boundary support vector set S and remaining sample set R; If c =0, the sample belongs to the set R; If 0<ξ c <C, then the sample belongs to E; If c =C, then the sample belongs to S; Where: C is the set threshold; (2.3) If the sample belongs to set S or set E, it is used to update the model parameters. If the sample belongs to set R, the model parameters are not updated. (2.4) The objective function of the supercapacitor model based on support vector machine is: Where: W is a i A vector composed of; ξ is a predefined error variable; ξ i and are slack variables, all greater than 0; The training of the supercapacitor model based on the support vector machine is completed by minimizing the objective function.

5. The supercapacitor SOC estimation method considering leakage current according to claim 1, characterized in that: The step (5) is specifically as follows: If the cell of the supercapacitor is not at rest, the SOC value of the cell is calculated using the ampere-hour integration method; If the cell of the supercapacitor is at rest, then the SOC is corrected according to the change value of the SOC; SOC(k)=SOC(k-1)-ΔSOC. Where: SOC(k) is the SOC value at the current moment, SOC(k-1) is the estimated SOC value at the previous moment, and △SOC is the change in SOC under the environmental variables of the supercapacitor to be tested.

Images