A method, system and device for estimating battery health status based on short-term linear characteristics

Abstract

The present invention discloses a method, system and device for estimating the state of health of a battery based on short-time linear characteristics. The method includes the following steps: obtaining the current and voltage data of constant current charging of battery aging cycle data; calculating the time Dt spent in different voltage intervals; calculating the Pearson correlation coefficient between Dt in different voltage intervals and the state of health SOH of the battery; selecting the Dt calculated in the voltage interval with the largest absolute value according to the Pearson coefficient as the linear characteristic HF of the health of the battery; establishing a linear regression model with HF as the independent variable and SOH as the dependent variable and solving the coefficient; obtaining the HF of the healthy loaded battery to be determined, and calculating the health state. The present invention is based on the improvement of the data-driven method, greatly reducing the experiments required, with low cost and high calculation efficiency, and can be better applied to engineering practice.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery health state estimation, and in particular relates to a battery health state estimation method, system and device based on short-time linear characteristics. Background Art

[0002] As one of the core components of electric vehicles, the power battery system directly affects the performance level of the vehicle. In addition to being used in transportation vehicles, power battery systems are also widely used in many fields such as power energy storage systems, mobile robot platforms, communication base stations, aerospace and military industries. The battery management system (BMS) is mainly responsible for monitoring and controlling the battery, including state of charge (SOC) estimation, state of health (SOH) estimation, battery balancing, fault diagnosis, thermal management, etc., among which SOC and SOH estimation are basic and core functions. SOH is an indicator that describes the current maximum available capacity of the battery. It is generally defined as the ratio of the current maximum available capacity to the nominal capacity. Accurate SOH estimation can effectively reflect the attenuation of capacity with battery use, which is conducive to timely battery replacement, improving battery safety and reliability, and extending battery life. It can also provide an effective basis for the classified recycling and cascade utilization of power batteries.

[0003] At present, the research on battery SOH estimation methods can be mainly divided into two categories: model-based methods and data-driven methods. Model-based methods usually set the rated capacity as the actual capacity, then estimate the SOC through the observer, and then use the SOC estimation result to estimate the actual capacity. This cycle of estimation has obvious error accumulation effect, complex algorithm, and low calculation efficiency. Data-driven methods obtain data on the entire health cycle of the battery through a large number of experiments, train and learn through black box methods such as machine learning, and establish the relationship between charging and discharging data and SOH. This type of method usually costs a lot to obtain enough data to build a model. At the same time, the calculation complexity is high, and it is usually necessary to collect 30%-80% of the data during the battery charging or discharging process for SOH estimation. The data collection process is time-consuming, and the battery SOH estimation in areas such as battery cascade utilization is extremely inefficient.

[0004] In summary, the cost of obtaining training data based on data-driven methods is high, and the model-based SOH estimation method has an error accumulation effect and low computational efficiency. Therefore, it is very necessary to propose a battery SOH estimation method that can be accurate and fast at the same time. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a battery health state estimation method, system and device based on short-time linear characteristics, aiming to solve the technical problems in the prior art of complex data processing process and model, low calculation efficiency and poor accuracy.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for estimating a battery health state based on short-term linear characteristics comprises the following steps:

[0008] The first step is to use battery charging and discharging equipment to perform constant current and constant voltage charging-constant current discharge aging cycles on the new battery to obtain the current and voltage data of the battery constant current charging stage and the SOH corresponding to each cycle in the process of the battery capacity decreasing from 100% to 95% to 99%.

[0009] The second step is to extract data from the battery constant current charging data at a specified voltage value changing by a specified voltage amount, and calculate the time Dt taken to complete the voltage change in this interval.

[0010] The third step is to calculate the Pearson correlation coefficient r between the voltage interval Dt and SOH by the following formula.

[0011]

[0012] Where N is the number of samples, X and Y are Dt and SOH respectively.

[0013] The fourth step is to calculate Dt and the correlation coefficient r in different voltage intervals within the constant current charging curve of the battery, and calculate the voltage interval with the largest absolute value of the correlation coefficient r, and use the Dt calculated in this voltage interval as the health feature HF for estimating SOH.

[0014] The fifth step is to use the selected health characteristic HF as the independent variable and SOH as the dependent variable, and establish the linear regression equation SOH = a*HF + b + e, where e is the noise error. The linear regression model coefficients a and b are determined by the least squares estimation or maximum likelihood estimation method.

[0015] The sixth step is to charge the battery with constant current. The constant current section needs to cross the selected voltage range, calculate HF, and calculate the estimated SOH of the battery in the desired health state by the following formula:

[0016]

[0017] In the formula, are the estimated coefficients of the linear regression model.

[0018] A further improvement of the present invention is that the battery used is a lithium battery, a nickel-cadmium battery or a lead-acid battery.

[0019] A further improvement of the present invention is that in the first step, when there are already several batteries with different aging degrees, three constant current constant voltage charging-constant current discharge aging cycles are performed on each battery to obtain current and voltage data in the constant current charging stage.

[0020] A further improvement of the present invention is that the capacity setting at which the battery aging cycle experiment is stopped in the first step is determined according to the battery aging speed, ensuring that data of more than three cycles can be collected.

[0021] A further improvement of the present invention is that the size of the voltage variation interval in the second step is mainly determined based on the battery characteristics and hardware parameters, and the voltage variation interval is selected as small as possible while ensuring that all data can be collected normally to calculate Dt.

[0022] A further improvement of the present invention is that, in the fifth step, when there are many samples and large noise, residual resampling is used to iteratively solve the coefficients a and b.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The estimation method proposed in the present invention quickly estimates the battery health status by establishing short-time linear features. The speed is mainly reflected in three aspects: on the one hand, the data acquisition process is fast, and it can be completed within 60 seconds under the condition that the hardware conditions permit; on the other hand, the calculation is fast. The model used in the present invention is a linear regression model, and the calculation complexity is extremely low; thirdly, the training data is acquired quickly. The linear regression model only needs a small amount of sample data to model, and the required data can be obtained quickly. In summary, the present invention establishes a linear regression model through short-time features to achieve SOH estimation, which has the characteristics of short data acquisition time, fast acquisition of training sample data, simple algorithm, simple operation, and high calculation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a flow chart of the method proposed by the present invention.

[0026] Figure 2 Schematic diagram of the short-time linear characteristic Dt structure in this embodiment.

[0027] Figure 3 is the Pearson linear correlation coefficient between Dt and SOH in different voltage intervals of the battery in this embodiment.

[0028] Figure 4 This is a flow chart of solving the linear regression model coefficients by residual resampling least squares estimation in this embodiment. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0032] See also Figure 1 The present invention provides a method for estimating a battery health state based on short-term linear characteristics, comprising the following steps:

[0033] The first step is to use battery charging and discharging equipment to perform constant current and constant voltage charging-constant current discharge aging cycles on the new battery to obtain constant current charging current and voltage data and the SOH data corresponding to each cycle.

[0034] The second step is to extract data from the battery constant current charging data at a specified voltage value changing by a specified voltage amount, and calculate the time Dt taken to complete the voltage change in this interval.

[0035] The third step is to calculate the Pearson correlation coefficient r between the voltage interval Dt and SOH by the following formula.

[0036]

[0037] The fourth step is to calculate Dt and the correlation coefficient r in different voltage intervals within the constant current charging curve of the battery, and calculate the voltage interval with the largest absolute value of the correlation coefficient r, and use the Dt calculated in this voltage interval as the health feature HF for estimating SOH.

[0038] The fifth step is to use the selected health characteristic HF as the independent variable and SOH as the dependent variable, and establish the linear regression equation SOH = a*HF + b + e, where e is the noise error. The linear regression model coefficients a and b are determined by the least squares estimation or maximum likelihood estimation method.

[0039] The sixth step is to charge the battery with constant current. The constant current section needs to cross the selected voltage range, calculate HF, and calculate the estimated SOH of the battery in the desired health state by the following formula:

[0040]

[0041] In the first step, when there are already several batteries with different aging degrees, three constant current constant voltage charge-constant current discharge aging cycles are performed on each battery to obtain the current and voltage data of the constant current charging stage. The capacity setting for stopping the battery aging cycle experiment is determined according to the battery aging speed to ensure that data of more than three cycles can be collected; in the second step, the size of the voltage change interval is mainly determined according to the battery characteristics and hardware parameters. The smallest possible voltage change interval is selected while ensuring that all data can be collected normally to calculate Dt; in the fifth step, when there are many samples and large noise, the residual resampling iteration is used to solve the coefficients a and b. The process is as follows Figure 4 .

[0042] Example:

[0043] The data used in this example include aging experiments on 8 commercial Kokam soft-pack batteries with a nominal capacity of 740 mAh, using a graphite negative electrode and a lithium cobalt oxide / lithium nickel cobalt oxide positive electrode. Aging cycles were performed in an incubator at an ambient temperature of 40°C, and 519 sets of cycle data were collected on all batteries, of which the capacity was obtained by integrating the 1C discharge curve.

[0044] The first 10% of the samples of the acquired battery aging data are used for analysis and modeling, and the rest are used for testing. From the battery constant current charging data, data is intercepted in the interval where the specified voltage value changes by the specified voltage amount, and the time Dt taken to complete the voltage change in the interval is calculated. In the constant current charging stage of this embodiment, the voltage rises from 2.8V to 4.2V, and the data of the voltage interval with a length of 0.03V is intercepted every 0.03V from 2.8V to calculate the time Dt taken for the voltage change.

[0045] Voltage range [V L，i , V L，i +ΔV], where V L = {V L，1 , V L，n2 , ..., V L，n},(V L，i ∈[V min , Vmax -ΔV]), V L is an arithmetic progression, from which a series of voltage interval segments can be obtained, and then a series of Dt can be further calculated.

[0046] The Pearson correlation coefficient r between the voltage interval Dt and SOH is calculated by the following formula.

[0047]

[0048] Calculate the Dt and correlation coefficient r of different voltage intervals in the constant current charging curve of the battery, and calculate the voltage interval with the largest absolute value of the correlation coefficient r, and use the Dt calculated in this voltage interval as the health feature HF for estimating SOH. Figure 2 In this embodiment, the Pearson coefficient of the charging curve in the voltage range [3.0V-3.3V] is close to 1, indicating that the characteristic Dt and SOH have a very obvious positive linear relationship in the voltage range [3.0V-3.3V]. There are points close to 1 or -1 near 2.8V, 3.75V, 3.8V and 3.95V, but they are stable. Therefore, in this embodiment, the Dt calculated by selecting the voltage interval with the highest Pearson coefficient in the voltage range [3.0V-3.3V] is used as the battery health characteristic HF for modeling and SOH estimation.

[0049] The selected health feature HF is used as the independent variable and SOH is used as the dependent variable to establish a linear regression equation SOH = a*HF + b + e, where e is the noise error. The linear regression model coefficients a and b are determined by the least squares estimation or maximum likelihood estimation method. In this embodiment, the least squares estimation is selected to solve the model coefficients a and b, and the calculation formula is as follows:

[0050]

[0051] Since the least squares estimation method is extremely sensitive to outliers, this embodiment uses an iterative modeling approach to overcome this problem. Figure 4 , X0, Y0 are the features and SOH in the original data set samples, a linear regression model is established, and the solution is calculated to obtain a k , b k , and get the residual e k , theoretically the residual e k is a random error that obeys a Gaussian distribution. Therefore, the 3σ principle is used to identify outliers, and the data identified as outliers in the sample are removed. The least squares estimation is used again to iterate the solution until the residual e k The root mean square value converges to obtain the final and the final linear regression model.

[0052] The battery analysis model is used as 90% of the data for the test of this embodiment. The battery is charged with a constant current, and the constant current section needs to cross the selected voltage range, and the HF is calculated. The estimated SOH of the battery in the health state to be calculated is calculated by the following formula: And the error of the estimated SOH is calculated. In this embodiment, the SOH estimation is realized, and the root mean square error (RMSE) is 0.5% on the best performing battery, and the maximum error (MaxE) is only 1.4%. The RMSE and MaxE of the SOH estimation of the remaining batteries are shown in Table 1. The overall effect of the SOH estimation of the 8 batteries is excellent, the average RMSE of the 8 batteries is 1.35%, and the maximum MaxE is 4.66%.

[0053] Table 1

[0054]

[0055] The proposed battery health state estimation method based on short-term linear characteristics has high accuracy. The root mean square value of the SOH estimation error tested in the embodiment is 1.35%, and the maximum estimation error does not exceed 4.66%. In the embodiment, the data acquisition time does not exceed 60s. Only the first 10% of the sample data in the battery aging data are used for analysis and modeling. Compared with the data-driven method that requires sample data of the entire health cycle for modeling, the scale of sample data that needs to be obtained is greatly reduced, and the cost of obtaining sample data is greatly reduced. At the same time, the algorithm is simple and the calculation is efficient.

[0056] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A battery health state estimation method based on short-term linear characteristics, characterized in that: The following steps are involved: Step 1: Obtain the current and voltage data of the battery aging cycle data constant current charging; Use battery charging and discharging equipment to perform constant current and constant voltage charging-constant current discharge aging cycles on new batteries, and obtain the current and voltage data of the battery constant current charging stage and the SOH corresponding to each cycle during the process of battery capacity decreasing from 100% to 95% to 99%; Step 2: Calculate the time Dt spent in different voltage intervals; From the battery constant current charging data, extract data at a specified voltage interval and calculate the time Dt taken to complete the voltage change in the interval; Step 3: Calculate the Pearson correlation coefficient between Dt and battery health status SOH in different voltage ranges; The Pearson correlation coefficient r between the voltage range Dt and SOH is calculated by the following formula: In the formula, N is the number of samples, X and Y are Dt and SOH respectively; Step 4: Select the Dt calculated in the voltage interval with the largest absolute value according to the Pearson correlation coefficient as the linear characteristic HF of the battery health; The voltage interval with the largest absolute value of the correlation coefficient r is calculated, and Dt calculated in this voltage interval is used as the health characteristic HF for estimating SOH; Step 5: Establish a linear regression model with the linear feature HF as the independent variable and the battery health status SOH as the dependent variable and solve the coefficient; Step 6: Obtain the HF of the battery whose health status is to be determined, and calculate the estimated SOH of the battery whose health status is to be determined.

2. The method for estimating the battery health state based on short-term linear characteristics according to claim 1, characterized in that: In step 5, a linear regression model is established and the coefficients are solved. The linear regression model equation established is as follows: SOH=a*HF+b+e; In the formula, e is the noise error, and the linear regression model coefficients a and b are determined by the least squares estimation or maximum likelihood estimation method.

3. The method for estimating the battery health state based on short-term linear characteristics according to claim 2, characterized in that: When there are many samples and the noise is large, residual resampling is used to iteratively solve the coefficients a and b.

4. The method for estimating the battery health state based on short-term linear characteristics according to claim 1, characterized in that: In step 6, the battery is charged with a constant current, and the constant current section crosses the selected voltage interval. The HF is calculated, and the estimated SOH of the battery in the desired health state is calculated by the following formula: In the formula, are the estimated coefficients of the linear regression model.

5. A system using the battery health state estimation method based on short-time linear characteristics according to any one of claims 1 to 4, comprising: A data acquisition module, used to acquire battery aging cycle data and current and voltage data of constant current charging; A time calculation module is used to calculate the time Dt spent in different voltage intervals; The correlation coefficient calculation module is used to calculate the Pearson correlation coefficient between Dt and battery health status SOH in different voltage ranges; A linear feature extraction module is used to select Dt calculated from the voltage interval with the largest absolute value according to the Pearson correlation coefficient as the linear feature HF of battery health; Linear regression model building module, used to build a linear regression model with linear feature HF as independent variable and battery health status SOH as dependent variable and solve the coefficient; The estimated SOH solving module is used to obtain the HF of the battery whose health state is to be determined, and calculate the estimated SOH of the battery whose health state is to be determined.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the battery health state estimation method based on short-time linear characteristics described in any one of claims 1 to 4 are implemented.

Images