Method for detecting outlier battery cell groups in real-time energy storage systems

Abstract

This invention discloses a method for diagnosing outliers in operational battery cell groups. The method targets an energy storage system composed of multiple parallel battery cell groups. It involves collecting battery operating data and calculating the skewness and kurtosis of each battery cell group, followed by computing the Z-score. Using a predefined Z-score threshold, the method filters out the outlier battery cell groups. Maintenance personnel can then pay close attention to these outlier groups or replace faulty cells if necessary at an earlier stage, before the energy storage system goes beyond control, ensuring the energy storage system continues to function normally.

Description

[Technical Field]

[0001] This invention relates to a method for outlier diagnosis of multiple battery cell groups in an energy storage system, particularly a method for screening outlier battery cell groups based on the calculation of skewness and kurtosis. [Previous Technology]

[0002] The conventional technology for energy storage devices is based on the quality inspection of individual battery cells, and no quality consistency inspection between groups has been found.

[0003] The inventors of this case first proposed a detection method for detecting the quality consistency between groups in a structure in which multiple series battery groups are connected in parallel. This method can detect outlier groups, correct or replace defective units in advance, maintain the normal quality of the system, and prevent the overall system from gradually deteriorating over time. [Summary of the Invention]

[0004] The energy storage system has the ability to charge and discharge quickly. When the energy storage system is electrically coupled to the grid, it can provide auxiliary services to the grid and perform automatic frequency control (AFC) to output or input electrical energy to correct grid frequency deviations. This helps to maintain the frequency drift of the grid caused by load fluctuations, thereby improving the stability of the grid.

[0005] This invention discloses a method for outlier diagnosis of multiple cell groups in an energy storage system. The energy storage system includes multiple cell groups operating in parallel. This invention calculates the skewness and kurtosis of each group, and then calculates its Z-score. Based on a set Z-score threshold, outlier cell groups are screened out. If an outlier group is detected, we can correct the group or replace the faulty cells in advance to maintain the basic operating function of the overall energy storage system.

Implementation Method

[0006] We use the following examples in conjunction with Tables 1-5 for illustration. The specifications in the examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0007] Table 1 shows the definitions of skewness and kurtosis.

[0008] Table 2 shows an example of the original data. There are 304 battery cells. The data of the voltage of the first five cells and three seconds are used as an example.

[0009] Table 3 shows the calculation of the average absolute value of skewness and the average value of kurtosis.

[0010] Table 4 shows the statistics of skewness and z-score.

[0011] Table 5 shows the statistics of kurtosis and Z-score.

[0012] Example Explanation:

[0013] An energy storage system comprises six groups of battery cells, each group containing 304 cells connected in series. These six groups of cells, along with devices such as a charge / discharge system (PCS), constitute an energy storage system. This energy storage system can be electrically coupled to the official power grid, providing automatic frequency control (AFC) services and improving the stability of the official power grid. In the operation of the energy storage system, this invention utilizes skewness and kurtosis, combined with Z-scores, to examine whether there are any outlier groups of cells among the six groups of cells.

[0014] For the definitions of skewness and Fisher's kurtosis, please refer to Table 1.

[0015] In this invention, six groups of battery cells are used in parallel to form an energy storage system. Each group of cells contains 304 cells. The goal of this invention is to identify any outliers among these six groups of cells. The diagnostic method of this invention is as follows: The voltage of each of the 304 cells in the group is statistically analyzed, specifically the voltage from the first cell (CV1) to the 304th cell (CV304). The skewness and kurtosis values ​​of each group of cells are then calculated. The operation steps are as follows: Step 1: Calculate the skewness and kurtosis for each group of cells. Step 2: Select data with current ranges of 4A~10A and -4A~-10A for analysis. The battery performance will vary under different currents. When the current is too low, it approaches the open-circuit voltage, and the voltage distribution difference is not significant; a larger current will also reduce the voltage distribution difference among the cells. Therefore, the time-domain analysis is converted to a current-domain analysis. This invention uses a cell group with a rated capacity of 200Ah. When the discharge current is 4A, its discharge rate is 0.02C (4A / 200A=0.02C); when the discharge current is 10A, its discharge rate is 0.05C (10A / 200A=0.05C). The current range selected in this invention is based on a C-rate of 0.02C to 0.05C. Other wider current ranges that can be selected include a C-rate of 0.02C to 0.20C. Step 3: Since the skewness signifies the direction of offset, the absolute value of the skewness is used to calculate the magnitude of the offset. Step 4: Calculate the mathematical mean of the absolute value of the skewness and the mathematical mean of the kurtosis for each cell group. Step 5: Calculate the outlier index Z-score for skewness / kurtosis of the cell group using the mean. Step 6: Set the outlier index threshold Z-score > 2, and filter out the third cell group as the outlier cell group.

[0016] Calculation of average cell voltage and standard deviation: , where is the number of cells in each cell group, is the voltage of the i-th cell, is the average cell voltage of the i-th cells, and is the standard deviation of the cell voltage of the i-th cells.

[0017] The original data sample is shown in Appendix Table 2. Table 2 only summarizes the data of the first three seconds and the voltage of the first five cells, including: time, cell group number, cell group voltage, cell group current, skewness (absolute value), skewness, kurtosis, and the voltage of the first five cells (CV1, CV2, CV3, CV4, CV5).

[0018] Table 2 shows an example of the original data, for 304 battery cells. The voltage and three-second data of the first five cells are used as an example:

[0019] The voltage of the 304 cells in the first group (CV1~CV304) was measured. At 22:00:00, the absolute value of the voltage skewness of the 304 cells was 0.435297048; the kurtosis was -0.00011244.

[0020] The voltage of the 304 cells in the second group (CV1~CV304) was measured. At 22:00:00, the absolute value of the voltage skewness of the 304 cells was 0.70140633; the kurtosis was -0.082520391.

[0021] The voltage of the 304 cells in the third group (CV1~CV304) was measured. At 22:00:00, the absolute value of the voltage skewness of the 304 cells was 7.127010436; the kurtosis was 94.23706603.

[0022] The voltage of the 304 cells in the 4th group (CV1~CV304) was measured. At 22:00:00, the absolute value of the voltage skewness of the 304 cells was 0.109338357; the kurtosis was -0.369991049.

[0023] The voltage of the 304 cells in the 5th group (CV1~CV304) was measured. At 22:00:00, the absolute value of the voltage skewness of the 304 cells was 1.379180162; the kurtosis was 4.478331217.

[0024] The voltage of the 304 cells in the 6th group (CV1~CV304) was measured. At 22:00:00, the absolute value of the voltage skewness of the 304 cells was 0.214862981; the kurtosis was -0.626448809.

[0025] The information at 22:00:01 and 22:00:02 has the same meaning as described above, and will not be repeated here.

[0026] Table 3 shows the calculation of the absolute average of skewness and the average of kurtosis.

[0027] Table 3 shows the calculation of the average absolute value of skewness and the average value of kurtosis, based on the data in Table 2. Table 3 is an example based on the data from the first three seconds of Table 2.

[0028] The data from the first three seconds were averaged to obtain the average absolute values ​​of the skewness for the first to sixth cell groups: 0.435297048; 0.70140633; 7.127010436; 0.172114929; 1.379180162; and 0.214862981.

[0029] The data from the first three seconds were averaged to obtain the average kurtosis of the cell groups from group 1 to group 6 as follows: -0.00011244; -0.082520391; 94.23706603; -0.249707213; 4.478331217; and -0.626448809.

[0030] Table 4 shows the statistics for skewness and Z-score. , , where is the mean; is the standard deviation. Skewness calculation: In a group of cells, if is the average skewness value of the i-th cell group, then the skewness Z-score of the i-th cell group can be calculated, as shown in Table 4. Kurtosis calculation: In a group of cells, if is the average kurtosis value of the i-th cell group, then the kurtosis Z-score of the i-th cell group can be calculated, as shown in Table 5.

[0031] Table 4 shows that the skewness values ​​of the six groups of battery cells are 0.435297048, 0.70140633, 7.127010436, 0.172114929, 1.379180162, and 0.214862981, respectively.

[0032] The mathematical mean of the skewness values ​​of the six cell groups is 1.671645314; its standard deviation is calculated to be 2.7087736. Then, its Z-score is calculated, Z-score = (popular mean of skewness values ​​- mathematical mean) / standard deviation. The Z-scores of groups 1 to 6 are: -0.45642363; -0.358183864; 2.013961291; -0.553582773; -0.107969583; and -0.537801442, respectively.

[0033] If the Z score threshold is set to 2, then when the Z score is greater than 2, it is determined to be an outlier cell group. The aforementioned third cell group has a Z score of 2.013961291. When the Z score threshold is 2, the third cell group will be determined to be an outlier cell group.

[0034] Table 5 shows the calculation methods for kurtosis and Z-score.

[0035] In this invention, the kurtosis value of the cell group is calculated using Fisher's kurtosis as an example, and the statistical method is the same as in the previous example. In this invention, the outlier index of the kurtosis value of each cell group is calculated using the Z-score (z-score) of the average kurtosis value of each cell group as an example. In this invention, the outlier index threshold Z-score (z-score) is 2 as an example.

[0036] Table 5 shows that the kurtosis values ​​of the six battery cell groups are -0.00011244, -0.082520391, 94.23706603, -0.249707213, 4.478331217 and -0.626448809, respectively.

[0037] The mathematical mean of the kurtosis values ​​of the six cell groups is 16.29276807; its standard deviation is calculated to be 38.231965. Then, its Z-score is calculated: Z-score = (population mean of kurtosis values ​​- mathematical mean) / standard deviation. The Z-scores of groups 1 to 6 are: -0.42616; -0.42831; 2.038721; -0.43269; -0.30902; and -0.44254, respectively.

[0038] If the Z-score threshold is set to 2, then when the Z-score is greater than 2, it is determined to be an outlier cell group. The aforementioned third cell group has a Z-score of 2.013961291. When the Z-score threshold is 2, the third cell group will be determined to be an outlier cell group.

[0039] In the above example, we use three seconds of data for analysis. In actual operation, we can use several minutes, several hours, or several days for data analysis.

[0040] Figure 1. Embodiment 1 of the present invention

[0041] Figure 1 shows a method for outlier diagnosis of battery cell groups in operation based on skewness. The steps include: Step 1: Measure the current of each battery cell group and record the voltage readings of each battery in each battery cell group at intervals over a period of time; as shown in Table 2; Step 2: Obtain the skewness of the voltage reading distribution of each battery cell group based on each set of voltage readings; Step 3: Obtain the average absolute skewness of each battery cell group during this period; as shown in Table 3; Step 4: Obtain the outlier index (Z score) of each battery cell group based on the total average skewness and the average absolute skewness of all battery cell groups; as shown in Table 4; Step 5: Set the outlier index (Z score) threshold; and Step 6: Output a list of battery cell groups that exceed the threshold.

[0042] Figure 2 Embodiment 2 of the present invention

[0043] Figure 2 shows a method for outlier diagnosis of battery cell clusters in operation based on kurtosis. The steps include: Step 1: Measure the current of each battery cell cluster and record the voltage readings of each battery in each battery cell cluster at intervals over a period of time; as shown in Table 2; Step 2: Obtain the kurtosis of the voltage reading distribution of each battery cell cluster based on each set of voltage readings; as shown in Table 3; Step 3: Obtain the average kurtosis of each battery cell cluster over this period of time; as shown in Table 5; Step 4: Obtain the outlier index (Z score) of each battery cell cluster based on the total average kurtosis of all battery cell clusters and the average kurtosis of each battery cell cluster; Step 5: Set the outlier index (Z score) threshold; and Step 6: Output a list of battery cell clusters that exceed the threshold.

[0044] Although the present invention specification has described several embodiments by way of example, various obvious modifications can be made by those skilled in the art without departing from the spirit of the appended claims, and these modifications still fall within the scope of the claims of this case. [Simplified Explanation of the Diagram]

[0045] Figure 1 shows an embodiment of the present invention.

[0046] Figure 2 shows the second embodiment of the present invention.

Claims

1. A method for detecting abnormal cell clusters in an instantaneous energy storage system based on skewness, wherein the energy storage system includes multiple cell clusters connected in parallel, each cluster containing multiple batteries connected in series, the method comprising: Step 1: Measure the current of each cell group and record the voltage readings of each battery in each cell group at intervals over a period of time; Step 2: Obtain the skewness of the voltage reading distribution of each cell group based on each set of voltage readings; Step 3: Obtain the average absolute skewness of each cell group over this period of time; Step 4: Obtain the outlier index (Z-score) of each cell group based on the total average skewness and the average absolute skewness of all cell groups; Step 5: Set the outlier index (Z-score) threshold; and Step 6: Output a list of cell groups that exceed the threshold.

2. The method for detecting abnormal cell clusters in a skewness-based instantaneous energy storage system as described in claim 1, wherein, The interval recording mentioned in step 1 refers to seconds, minutes, hours, days, weeks, or months.

3. The method for detecting abnormal cell clusters in a skewness-based instantaneous energy storage system as described in claim 1, wherein, The calculation of the skewness of the cell voltage distribution in step 2 is performed using data from a certain current range of the cell group.

4. The method for detecting abnormal cell clusters in a skewness-based instantaneous energy storage system as described in claim 3, wherein, The selected current range is based on 0.02C to 0.20C; where C is the charge / discharge rate (C-rate).

5. The method for detecting abnormal cell clusters in a skewness-based instantaneous energy storage system as described in claim 4, wherein, The selected current range is based on 0.02C to 0.05C; where C is the charge / discharge rate (C-rate).

6. A method for detecting abnormal cell clusters in a kurtosis-based instantaneous energy storage system, wherein, The energy storage system comprises multiple groups of battery cells connected in parallel, each group containing multiple batteries connected in series. The method includes: Step 1: Measuring the current of each battery cell group and recording the voltage readings of each battery within each group at intervals over a period of time; Step 2: Obtaining the kurtosis of the voltage reading distribution for each battery cell group based on each voltage reading set; Step 3: Obtaining the average kurtosis of each battery cell group over this period; Step 4: Obtaining the outlier index (Z-score) for each battery cell group based on the overall average kurtosis of all battery cell groups and the average kurtosis of each battery cell group; Step 5: Setting an outlier index (Z-score) threshold; and Step 6: Outputting a list of battery cell groups that exceed the threshold.

7. A method for detecting abnormal cell clusters in a kurtosis-based instantaneous energy storage system as described in claim 6, wherein, The interval recording mentioned in step 1 refers to seconds, minutes, hours, days, weeks, or months.

8. A method for detecting abnormal cell clusters in a kurtosis-based instantaneous energy storage system as described in claim 6, wherein, The calculation of the kurtosis of the cell voltage distribution in step 2 is performed by taking data from a certain current range of the cell group.

9. The method for detecting abnormal cell clusters in a kurtosis-based instantaneous energy storage system as described in claim 6, wherein, The selected current range is based on 0.02C to 0.20C; where C is the charge / discharge rate (C-rate).

10. The method for detecting abnormal cell clusters in a kurtosis-based instantaneous energy storage system as described in claim 9, wherein, The selected current range is based on 0.02C to 0.05C; where C is the charge / discharge rate (C-rate).

Images