A method and system for detecting activation of a power cell

Abstract

The present application relates to the technical field of electric variable measurement, and especially relates to a detection method and a detection system for a starting power cell. The method comprises the following steps: obtaining a local abnormal trend of a data point according to the difference between the instantaneous amplitude of the data point and the average of the instantaneous amplitudes of the window of the data point, the standard deviation of the instantaneous amplitude, and the average of the window fitting slopes of the data point and its two adjacent data points, so as to obtain the proportion of the suspicious data point in the current time sequence, and constructing a slope value sequence based on the instantaneous amplitude slope values between the adjacent data points to obtain a global abnormal trend of the current time sequence; calculating the bin width of the current time sequence, and the bin width is negatively correlated with the global abnormal trend; using the bin width of the current time sequence in the HBOS algorithm to obtain the abnormal detection result of each data point, and effectively improving the detection result accuracy of the starting power cell.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric variable measurement, and in particular to a detection method and a detection system for a starting power cell. BACKGROUND

[0002] As a core power supply unit of a vehicle, the performance of the starting power directly affects the reliability of engine starting, the stability of the electrical system and the safety of driving. Among them, the current data of the starting power cell is a key indicator reflecting its performance state. Abnormal fluctuations in current (such as sudden increase, sudden decrease, irregular oscillation, etc.) may indicate that there are process defects or potential faults in the cell, thereby increasing the risk of thermal runaway.

[0003] The HBOS (Histogram-based Outlier Score, referred to as HBOS) algorithm is an unsupervised algorithm based on histogram. The core idea is to regard each feature of the current data as an independent variable. By constructing a histogram to calculate the abnormal score, the abnormal detection of the starting power cell can be realized. The characteristic of the HBOS algorithm is high calculation efficiency, which can quickly process the current data of the starting power cell generated in real time, and is therefore suitable for abnormal detection when testing real-time vehicle-mounted systems.

[0004] When the current data is skewed, the abnormal current data will be surrounded by a large number of normal current data, and its neighborhood points are mostly normal current data. At this time, if the conventional HBOS algorithm is still used to process and detect abnormal current data, the abnormal current data will be submerged in a large number of normal data, which will eventually reduce the detection accuracy of the abnormal current data of the starting power cell and cannot meet the demand of high-precision detection.

[0005] In summary, how to accurately obtain the detection result of the starting power cell is a problem to be solved at present. SUMMARY

[0006] In order to solve the technical problem of how to accurately obtain the detection result of the starting power cell, the present application provides a detection method and a detection system for a starting power cell.

[0007] In the first aspect, the present application provides a detection method for a starting power cell, which adopts the following technical scheme:

[0008] A detection method for a starting power cell, comprising the steps of:

[0009] The instantaneous amplitude of each data point in the current time sequence of the starting power supply cell is obtained, a window of each data point is constructed based on a preset window size, the instantaneous amplitude in the window is fitted to obtain a window fitting slope of the data point, a local abnormal trend of the data point is obtained according to the difference between the instantaneous amplitude of the data point and the average of the instantaneous amplitude of the window, the standard deviation of the instantaneous amplitude, and the average of the window fitting slopes of the data point and its two adjacent data points, a suspicious data point is determined according to the comparison result of the local abnormal trend of the data point and a local threshold, the proportion of the suspicious data point in the current time sequence is obtained, and a slope value sequence is constructed based on the instantaneous amplitude slope values between adjacent data points; the ratio of the standard deviation and the average of the slope value sequence is multiplied by the proportion of the suspicious data point in the current time sequence to obtain a global abnormal trend of the current time sequence; the bin width of the current time sequence is calculated, and the bin width is negatively correlated with the global abnormal trend; the bin width of the current time sequence is used in the HBOS algorithm to obtain an abnormal detection result of each data point.

[0010] The application can accurately obtain the detection result of the starting power supply cell by processing the current data sequence of the starting power supply cell through the HBOS algorithm. In the processing process of the HBOS algorithm, it is considered that the default bin width in the algorithm cannot adapt to data skew, so that when the abnormal data point is surrounded by normal data points, the bin width is too high to accurately detect the abnormality. Therefore, the local trend change and the global instantaneous amplitude trend change of the instantaneous amplitude of each data point in the current data sequence are analyzed, and the default bin width is adjusted, which effectively avoids the misjudgment caused by local fluctuations and improves the accuracy of abnormal detection. When obtaining the local trend change, the abnormal trend in the local window is captured by the offset between the instantaneous amplitude of the data point and the average level and the change of the instantaneous amplitude slope in the window, so that the local abnormal trend can be accurately obtained, and the accuracy of the detection of the starting power supply cell is effectively improved.

[0011] According to the detection method of the starting power supply cell provided by the application, the instantaneous amplitude of each data point in the current time sequence of the starting power supply cell is obtained, which comprises: collecting the current time sequence of the starting power supply cell in the detection process, and obtaining the current time sequence after pretreatment; the current time sequence is processed by using Hilbert transform to obtain the instantaneous amplitude of each data point in the current time sequence.

[0012] The application considers that noise may be collected when the sensor collects data, and such noise will interfere with the identification of abnormal trends. Therefore, the application can effectively avoid the influence of noise data on the subsequent detection result by pretreating the collected current data.

[0013] According to the present invention, a method for detecting a power supply cell includes, in which a window for each data point is constructed based on a preset window size, and the window fitting slope for the data point is obtained by fitting the instantaneous amplitude within the window, comprising: a preset window size of... ; Obtain continuous data centered on a single data point A set of data points are used as a window for that data point; the window fitting slope for that data point is obtained by fitting the instantaneous amplitude in the window using the least squares method.

[0014] This invention takes into account that when adjusting the default bin width, local abnormal trends can reflect local abnormal fluctuations in data points. By constructing a window of data points and analyzing the abnormal trends in the window, the bin width can be accurately reduced near the abnormal points to improve resolution.

[0015] According to the method for detecting a startup power supply cell provided by the present invention, obtaining the local anomaly trend of the data point includes:

[0016] ;

[0017] , The first Local anomaly trends and instantaneous magnitudes of individual data points , The first The instantaneous amplitude mean and instantaneous amplitude standard deviation of each data point window. For the first The window fitting slope for each data point For the first The mean slope of the window fitting between each data point and its two adjacent data points. To prevent a zero coefficient, For normalization function, It is the absolute value symbol.

[0018] According to the present invention, a method for detecting a power supply cell is provided, wherein determining a suspicious data point based on a comparison between the local anomaly trend of the data point and a local threshold includes: if the local anomaly trend of the data point is greater than the local threshold, then the data point is a suspicious data point.

[0019] According to the present invention, a method for detecting a power supply cell is provided, wherein calculating the bin width of the current time series includes: using the negative value of the global abnormal trend of the current time series as the exponential input of an exponential function with base e to obtain a correction factor; and using the product of the correction factor and the default bin width of the HBOS algorithm as the bin width of the current time series.

[0020] The application can effectively avoid misjudgment caused by local fluctuations and improve the accuracy of anomaly detection by adjusting the default bin width according to local abnormal trends and global abnormal trends.

[0021] According to the application, the method for obtaining the default bin width of the HBOS algorithm comprises the following steps: obtaining the difference between the maximum current value and the minimum current value in the current time sequence, and obtaining the default bin width according to the ratio of the difference to the preset bin number.

[0022] According to the application, the method for obtaining the default bin width of the HBOS algorithm comprises the following steps: obtaining the difference between the maximum current value and the minimum current value in the current time sequence, and obtaining the default bin width according to the ratio of the difference to the preset bin number.

[0023] According to the application, the method for obtaining the default bin width of the HBOS algorithm comprises the following steps: obtaining the difference between the maximum current value and the minimum current value in the current time sequence, and obtaining the default bin width according to the ratio of the difference to the preset bin number.

[0024] According to the application, the method for obtaining the default bin width of the HBOS algorithm comprises the following steps: obtaining the difference between the maximum current value and the minimum current value in the current time sequence, and obtaining the default bin width according to the ratio of the difference to the preset bin number.

[0025] In the second aspect, the application provides a detection system for the starting power supply cell, which adopts the following technical scheme.

[0026] The detection system for the starting power supply cell comprises a processor and a memory, and the memory stores computer program instructions.

[0027] The detection system for the starting power supply cell comprises a processor and a memory, and the memory stores computer program instructions.

[0028] The application has the following technical effects:

[0029] Based on the technical scheme, the detection method and the detection system of the starting power cell are provided, the current data sequence of the starting power cell is processed by the HBOS algorithm, and the detection result of the starting power cell can be accurately obtained. In the process of processing by the HBOS algorithm, it is considered that the default bin width in the algorithm cannot adapt to data skew, so that when the abnormal data points are surrounded by normal data points, the bin width is too high to accurately detect the abnormality. Based on this, the local trend change and the global instantaneous amplitude trend change of the instantaneous amplitude of each data point in the current data sequence are analyzed, and the default bin width is adjusted in combination with the two, so as to effectively avoid the misjudgment caused by local fluctuation and improve the accuracy of abnormal detection. When the local trend change is obtained, the abnormal trend in the local window is captured by the offset between the instantaneous amplitude of the data point and the average level and the instantaneous amplitude slope change in the window, so that the local abnormal trend can be accurately obtained, and the accuracy of the detection of the starting power cell is effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 A flowchart of a detection method of a starting power cell is provided in the embodiment of the present application. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, but not all the embodiments.

[0032] The detection method of the starting power cell provided by the present application can accurately adjust the default bin width of the HBOS algorithm by analyzing the local abnormal trend of each current data and the global abnormal trend of the current data as a whole when detecting the starting power cell, so as to effectively improve the detection accuracy of the abnormal current data of the starting power cell.

[0033] For details, please refer to Figure 1 as shown, Figure 1 A flowchart of a detection method of a starting power cell is provided in the embodiment of the present application, which specifically includes the following steps:

[0034] S1: Obtain the current value of each data point in the current time sequence sequence of the starting power cell.

[0035] For example, in the embodiment of the present application, the current time sequence data of the starting power cell in the detection process is collected and preprocessed to obtain the current time sequence.

[0036] Specifically, when simulating the starting working condition of the vehicle, the test environment can be set to an environment of-30°C to 85°C, and then the vehicle can be subjected to a charging and discharging cycle process. This process can be in turn a first static state, constant current charging, constant voltage charging, a second static state, and pulse discharging. The duration of the first static state is less than that of the second static state, and the duration of the first static state can be set to 10 seconds, and the duration of the second static state can be set to 30 seconds.

[0037] At each collection time, the current sensor is used to collect the current value of the starting power cell in the detection process. A collection time and its corresponding current value are taken as a data point to obtain current time series data. The current value collection frequency can be set to 500Hz, and the collection frequency can be set according to actual needs, which is not limited herein.

[0038] For example, when the current time series data is preprocessed, an adaptive filtering algorithm can be used to remove noise in the collected current time series data. Interpolation is used to fill in missing values. The preprocessing method and specific implementation method can be set according to actual needs, which is not repeated herein.

[0039] It should be noted that abnormal current data appears as a microscopic peak in the current time series sequence. The default bin width in the HBOS algorithm can smooth the abnormal value, causing the peak to be submerged in normal data. Therefore, the present embodiment provides a new bin width adjustment method. By analyzing the abnormal trend of the current data, the bin width can be optimized, and a finer bin width is used in the abnormal area, thereby amplifying the difference between the abnormal point and the normal point, improving the accuracy of abnormal detection, i.e., performing the following steps.

[0040] S2: Obtain the instantaneous amplitude of each data point in the current time series sequence of the starting power cell; construct a window of each data point based on a preset window size, fit the instantaneous amplitude in the window to obtain a window fitting slope of the data point; and calculate the local abnormal trend of each data point.

[0041] It should be noted that when adjusting the bin width, the local abnormality and the global abnormality of the data point need to be considered at the same time, so that the bin width can be reduced near the abnormal point, and the bin width can be generally reduced when the overall abnormal level is increased. The Hilbert transform is used to extract the envelope line of the current time series sequence to obtain the instantaneous amplitude of each data point. This value reflects the trend change of the current signal, and the signal remains stable in a local time. The slope of the instantaneous amplitude should be consistent with the overall trend of the previous and subsequent time.

[0042] Based on this, embodiments of the present invention can construct a local window for a data point. Within the window, based on the difference between the instantaneous amplitude of the data point and the mean instantaneous amplitude of the window, as well as the mean window fitting slope of the data point and its two adjacent data points, the local abnormal trend of the data point can be obtained. Here, the mean instantaneous amplitude of the window represents the average level of instantaneous amplitude within the local range, the standard deviation of the instantaneous amplitude represents the degree of fluctuation in the local instantaneous amplitude, and the mean window fitting slope represents the difference between the data point and the overall trend formed by its adjacent data points.

[0043] For example, in an embodiment of the present invention, obtaining the instantaneous amplitude of each data point in the current time sequence of the startup power cell includes: processing the current time sequence using Hilbert transform to obtain the instantaneous amplitude of each data point in the current time sequence.

[0044] The specific steps for using Hilbert transform to process the current time series and obtain the instantaneous amplitude of each data point in the current time series can be implemented by existing technology, and will not be elaborated here in the embodiments of the present invention.

[0045] For example, in an embodiment of the present invention, a window for each data point is constructed based on a preset window size, and the window fitting slope for that data point is obtained by fitting the instantaneous amplitude within the window, including: the preset window size is... ; Obtain continuous data centered on a single data point A set of data points are used as a window for that data point; the window fitting slope for that data point is obtained by fitting the instantaneous amplitude in the window using the least squares method.

[0046] in, It can be set to 5; the window size can be set according to actual needs.

[0047] For example, during fitting, the acquisition time can be used as the independent variable, and the instantaneous amplitude corresponding to the acquisition time can be used as the dependent variable. The fitting step can be implemented using existing technologies, and will not be elaborated upon here in this embodiment of the invention.

[0048] For example, in an embodiment of the present invention, calculating the local anomaly trend of data points includes:

[0049] ;

[0050] For the first Local anomalies in individual data points For the first The instantaneous amplitude of each data point For the first The instantaneous mean amplitude of a window of data points. For the first The instantaneous amplitude standard deviation of a data point window. For the first The window fitting slope for each data point For the first The mean slope of the window fitting between each data point and its two adjacent data points. To prevent a zero coefficient, For normalization function, It is the absolute value symbol.

[0051] An example illustrating how to obtain the mean slope of the window fitting: If If the current data point is the 4th data point, then the window fitting slopes of the 3rd, 4th, and 5th data points are obtained respectively, and the average of the window fitting slopes of these three data points is used as the average of the window fitting slopes of the 4th data point and its two adjacent data points.

[0052] In this calculation method, the zero-prevention coefficient is an extremely small positive number, which can be set to 0.01. The specific setting can be adjusted according to actual needs.

[0053] It is the first The absolute deviation of the instantaneous amplitude of each data point from the mean instantaneous amplitude of its window. This indicates the degree of deviation between the instantaneous amplitude and the average instantaneous amplitude. The smaller the deviation, the more stable the instantaneous amplitude of the current remains in a short period of time; the larger the deviation, the more the fluctuation intensity at the current moment deviates significantly from the recent normal level. Standardization is used to standardize the molecular part. The instantaneous amplitude of current time series data may exhibit different magnitudes at different times. Directly comparing absolute deviations will produce errors. Through standardization, it is possible to uniformly assess whether the deviation of the data point is significant under the current fluctuation level.

[0054] It is the first The window fitting slope of the nth data point and the nth data point The absolute difference between the mean slope of the window fit of each data point and its two adjacent data points is calculated by dividing this absolute difference by... This difference can be converted into a relative difference. In the signal stationary region, The value of approaches 0, so even tiny instantaneous amplitude anomalies can be sensitively captured; conversely, in areas of drastic signal change, anomalies are only triggered when the window fitting slope of the current data point deviates, thus avoiding mistaking normal fluctuations for anomalies.

[0055] In summary, in the first The greater the absolute difference between the instantaneous amplitude of the data point and the mean instantaneous amplitude of its window, the more likely the data point will be to have a larger absolute difference. The window fitting slope of the nth data point and the nth data point The greater the absolute difference between the data point and the average of the window fitting slopes of the adjacent data points on both sides, the greater the local abnormal trend of the data point, and the greater the possibility of the data point being an abnormal suspicious data point.

[0056] According to the above steps, the local abnormal trend of each data point in the current time sequence of the starting power cell can be obtained, and the suspicious data points can be preliminarily screened out. The more the suspicious data points, the greater the global trend abnormality degree of the current time sequence.

[0057] S3: determining the suspicious data points according to the comparison result of the local abnormal trend of the data points and the local threshold value, obtaining the proportion of the suspicious data points in the current time sequence, and constructing a slope value sequence based on the instantaneous amplitude slope values between the adjacent data points; and calculating the global abnormal trend of the current time sequence.

[0058] The local threshold value can be set to 0.5, and the local threshold value can be set according to actual needs.

[0059] It should be noted that in the case that the global abnormal trend of the current time sequence is relatively high, it indicates that the starting power cell is in an unstable state as a whole. At this time, even if the local abnormal trend of the data point is temporarily low, a smaller bin width also needs to be set to maintain the detection sensitivity. The overall abnormal trend is calculated by the local abnormal trend, and the combination of the two can effectively avoid misjudgment caused by local fluctuations and form a more stable binning strategy.

[0060] For example, in the embodiment of the application, the suspicious data points are determined according to the comparison result of the local abnormal trend of the data points and the local threshold value, which includes: if the local abnormal trend of the data point is greater than the local threshold value, the data point is a suspicious data point.

[0061] For example, the ratio of the standard deviation and the average value of the slope value sequence is multiplied by the proportion of the suspicious data points in the current time sequence to obtain the global abnormal trend of the current time sequence.

[0062] In this calculation method, the ratio of the standard deviation and the average value of the slope value sequence is the coefficient of variation of the slope value sequence, which can measure the relative fluctuation intensity of the slope sequence. The greater the value, the more intense the slope fluctuation, and the more intense the instantaneous amplitude change of the current, which may correspond to a sudden change in the local abnormal trend.

[0063] The greater the proportion of the suspicious data points in the current time sequence, the more times the local mutation trend appears in the current time sequence, and the greater the global abnormal trend of the corresponding instantaneous amplitude value.

[0064] ​After obtaining the global abnormal trend of the current time series based on the above steps, the default bin width of the current time series can be adjusted according to the global abnormal trend of the current time series, so as to accurately detect abnormal values.

[0065] S4: Calculate the bin width of the current time series, and the bin width is negatively correlated with the global abnormal trend; use the bin width of the current time series in the HBOS algorithm to obtain the abnormal detection result of each data point.

[0066] For example, in the embodiment of the application, the bin width of the current time series is calculated, including: taking the negative number of the global abnormal trend of the current time series as the exponential input of the exponential function with base e to obtain a correction factor; and taking the product of the correction factor and the default bin width of the HBOS algorithm as the bin width of the current time series.

[0067] It can be understood that the greater the global abnormal trend of the current time series is, the default bin width of the traditional HBOS algorithm should be appropriately reduced to enlarge the difference between abnormal points and normal points, thereby improving the detection sensitivity, and therefore the bin width is negatively correlated with the global abnormal trend.

[0068] For example, in the embodiment of the application, the method for obtaining the default bin width of the HBOS algorithm includes: obtaining the difference between the maximum current value and the minimum current value in the current time series, and obtaining the default bin width according to the ratio of the difference to the preset bin number.

[0069] The preset bin number can be set to 30, and the preset bin number can be set according to actual needs, which is not limited in the embodiment of the application.

[0070] For example, in the embodiment of the application, the bin width of the current time series is used in the HBOS algorithm to obtain the abnormal detection result of each data point, including: constructing a histogram of the current time series using the bin width of the current time series, counting the data point distribution frequency in each bin interval to obtain the probability density of each bin interval; taking the negative logarithm of the probability density of the bin interval as the HBOS abnormal score of each data point in the bin interval; and obtaining the detection result of the starting power cell according to the HBOS abnormal score of the data point.

[0071] The specific steps of obtaining the probability density of each bin interval according to the data point distribution frequency can be obtained by the specific steps in the HBOS algorithm, which is not described in detail in the embodiment of the application.

[0072] For example, when obtaining the detection result of the starting power cell according to the HBOS abnormal score of the data point, an abnormal score threshold can be preset, and the data point greater than the abnormal score threshold is taken as abnormal current data.

[0073] The abnormal score threshold can be set according to actual needs.

[0074] It can be understood that the starting power supply can be continuously optimized and adjusted based on the repeated tests on the starting power supply cell, so as to improve the product performance. In order to facilitate subsequent processing of the staff, a report can be generated after each detection result of the current time sequence is obtained.

[0075] For example, in the embodiment of the application, the abnormal detection result of each data point is obtained, and then the detection report is generated according to the detection result of the current time sequence of the starting power supply cell.

[0076] The detection report can be generated online, and the form of the detection report can be in the form of pdf. The manner of generating the detection report and the form of the detection report can be set according to actual needs, and the embodiment of the application does not make too many limitations here.

[0077] As can be seen, in the embodiment of the application, the instantaneous amplitude of each data point in the current time sequence of the starting power supply cell is obtained; a window of each data point is constructed based on a preset window size, and the instantaneous amplitude in the window is fitted to obtain a window fitting slope of the data point; the local abnormal trend of the data point is obtained according to the difference between the instantaneous amplitude of the data point and the average of the instantaneous amplitude of the window, the standard deviation of the instantaneous amplitude, and the average of the window fitting slopes of the data point and its two adjacent data points; the suspicious data point is determined according to the comparison result of the local abnormal trend of the data point and the local threshold, the proportion of the suspicious data point in the current time sequence is obtained, and the slope value sequence is constructed based on the instantaneous amplitude slope values between adjacent data points; the ratio of the standard deviation and the average of the slope value sequence is multiplied by the proportion of the suspicious data point in the current time sequence, to obtain the global abnormal trend of the current time sequence; the bin width of the current time sequence is calculated, and the bin width is negatively correlated with the global abnormal trend; the bin width of the current time sequence is used in the HBOS algorithm to obtain the abnormal detection result of each data point, which effectively improves the detection result accuracy of the starting power supply cell.

[0078] The embodiment of the application also discloses a detection system for a starting power supply cell, which comprises a processor and a memory, and the memory stores computer program instructions.

[0079] The above system also includes a communication bus and a communication interface and other components familiar to those skilled in the art, the settings and functions of which are known in the art, and therefore will not be described here.

[0080] In this application, the aforementioned memory can be any tangible media that includes or stores a program or programs that can be used by or in connection with an instruction execution system, apparatus, or device.

[0081] The above are preferred embodiments of the present application, not to limit the protection scope of the present application, therefore: all equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for detecting a startup power supply cell, characterized in that, include: Obtain the instantaneous amplitude of each data point in the current timing sequence of the starting power supply cell; A window for each data point is constructed based on a preset window size. The window fitting slope for the data point is obtained by fitting the instantaneous amplitude in the window. The local abnormal trend of the data point is obtained based on the difference between the instantaneous amplitude of the data point and the mean instantaneous amplitude of its window, the standard deviation of the instantaneous amplitude, the difference between the window fitting slope of the data point and the mean window fitting slope of the data point and its two adjacent data points, and the mean window fitting slope of the data point and its two adjacent data points. Suspicious data points are identified by comparing the local anomaly trends of data points with local thresholds, the proportion of suspicious data points in the current time series is obtained, and a slope value sequence is constructed based on the instantaneous amplitude slope values ​​between adjacent data points. The global anomaly trend of the current time series is obtained by multiplying the ratio of the standard deviation to the mean of the slope value series with the proportion of suspicious data points in the current time series. Calculating the bin width of the current time series includes: taking the negative of the global anomaly trend of the current time series as the exponential input of an exponential function with base e to obtain the correction factor; The product of the correction factor and the default bin width of the HBOS algorithm is used as the bin width of the current time series. The bin width is negatively correlated with the global anomaly trend. The HBOS algorithm uses the bin width of the current time series to obtain the anomaly detection results for each data point.

2. The method for detecting a startup power supply cell according to claim 1, characterized in that, The instantaneous amplitude of each data point in the current timing sequence of the startup power cell is obtained, including: After collecting the current timing data of the power supply cell during the detection process, preprocessing is performed to obtain the current timing sequence; Hilbert transform is used to process the current timing sequence to obtain the instantaneous amplitude of each data point in the current timing sequence.

3. The method for detecting a startup power supply cell according to claim 1, characterized in that, The process of constructing windows for each data point based on a preset window size and obtaining the window fitting slope for that data point by fitting the instantaneous amplitude within the window includes: The default window size is ; Obtain continuous data centered on a single data point A set of data points are used as a window for that data point; the window fitting slope for that data point is obtained by fitting the instantaneous amplitude in the window using the least squares method.

4. The method for detecting a startup power supply cell according to claim 1, characterized in that, The process of obtaining the local anomaly trend of the data point includes: ； , The first Local anomaly trends and instantaneous magnitudes of individual data points , The first The instantaneous amplitude mean and instantaneous amplitude standard deviation of each data point window. For the first The window fitting slope for each data point For the first The mean slope of the window fitting between each data point and its two adjacent data points. To prevent a zero coefficient, For normalization function, It is the absolute value symbol.

5. The method for detecting a startup power supply cell according to claim 1, characterized in that, The step of determining suspicious data points based on the comparison results of local anomaly trends and local thresholds includes: If the local anomaly trend of a data point is greater than the local threshold, then the data point is a suspicious data point.

6. The method for detecting a startup power supply cell according to claim 1, characterized in that, The default methods for obtaining bin width in the HBOS algorithm include: Obtain the difference between the maximum and minimum current values ​​in the current time sequence, and calculate the default compartment width based on the ratio of the difference to the preset number of compartments.

7. The method for detecting a startup power supply cell according to claim 1, characterized in that, The method of using the bin width of the current time series in the HBOS algorithm to obtain the anomaly detection results for each data point includes: A histogram of the current time series is constructed using the bin width of the current time series, and the distribution frequency of data points in each bin interval is counted to obtain the probability density of each bin interval. The negative logarithm of the probability density of the binning interval is used as the HBOS anomaly score of each data point in the binning interval; the detection result of the starting power cell is obtained based on the HBOS anomaly score of the data point.

8. The method for detecting a startup power supply cell according to claim 1, characterized in that, After obtaining the anomaly detection results for each data point, the process further includes: A test report is generated based on the test results of the current timing sequence of the starting power supply cell.

9. A detection system for a startup power supply cell, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a method for detecting a startup power cell according to any one of claims 1-8.

Images