Circuit operation monitoring method for new energy automobile direct current high voltage contactor

Abstract

The present application relates to the technical field of current data processing, in particular to a circuit operation monitoring method for new energy automobile direct current high-voltage contactor. First, the current data to be analyzed is obtained; further analyze the local density deviation to screen out the initial abnormal data; further divide the data interval according to the data fluctuation and data difference between the normal data and the adjacent initial abnormal data; further extract the characteristic value of the data distribution in the data interval, and perform clustering classification based on the characteristic value, analyze the difference of the data in different clustering clusters, and obtain the independent sub-parameters of each data interval combined with the characteristic value; finally, according to the fluctuation of the independent sub-parameters, combined with the distribution of the data difference coefficient, the hyperplane parameters are obtained, and the real abnormal data is obtained, the accurate identification of the abnormal current of the high-voltage contactor is realized, and the classification robustness and accuracy are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of current data processing, and in particular to a circuit operation monitoring method for a direct-current high-voltage contactor of a new energy vehicle. BACKGROUND

[0002] As a key component in the power system of an electric vehicle, a direct-current high-voltage contactor is mainly used to control the on-off of current between a battery pack and a motor, and to ensure the normal operation of the vehicle under different working conditions. In actual application, due to the high voltage and large current borne by the contactor, long-time operation may cause problems such as wear of contact points, temperature rise, and arc generation, which may lead to contactor failure and further affect the safety of the vehicle.

[0003] The monitoring system integrates sensors to detect current, voltage, temperature, and the opening and closing state of the contactor, and uses a support vector machine (SVM) algorithm to analyze the data to determine the operation of the contactor. However, due to the complex environment of high voltage and large current in which the direct-current high-voltage contactor operates, interference factors such as temperature fluctuations, vibration, and electromagnetic interference will affect the measurement accuracy and stability of the current sensor, and will cause the support vector machine algorithm to be disturbed when finding the optimal classification hyperplane, causing the classification boundary to deviate from the ideal position, and ultimately leading to inaccurate identification of abnormal current data. SUMMARY

[0004] To solve the technical problem that the current sensor data is disturbed when the direct-current high-voltage contactor operates in a complex environment, making it difficult for the support vector machine algorithm to accurately find the optimal classification hyperplane, and further affecting the accuracy of abnormal current identification, the purpose of the present application is to provide a circuit operation monitoring method for a direct-current high-voltage contactor of a new energy vehicle, and the technical solution adopted is as follows:

[0005] Obtain current data of the direct-current high-voltage contactor to be analyzed;

[0006] According to the local density deviation of the current data, initial abnormal data is screened out; according to the data fluctuation and data difference between each normal data and adjacent initial abnormal data, data intervals are divided; according to the distribution of data in each data interval, feature values are obtained, and clustering classification is performed based on the feature values; according to the difference of data in different clustering clusters, a data difference coefficient of each clustering cluster is obtained;

[0007] According to the difference between the data difference coefficients, in combination with the feature values, an independent sub-parameter of each data interval is obtained; according to the fluctuation of the independent sub-parameters, in combination with the distribution of the data difference coefficients, a hyperplane parameter is obtained, and real abnormal data is obtained.

[0008] Further, the data interval acquisition method comprises:

[0009] When the normal data has only one adjacent initial abnormal data, the data interval is divided to the adjacent initial abnormal data;

[0010] In the normal data with two adjacent initial abnormal data, the normal data are selected one by one as the data to be analyzed; according to the variance of the data between the data to be analyzed and each adjacent initial abnormal data, the belonging degree corresponding to the data to be analyzed and each adjacent initial abnormal data is acquired by combining the absolute value of the difference and the Euclidean distance between the two data;

[0011] The data interval is divided according to the belonging degrees corresponding to all the data to be analyzed.

[0012] Further, the method for dividing the data interval according to the belonging degrees corresponding to all the data to be analyzed comprises:

[0013] In the normal data between the two adjacent initial abnormal data, the data interval of the initial abnormal data corresponding to the maximum belonging degree of the normal data is taken as the division result; and the isolated normal data in the time domain of the data interval is divided into the data interval on the other side.

[0014] Further, the feature value acquisition method comprises:

[0015] The feature value of each data interval is acquired by fusing the maximum value, the variance and the number of the data in each data interval.

[0016] Further, the data difference coefficient acquisition method comprises:

[0017] The data difference coefficient of each cluster is acquired according to the difference between the average value of the feature values in each cluster and the average value of all the feature values, and the absolute average deviation of the feature values in the cluster.

[0018] Further, the independent sub-parameter acquisition method comprises:

[0019] The independent sub-parameter of each data interval is acquired by fusing the ratio of the range to the average value of the data difference coefficients and the feature value of each data interval.

[0020] Further, the hyperplane parameter acquisition method comprises:

[0021] The hyperplane parameter is acquired according to the variance of the independent sub-parameter, the data quantity distribution on both sides of the average value of the data difference coefficients, and the average value of the data difference coefficients.

[0022] Further, the method for obtaining the hyperplane parameters based on the variance of the independent sub-parameters, combined with the data quantity distribution on both sides of the average value of the data difference coefficients, and the average value of the data difference coefficients, includes:

[0023] The hyperplane parameters are obtained by combining the ratio of the number of data difference coefficients greater than the average value to the number of data difference coefficients less than or equal to the average value, the variance of the independent sub-parameters, and the average value of the data difference coefficients.

[0024] Furthermore, the method for obtaining the initial abnormal data includes:

[0025] Initial outlier data was filtered out using the LOF algorithm.

[0026] Furthermore, the clustering method used for clustering and classification based on feature values ​​is a hierarchical clustering algorithm.

[0027] The present invention has the following beneficial effects:

[0028] First, the current data to be analyzed provides the foundation for the analysis. Further analysis of local density deviations filters out initial anomalous data, identifying abrupt anomalous data points to facilitate subsequent data interval division. Focusing on different data intervals allows for adjustments to the support vector machine (SVM). Further analysis of data fluctuations and differences between normal data and adjacent initial anomalous data further divides the data intervals, enabling the analysis of the entire local anomaly process and ultimately improving the accuracy of the hyperplane parameters. Next, feature values ​​of the data distribution within each interval are extracted to characterize the anomalous behavior or operating state pattern of that interval, and clustering is performed based on these feature values. Then, the differences in data within different clusters are analyzed to obtain data difference coefficients, reflecting the degree of differentiation within the current data to be analyzed and providing guidance for subsequent hyperplane construction. Based on the differences between data difference coefficients and the feature values, independent sub-parameters for each data interval are obtained from both a global difference perspective and local feature representation, characterizing the individual features of the data interval. Finally, based on the fluctuations of the independent sub-parameters and the distribution of the data difference coefficients, hyperplane parameters are obtained, and real anomalous data is acquired using these parameters. This mitigates the impact of environmental interference and measurement noise on the traditional SVM boundary offset, achieving more accurate identification of real anomalies. Attached Figure Description

[0029] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A flowchart illustrating a circuit operation monitoring method for DC high-voltage contactors in new energy vehicles, provided as an embodiment of the present invention;

[0031] Figure 2 This is a flowchart illustrating a method for obtaining a data range according to an embodiment of the present invention. Detailed Implementation

[0032] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a circuit operation monitoring method for DC high-voltage contactors in new energy vehicles proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0034] The following description, in conjunction with the accompanying drawings, details a specific scheme for a circuit operation monitoring method for DC high-voltage contactors in new energy vehicles provided by the present invention.

[0035] Please see Figure 1 The diagram illustrates a flowchart of a circuit operation monitoring method for DC high-voltage contactors in new energy vehicles, provided by an embodiment of the present invention, specifically including:

[0036] Step S1: Obtain the current data to be analyzed from the DC high-voltage contactor.

[0037] In one embodiment of the present invention, a Hall effect sensor or shunt resistor is used to monitor the current passing through the contactor in real time; a voltage sensor (such as a voltage divider) is used to monitor the voltage across the contactor; a temperature sensor (such as a thermocouple or thermistor) is installed to monitor the temperature data of the contactor; and an embedded system or PLC technology is used to control each sensor to collect data.

[0038] The collected data is transmitted wirelessly or via wired connection to a central monitoring system or cloud platform to enable remote monitoring and data analysis.

[0039] The main faults of DC high-voltage contactors in new energy vehicles are as follows:

[0040] 1) Insulation faults (decreased insulation resistance or abnormal leakage current) can lead to high voltage leakage risk, which may cause electric shock or short circuit in the battery system.

[0041] 2) Abnormal contact resistance (contact oxidation or wear) can cause the contactor to overheat or even weld, resulting in the high-voltage circuit failing to disconnect properly;

[0042] 3) Mechanical failure (contact sticking or delayed action) may cause the vehicle to fail to power on or fail to disconnect power in an emergency;

[0043] Among them, abnormal contact resistance is a common fault. This fault is caused by the rapid formation of an oxide film on the metal surface when exposed to air (such as copper, which can form an oxide layer within 2-3 minutes) or by the adhesion of oil and dust, which increases the film resistance between conductive spots and leads to poor contact; or by the aging of springs, mechanical deformation or assembly errors, which leads to a decrease in the clamping force between contacts, a reduction in the number of contact points and a significant increase in resistance.

[0044] Therefore, by monitoring the current data during the operation of the DC high-voltage contactor and analyzing the changes in the current data, abnormal states during the operation of the DC high-voltage contactor can be obtained. First, the current data of the DC high-voltage contactor to be analyzed is obtained to provide a basis for subsequent analysis.

[0045] Step S2: Filter out initial abnormal data based on the local density deviation of the current data; divide the data intervals based on the data fluctuations and differences between each normal data and adjacent initial abnormal data; obtain feature values ​​based on the distribution of data within each data interval, and perform clustering classification based on the feature values; obtain the data difference coefficient of each cluster based on the differences in data within different clusters.

[0046] Under normal operating conditions, current data changes smoothly with relatively consistent local density. When anomalies occur, current data fluctuates, and different anomalies produce different data fluctuations. For example, when there is an insulation fault, a short circuit will occur, causing an increase in instantaneous current. When there is poor contact, the current will be unstable, fluctuating between large and small values. When a fault occurs, the current value will suddenly change or deviate significantly from neighboring values. Therefore, by screening the initial abnormal data based on the local density deviation of the current data, identifying abrupt abnormal data points, adapting to diverse anomaly patterns, and facilitating subsequent data interval division, focusing on different data intervals, and adjusting the support vector machine.

[0047] Preferably, in one embodiment of the present invention, the initial abnormal data is filtered out by the LOF algorithm, and the remaining current data is normal data.

[0048] In another embodiment of the present invention, other algorithms for analyzing the local density deviation characteristics of data, such as the KNN local distance method, can also be used. These are all existing technologies and will not be described in detail here.

[0049] Considering that different initial abnormal data may correspond to different fault types or environmental disturbances, these abnormal points are not completely isolated points. They often have a "disturbance effect" on neighboring data within a certain range. Normal data may have two adjacent initial abnormal data, and the degree of attribution of the adjacent abnormal points on both sides is different. Therefore, based on the data fluctuations and data differences between each normal data and the adjacent initial abnormal data, the data interval is divided to facilitate the analysis of the entire local abnormal process and ultimately improve the accuracy of the hyperplane parameters.

[0050] Preferably, in one embodiment of the present invention, please refer to Figure 2 The flowchart illustrates a method for obtaining a data range according to an embodiment of the present invention, specifically including:

[0051] Step S201: When there is only one adjacent initial abnormal data in the normal data, divide the data range to the adjacent initial abnormal data.

[0052] When normal data has only one adjacent initial abnormal data, there is no dispute over its attribution, and it is directly assigned to the data range of the adjacent initial abnormal data.

[0053] Step S202: In the normal data where there are two adjacent initial outliers, select normal data one by one as the data to be analyzed; based on the variance between the data to be analyzed and each adjacent initial outlier, and combined with the absolute value of the difference between the two data and the Euclidean distance, obtain the degree of belonging of the data to be analyzed to each adjacent initial outlier.

[0054] First, the data to be analyzed is analyzed one by one. Considering that the smaller the variance between the data to be analyzed and each adjacent initial outlier, it indicates that the data fluctuation is smaller, belongs to the same data change characteristics, and the degree of belonging is greater. The smaller the absolute value of the data difference between the data to be analyzed and each adjacent initial outlier, it indicates that the amplitude difference of the data points is smaller and the degree of belonging is higher. At the same time, the Euclidean distance integrates the amplitude difference of the data and the time interval. The smaller the Euclidean distance, the higher the degree of belonging.

[0055] As an example, the product of the variance between the data to be analyzed and each adjacent initial outlier, the absolute value of the difference between the data to be analyzed and each adjacent initial outlier, and the Euclidean distance between the data to be analyzed and each adjacent initial outlier is used as independent variables. A negative correlation mapping is performed using an exponential function exp(-x) with the natural constant e as the base, where x is the independent variable. The mapping result is used as the degree of belonging of the data to be analyzed to each of its adjacent initial outliers.

[0056] The variance represents the data fluctuation, the absolute value of the difference in amplitude angle, and the Euclidean distance, which combines the two angles of data amplitude difference and time interval, represent the data difference, thus obtaining the two degrees of belonging to the data to be analyzed.

[0057] In other embodiments of the present invention, negative correlation mapping can also be performed by taking the reciprocal; positive correlation can also be performed by adding or weighted summing to fuse the variance, absolute value of the data difference and Euclidean distance between the data to be analyzed and each adjacent initial abnormal data. Only the data values ​​are used in the calculation, and their dimensions are ignored, which will not be elaborated further.

[0058] Step S203: Divide the data intervals according to the degree of belonging to all the data to be analyzed.

[0059] Considering that in the normal data between two adjacent initial outliers, each normal data point corresponds to two membership degrees (one for each of the adjacent initial outliers), the data range of the initial outlier corresponding to the highest membership degree of the normal data is used as the partitioning result.

[0060] Considering that some normal data may exhibit isolation in the time domain, isolated normal data in the time domain of the data interval are further divided into the data interval on the other side.

[0061] For example, in the original partitioning results, the data interval of the 5th initial anomalous data with time series number 800 was {700-701, 730-810, 812-840}, the data interval of the 4th initial anomalous data with time series number 658 was {640-699, 702-729}, and the data interval of the 6th initial anomalous data with time series number 870 was {811, 841-920}. Among them, {700-701, 811} are isolated in the time domain of the data interval. The data intervals of 700-701 and 811 are reassigned to the data interval of the 4th initial anomalous data and 811 are reassigned to the data interval of the 5th initial anomalous data. After the reassignment, the data intervals are {640-729}, {730-840}, and {841-920}.

[0062] Each data interval represents a cluster of similar perturbation features, which may correspond to a certain type of fault impact. By obtaining feature values ​​based on the distribution of data within each data interval, the abnormal behavior or working state pattern of that interval can be effectively characterized, providing a foundation for subsequent hyperplane training and optimization.

[0063] Furthermore, considering that the characteristic values ​​of similar intervals are similar, they can be classified into the same cluster. Therefore, clustering based on characteristic values ​​enhances the overall ability to identify abnormal evolution trends and ultimately improves the performance of the SVM algorithm. Considering that the differences in data within different clusters reflect the degree of differentiation within the current data to be analyzed, providing guidance for subsequent hyperplane construction, the data difference coefficient of each cluster is obtained based on the differences in data within different clusters.

[0064] Preferably, in one embodiment of the present invention, considering that the maximum value of the data within a data interval represents the maximum abnormal performance of the data, the variance of the data reflects the severity of the fluctuation of the data, and the number of data represents the duration, the distribution of the data within the data interval is represented by combining these three perspectives. Therefore, the maximum value, variance and number of data within each data interval are fused to obtain the feature value of each data interval.

[0065] As an example, the product of the maximum value, variance, and number of data points within each data interval is used as the feature value for each data interval.

[0066] Furthermore, a hierarchical clustering algorithm is used to cluster and classify the feature values.

[0067] In other embodiments of the present invention, the implementer may also fuse the maximum value, variance and quantity of data within the data interval by positive correlation methods such as addition or weighted summation, which will not be elaborated further.

[0068] Preferably, in one embodiment of the present invention, considering that the larger the absolute average deviation of the feature values ​​within a cluster, the greater the data difference within the cluster; the greater the difference between the average value of the feature values ​​within a cluster and the average value of all feature values, the greater the difference between the current data corresponding to the current cluster and the current data of other time periods, and the higher the data difference coefficient.

[0069] Based on this, the data difference coefficient of each cluster is obtained by combining the difference between the average value of the feature values ​​within each cluster and the average value of all feature values, and the absolute average deviation of the feature values ​​within the cluster.

[0070] As an example, the differences between data can be represented by the absolute value of the difference. The product of the absolute value of the difference between the mean of the feature values ​​within each cluster and the mean of all feature values, and the corresponding absolute mean deviation, is used as the data difference coefficient for each cluster.

[0071] Step S3: Based on the differences between the data difference coefficients and the feature values, obtain the independent sub-parameters for each data interval; based on the fluctuations of the independent sub-parameters and the distribution of the data difference coefficients, obtain the hyperplane parameters and use them to obtain the real outlier data.

[0072] When obtaining the hyperplane, the main focus is on the differences in the original sample data. The greater the differences between the sample data, the easier it is to select the hyperplane; conversely, the smaller the differences between the samples, the more difficult it is to select the hyperplane. The differences between the data difference coefficients reflect the differences within the current data. Combined with the eigenvalues, the individual characteristics of each data interval are characterized from both the global difference perspective and the local feature representation. Therefore, based on the differences between the data difference coefficients and the eigenvalues, independent sub-parameters for each data interval are obtained, representing the abnormal independence or representativeness of the data in each interval within the current overall current data, thus preparing for the subsequent construction of a highly robust hyperplane.

[0073] Preferably, in one embodiment of the present invention, the ratio of the range of the data difference coefficients to the mean of the data difference coefficients is taken into account, which effectively measures the relative degree of difference between the data difference coefficients, avoids the influence of the order of magnitude of the data difference coefficients, and facilitates fusion with feature values.

[0074] Based on this, the ratio of the range to the mean of the data difference coefficient is fused with the feature value of each data interval to obtain the independent sub-parameters of each data interval.

[0075] As an example, the ratio of the range to the mean of the data difference coefficient is multiplied by the feature value of each data interval, and the product is used as an independent sub-parameter for each data interval.

[0076] The fluctuation of independent sub-parameters reflects the sparsity and dispersion of the distribution of abnormal features in each interval in the data; the distribution pattern of the data difference coefficient provides the distribution tendency of the overall phenomenon. Therefore, based on the fluctuation of independent sub-parameters and the distribution of data difference coefficient, the hyperplane parameters are obtained and real abnormal data are obtained. This helps to resist the influence of environmental interference, measurement noise and other factors on the boundary offset of traditional SVM, achieve more accurate identification of real abnormal points, and avoid misjudgment and missed judgment in the initial anomaly screening.

[0077] Preferably, in one embodiment of the present invention, considering that the variance of the data reflects the dispersion of the "characteristic representativeness" of all data intervals, the larger the variance, the more significant the performance difference between each interval, and the stronger the structural boundary, the higher the hyperplane parameter can be; the distribution of the number of data on both sides of the average value of the data difference coefficient reflects the distribution ratio of high difference clusters and low difference clusters, reflects the trend of "abnormal dominance", and can be used to guide the tightness of the boundary; the average value of the data difference coefficient itself represents the overall abnormality of the cluster and is the reference center for hyperplane positioning;

[0078] Based on this, the hyperplane parameters are obtained by combining the variance of the independent sub-parameters with the distribution of the number of data on both sides of the average value of the data difference coefficients, as well as the average value of the data difference coefficients.

[0079] Furthermore, considering that the number of high-discrepancy clusters is greater than the number of low-discrepancy clusters, there are a large number of significant differences in the data. Therefore, the anomaly identification boundary should be widened and the hyperplane parameter should be larger. Thus, the hyperplane parameter is obtained by fusing the ratio of the number of data difference coefficients greater than the mean to the number of data difference coefficients less than or equal to the mean, the variance of independent sub-parameters, and the mean of data difference coefficients.

[0080] As an example, the product of the number of data difference coefficients greater than the mean and the number of data difference coefficients less than or equal to the mean, the variance of the independent sub-parameters, and the mean of the data difference coefficients is used as the hyperplane parameter.

[0081] The obtained hyperplane parameters represent the slope of the hyperplane, and their function is to accurately distinguish data with greater anomalies when classifying data.

[0082] Furthermore, based on the obtained hyperplane parameters, the current data during the operation of the DC high-voltage contactor is classified using the support vector machine algorithm to obtain the true abnormal data in the current data; the detection of abnormal data in the data by the support vector machine algorithm is a well-known existing technology and will not be elaborated here.

[0083] In another embodiment of the present invention, after obtaining the abnormal current data (real abnormal data), the abnormal current data is also transmitted to the abnormal analysis unit or relevant personnel to determine the fault in the DC high voltage contactor of the new energy vehicle and ensure the safe operation of the circuit.

[0084] In summary, to address the technical problem that interference with current sensor data makes it difficult for Support Vector Machine (SVM) algorithms to accurately find the optimal classification hyperplane, thus affecting the accuracy of abnormal current identification, this invention proposes a circuit operation monitoring method for DC high-voltage contactors in new energy vehicles. This invention first acquires the current data to be analyzed; further, it analyzes local density deviations to filter out initial abnormal data; then, based on data fluctuations and differences between normal data and adjacent initial abnormal data, it divides the data into intervals; further, it extracts feature values ​​of the data distribution within each interval and performs clustering based on these feature values, analyzing the differences in data within different clusters, and combining the feature values ​​to obtain independent sub-parameters for each data interval; finally, based on the fluctuations of the independent sub-parameters and the distribution of the data difference coefficients, it obtains the hyperplane parameters and uses these parameters to acquire the true abnormal data. Addressing the problem of SVM's inaccurate classification due to current sensor interference, this invention uses density deviation to identify initial anomalies, interval division, feature extraction, and clustering, quantifying data differences and optimizing hyperplane parameters to achieve accurate identification of abnormal currents in high-voltage contactors, improving classification robustness and accuracy.

[0085] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0086] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A circuit operation monitoring method for a new energy vehicle direct current high voltage contactor, characterized in that, The method comprises: Obtaining current data of a DC high-voltage contactor to be analyzed; Screening initial abnormal data according to local density deviation of the current data; dividing data intervals according to data fluctuation and data difference between each normal data and adjacent initial abnormal data; obtaining characteristic values according to distribution of data in each data interval, and performing clustering classification based on the characteristic values; obtaining data difference coefficients of each clustering cluster according to difference of data in different clustering clusters; Obtaining independent sub-parameters of each data interval according to difference between the data difference coefficients, in combination with the characteristic values; obtaining hyperplane parameters according to fluctuation of the independent sub-parameters, in combination with distribution of the data difference coefficients, and obtaining real abnormal data based on the hyperplane parameters. The method for obtaining the data intervals comprises: When the normal data has only one adjacent initial abnormal data, the data interval of the adjacent initial abnormal data is divided; In normal data having two adjacent initial abnormal data, normal data are selected one by one as data to be analyzed; the attribution degrees of the data to be analyzed and each adjacent initial abnormal data are obtained according to variance of data between the data to be analyzed and each adjacent initial abnormal data, in combination with absolute value of difference between the two data and Euclidean distance; Data intervals are divided according to the attribution degrees of all the data to be analyzed.

2. The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 1, characterized in that, The method for dividing data intervals according to the attribution degrees of all the data to be analyzed comprises: In normal data between two adjacent initial abnormal data, the data interval of the initial abnormal data corresponding to the maximum attribution degree of the normal data is taken as a division result; and the normal data isolated in the time domain of the data interval are divided into the data interval on the other side.

3. The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 1, characterized in that, The method for obtaining the characteristic values comprises: The characteristic values of each data interval are obtained by fusing maximum value, variance and quantity of data in each data interval.

4. The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 1, characterized in that, The method for obtaining the data difference coefficients comprises: The data difference coefficients of each clustering cluster are obtained according to difference between average value of the characteristic values in each clustering cluster and average value of all the characteristic values, in combination with absolute average deviation of the characteristic values in the clustering cluster.

5. The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 1, characterized in that, The method for obtaining the independent sub-parameters comprises: The independent sub-parameters of each data interval are obtained by fusing ratio of range and mean value of the data difference coefficients, and the characteristic values of each data interval.

6. The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 1, characterized in that, The method for obtaining the hyperplane parameters comprises: The hyperplane parameters are obtained according to variance of the independent sub-parameters, in combination with data quantity distribution on both sides of average value of the data difference coefficients, and the average value of the data difference coefficients.

7. The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 6, characterized in that, The method for obtaining the hyperplane parameters according to variance of the independent sub-parameters, in combination with data quantity distribution on both sides of average value of the data difference coefficients, and the average value of the data difference coefficients comprises: The hyperplane parameters are obtained by fusing quantity of the data difference coefficients greater than the average value, ratio of the data difference coefficients less than or equal to the average value, variance of the independent sub-parameters and the average value of the data difference coefficients. 8.The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 1, wherein, The method for obtaining the initial abnormal data comprises: The initial abnormal data is screened out by the LOF algorithm. 9.The circuit operation monitoring method for a new energy vehicle direct current high voltage contactor according to claim 1, wherein, The clustering method used in the clustering classification based on the feature values is a hierarchical clustering algorithm.

Images