Direct current transmission project stray current event identification method considering influence of multiple working conditions

Abstract

The invention provides a direct-current transmission project stray current event identification method considering multi-working-condition influence, and the method comprises the following steps: A, obtaining related data of ground stray current, carrying out the preprocessing, and outputting a smoothed data sequence; b, according to the smoothed data sequence output in the step A, constructing a catastrophe point detection model considering an adaptive threshold value, detecting catastrophe points of the ground stray current and determining a catastrophe interval of the ground stray current; and C, extracting ground stray current data according to the ground stray current abrupt change interval obtained in the step B, clustering the data of the abrupt change interval through down-sampling K-Means, and judging the type of the stray current of the direct current transmission project. According to the method, the mutation points are accurately detected through the self-adaptive threshold model; by means of down-sampling K-Means, the clustering accuracy is guaranteed, and meanwhile the operation speed is increased; and the stray current type is judged in combination with a clustering result, and a data basis is provided for direct-current transmission project stray current treatment.

Description

Technical Field

[0001] This invention relates to the field of stray current technology, specifically a method for identifying stray current events in DC transmission projects that takes into account the influence of multiple operating conditions. Background Technology

[0002] Stray currents are generated by unipolar operation of ultra-high voltage and extra-high voltage direct current (UHVDC) projects and geomagnetic storms. These stray currents flow through urban utility tunnels, energy utility tunnels, and power grids, eventually flowing into transformers and causing DC bias magnetization, leading to transformer vibration. This not only affects transformer lifespan and jeopardizes the safe operation of the main power grid, but also causes noise pollution, leading to complaints from nearby residents and damaging the power grid company's public image. The types and characteristics of stray currents generated by DC transmission projects, urban rail transit, and geomagnetic storms differ. DC transmission projects can operate under unipolar grounding or asymmetrical conditions; therefore, it is necessary to determine the specific operating condition of the DC transmission project to manage stray currents.

[0003] Chinese invention patent application number 202510814834.4, entitled "A Time-Series-Based Landslide Remote Sensing Monitoring Method, System, and Device," provides a time-series-based landslide remote sensing monitoring method, system, and device. The method includes: acquiring a long-term multispectral remote sensing dataset of a target area, preprocessing it, and filtering effective pixels; calculating the Normalized Difference Vegetation Index (NDVI) of all effective pixels within the target area to construct an annual NDVI dataset for the target area; calculating the median NDVI for each pixel within the year to construct a time-series NDVI curve for each pixel; detecting abrupt change points for each pixel using linear regression, calculating and comparing the residuals of the NDVI values ​​to obtain the landslide monitoring result for each pixel; and correcting the landslide monitoring results to obtain the final landslide monitoring result for the target area. This invention can quickly identify the spatiotemporal distribution characteristics of landslides that have already occurred within a target area, which is of great significance for mastering landslide disaster monitoring and risk assessment in key areas and significantly improves the efficiency of landslide disaster information extraction. Chinese invention patent application number 202510761795.6, entitled "Method, System, Device, and Storage Medium for Determining Electric Bicycle Charging at Home," discloses a method, system, device, and storage medium for determining whether an electric bicycle is being charged at home. Belonging to the field of electric bicycle charging, the method includes the following steps: constructing an energy consumption characteristic function of an electric bicycle in a charging state; obtaining the total household electricity consumption and generating a total electricity consumption data sequence based on the total household electricity consumption; detecting abrupt changes in the total electricity consumption data sequence and removing energy consumption abrupt changes not caused by electric bicycle electricity consumption; extracting energy consumption features from the removed total electricity consumption data sequence; comparing the extracted energy consumption features with the energy consumption characteristic function, and determining whether an electric bicycle is being charged at home based on the comparison result. This method can quickly and accurately determine whether an electric bicycle is being charged at home.

[0004] The method proposed in Chinese invention patent application number 202510814834.4 mainly targets the detection of abrupt changes in image data, while the method proposed in Chinese invention patent application number 202510761795.6 targets electricity consumption data. Neither of these methods is applicable to the detection of abrupt changes in stray currents in DC transmission projects. Summary of the Invention

[0005] To address the aforementioned technical problems in existing technologies, this invention provides a stray current event identification method for DC transmission projects that considers the influence of multiple operating conditions. By constructing a mutation point detection model that considers adaptive thresholds, it can accurately detect ground stray current mutation points. Through downsampling K-Means clustering, it improves the clustering speed while ensuring the accuracy of the results. By analyzing the clustering results, it can determine the type of stray current in DC transmission projects, providing data support for the management of ground stray currents in DC transmission projects.

[0006] The technical solution adopted in this invention is as follows:

[0007] A method for identifying stray current events in DC transmission projects that considers the effects of multiple operating conditions includes the following steps:

[0008] A. Acquire relevant data on stray ground currents and preprocess them to output a smoothed data sequence;

[0009] B. Based on the smoothed data sequence output in step A, construct a mutation point detection model that considers adaptive thresholds to detect mutation points of stray ground currents and determine the mutation intervals of stray ground currents.

[0010] C. Based on the abrupt change intervals of stray ground currents obtained in step B, extract the stray ground current data, and cluster the data in the abrupt change intervals using downsampled K-Means to determine the stray current type of the DC transmission project.

[0011] Furthermore, step A specifically includes:

[0012] Acquire stray current data of the ground, denoted as Where N is the total number of monitoring points, This is the monitoring data of the i-th point; the stray current data of the ground is preprocessed, and the stray current of the ground is smoothed by the third order using the Savitzky-Golay filtering method; the filter length is set to... ,in If the number is odd, the order of the fitted polynomial is . For each sampling point i, in its neighborhood window Constructing polynomials ,in The relative position index with respect to the center of the window has a range of values. , The coefficients of the polynomial to be solved are obtained by using the least squares method, satisfying... The obtained coefficients are substituted into the polynomial, and the fitted value at the center point of the window is calculated as the smoothed output at that point. The sliding window traverses the entire data sequence to obtain the complete smoothed data sequence. .

[0013] Furthermore, step B specifically includes:

[0014] 1) For the smoothed data sequence The stray current gradient is calculated using the central difference method. , The first and last data points of the stray ground current sequence are set to zero.

[0015] 2) Regarding stray current gradient Non-maximum suppression is performed, specifically if... and and ,in This is the threshold for the stray current gradient. After non-maximum suppression, the stray current gradient is: ;

[0016] 3) Calculate significant stray current gradients The specific calculation method is as follows: ;

[0017] 4) Calculate the adaptive upper and lower thresholds using significant stray current gradients. The percentage is derived from the upper and lower limits of the adaptive confirmation threshold, which will significantly reduce stray current gradients. The H% quantile is used as the upper threshold and the L% quantile as the lower limit, where the value of H ranges from [50, 90] and the value of L ranges from [10, 75]. Different adaptive upper and lower thresholds are obtained by permuting and combining H and L.

[0018] 5) Based on adaptive upper and lower thresholds, data points greater than the upper threshold are marked as strong mutation points, and data points greater than the lower threshold but less than the upper threshold are marked as weak mutation points. A dual threshold connection mechanism is adopted, using strong edge points as seeds, and connecting adjacent weak edge points into continuous event segments through iteration, so as to achieve complete boundary extraction of stray current mutation process.

[0019] 6) Sort the stray current abrupt change points in ascending order. For each abrupt change point i, if the distance between a subsequent abrupt change point b and abrupt change point i is less than min_distance, and the gradient of b is greater than the gradient of abrupt change point i, then replace abrupt change point i with abrupt change point b and delete abrupt change point i. This finalizes the set of abrupt change points. Where min_distance is the defined minimum distance between mutation points;

[0020] 7) Record the set of abrupt changes for different adaptive threshold upper and lower limits. The most frequent abrupt change point in the set of all abrupt change points is identified, and the abrupt change interval of the stray current is calculated based on this interval, denoted as . , It is the time corresponding to the lower boundary of the mutation interval. It is the time corresponding to the upper boundary of the mutation interval.

[0021] Furthermore, step C specifically includes:

[0022] 1) For the abrupt change in stray current range The data in the sample is downsampled at a rate of 100%. ;

[0023] 2) Perform K-Means cluster analysis on the stray current abrupt change interval data after downsampling, calculate the silhouette coefficient of the clusters, and take the number of clusters with the largest silhouette coefficient as the optimal number of clusters;

[0024] 3) Using different downsampling rates, the optimal number of clusters under different downsampling rates is calculated, and the frequency of the optimal number of clusters under different downsampling rates is counted. The highest frequency number of clusters is the number of types of stray currents in the abrupt change range of ground stray currents.

[0025] 4) If the highest frequency cluster number is ≥2, then it is determined that there is a DC asymmetric operation condition.

[0026] The present invention provides a stray current event identification method for DC transmission projects that considers the influence of multiple operating conditions. By constructing a mutation point detection model that considers adaptive thresholds, it can accurately detect ground stray current mutation points. By using downsampling K-Means clustering, it can improve the clustering speed while ensuring the accuracy of the results. By analyzing the clustering results, it can determine the type of stray current in DC transmission projects, and provide data basis for the management of ground stray current in DC transmission projects. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of one embodiment of the stray current event identification method for DC transmission projects that takes into account the influence of multiple operating conditions according to the present invention;

[0028] Figure 2 This is a schematic diagram of the smoothed curve according to an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram of the mutation interval determined in an embodiment of the present invention;

[0030] Figure 4 This is a schematic diagram illustrating the relationship between the number of clusters and the silhouette coefficient in an embodiment of the present invention;

[0031] Figure 5 This is a schematic diagram illustrating the results of clustering data in mutation intervals using downsampling K-Means in an embodiment of the present invention. Detailed Implementation

[0032] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings.

[0033] Please see Figure 1 This is a schematic diagram of one embodiment of a stray current event identification method for DC transmission projects considering the influence of multiple operating conditions according to the present invention. The method includes the following steps:

[0034] Step A: Obtain relevant data on stray ground currents and preprocess the data, specifically as follows:

[0035] Acquire stray current data of the ground, denoted as Where N is the total number of monitoring points, This is the monitoring data of the i-th point; the stray current data of the ground is preprocessed, and the stray current of the ground is smoothed by the third order using the Savitzky-Golay filtering method; the filter length is set to... ,in If the number is odd, the order of the fitted polynomial is . For each sampling point i, in its neighborhood window Constructing polynomials ,in The relative position index with respect to the center of the window (value range is...) ), The coefficients of the polynomial to be solved can be obtained using the least squares method, satisfying the following conditions: The obtained coefficients are substituted into the polynomial, and the fitted value at the center point of the window is calculated as the smoothed output at that point. The sliding window traverses the entire data sequence to obtain the complete smoothed data sequence. ;

[0036] Step B: Based on the preprocessed stray ground current data from Step A, construct a mutation point detection model considering adaptive thresholds to detect mutation points in the stray ground current and determine mutation intervals. Specifically:

[0037] 1) For the smoothed data sequence The stray current gradient is calculated using the central difference method. , The first and last data points of the stray ground current sequence are set to zero.

[0038] 2) Regarding stray current gradient Non-maximum suppression is performed, specifically if... and and ,in This is the threshold for the stray current gradient. After non-maximum suppression, the stray current gradient is: ;

[0039] 3) Calculate significant stray current gradients The specific calculation method is as follows: ;

[0040] 4) Calculate the adaptive upper and lower thresholds using significant stray current gradients. The percentage is derived from the upper and lower limits of the adaptive confirmation threshold, which will significantly reduce stray current gradients. The H% quantile is used as the upper limit of the threshold, and the L% quantile is used as the lower limit. The value range of H is [50, 90], and the value range of L is [10, 75]. By permuting and combining H and L, different adaptive upper and lower limits of the threshold can be obtained.

[0041] 5) Based on adaptive upper and lower thresholds, data points greater than the upper threshold are marked as strong mutation points, and data points greater than the lower threshold but less than the upper threshold are marked as weak mutation points. A dual threshold connection mechanism is adopted, using strong edge points as seeds, and connecting adjacent weak edge points into continuous event segments through iteration, so as to achieve complete boundary extraction of stray current mutation process.

[0042] 6) Sort the stray current abrupt change points in ascending order. For each sorted abrupt change point i, if the distance between a subsequent abrupt change point b and abrupt change point i is less than min_distance, and the gradient of b is greater than the gradient of abrupt change point i, then replace abrupt change point i with abrupt change point b and delete abrupt change point i. The final set of abrupt change points is determined according to this method. Where min_distance is the defined minimum distance between mutation points;

[0043] 7) Record the set of abrupt changes for different adaptive threshold upper and lower limits. The most frequent abrupt change point in the set of all abrupt change points is identified, and the abrupt change interval of the stray current is calculated based on this interval, denoted as . ;

[0044] Step C: Based on the abrupt change intervals of stray ground current obtained in Step B, extract stray ground current data, and cluster the data in the abrupt change intervals using downsampled K-Means to determine the stray current type of the DC transmission project. Specifically:

[0045] 1) For the abrupt change in stray current range The data in the sample is downsampled at a rate of 100%. ;

[0046] 2) Perform K-Means clustering analysis on the stray current abrupt change interval data after downsampling, calculate the silhouette coefficient of the clusters, and take the number of clusters with the largest silhouette coefficient as the optimal number of clusters;

[0047] 3) Using different downsampling rates, the optimal number of clusters under different downsampling rates is determined, and the number of clusters with the most occurrences of the optimal number of clusters is counted. This number of clusters is the number of types of stray currents in the abrupt change range of ground stray currents.

[0048] 4) If there are two or more types of stray current in the stray current abrupt change interval, it indicates that the operating conditions of the DC transmission project are changing in the abrupt change interval, and there is a DC asymmetrical operation.

[0049] The technical solution and effects of the present invention will be described in detail below with a specific embodiment:

[0050] In step A, relevant data on stray ground currents are acquired, and the data is preprocessed, such as... Figure 2 The image shown is an example of a smoothed curve;

[0051] In step B, based on the preprocessed stray ground current data from step A, a mutation point detection model considering adaptive thresholds is constructed to detect mutation points in the stray ground current and determine mutation intervals; for example... Figure 3 The diagram shows the defined mutation intervals;

[0052] In step C, based on the abrupt change intervals of stray ground currents obtained in step B, stray ground current data is extracted. The data within these abrupt change intervals is then clustered using downsampled K-Means to determine the stray current type for the DC transmission project. Figure 4 As shown, the silhouette coefficient is largest when the cluster number is 2. Therefore, there are two types of stray currents in the abrupt change region, such as... Figure 5 The results shown are from clustering, which indicates that the operating conditions of the DC transmission project are changing and that there is DC asymmetric operation.

[0053] The present invention provides a stray current event identification method for DC transmission projects that considers the influence of multiple operating conditions. By constructing a mutation point detection model that considers adaptive thresholds, it can accurately detect ground stray current mutation points. By using downsampling K-Means clustering, it can improve the clustering speed while ensuring the accuracy of the results. By analyzing the clustering results, it can determine the type of stray current in DC transmission projects, and provide data basis for the management of ground stray current in DC transmission projects.

[0054] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying HVDC project stray current events considering multi-condition effects, characterized in that The method comprises the following steps: A, obtaining relevant data of earth's stray current and preprocessing, outputting smoothed data sequence; B, according to the smoothed data sequence outputted in step A, constructing a mutation point detection model considering adaptive threshold, detecting the mutation point of earth's stray current and determining the mutation interval of earth's stray current; C, according to the mutation interval of earth's stray current obtained in step B, extracting earth's stray current data, clustering the data in the mutation interval through K-Means with down-sampling, and judging the type of stray current of the direct current power transmission project.

2. The method of claim 1, wherein the method is characterized in that: Step A specifically comprises: Acquire stray current data of the ground, denoted as Where N is the total number of monitoring points, This is the monitoring data of the i-th point; the stray current data of the ground is preprocessed, and the stray current of the ground is smoothed by the third order using the Savitzky-Golay filtering method; the filter length is set to... ,in If the number is odd, the order of the fitted polynomial is . For each sampling point i, in its neighborhood window Constructing polynomials ,in The relative position index with respect to the center of the window has a range of values. , The coefficients of the polynomial to be solved are obtained by using the least squares method, satisfying... The obtained coefficients are substituted into the polynomial, and the fitted value at the center point of the window is calculated as the smoothed output at that point. The sliding window traverses the entire data sequence to obtain the complete smoothed data sequence. .

3. The method of claim 2, wherein the method further comprises: Step B specifically comprises: 1) For the smoothed data sequence , the gradient of the earth's stray current is calculated by using the central difference method , , the first data point and the last data point of the earth's stray current sequence are set to zero. 2) stray current gradient Non-maximum suppression is performed, and the specific processing method is that if and and wherein is a threshold value of the stray current gradient, and the stray current gradient after the non-maximum suppression is ; 3) calculating the significant stray current gradient , in particular by ; 4) calculating adaptive upper and lower thresholds by the percentage of significant stray current gradient Adaptive upper and lower thresholds are confirmed by the H% quantile and L% quantile of the significant stray current gradient, where H ranges from 50 to 90 and L ranges from 10 to 75. Arranging H and L in different combinations, different adaptive upper and lower thresholds are obtained. ​ 5) based on the adaptive upper and lower thresholds, marking the data points greater than the upper threshold as strong mutation points, and marking the data points greater than the lower threshold but less than the upper threshold as weak mutation points; adopting a double-threshold connection mechanism, taking the strong edge points as seeds, and connecting the adjacent weak edge points into continuous event segments through iteration to realize complete boundary extraction of the mutation process of the stray current; 6) ascendingly order the stray current mutation points, for the ordered mutation point i, if the distance between the subsequent mutation point b and the mutation point i is less than min_distance, and the gradient corresponding to the mutation point b is greater than the gradient of the mutation point i, then replace the mutation point i with the mutation point b, delete the mutation point i, and finally determine the set of mutation points ; wherein min_distance is the minimum distance between the defined mutation points; 7) record the mutation point set of different adaptive threshold upper and lower limits , the most frequent mutation point in all mutation point sets is counted, and the mutation interval of the stray current is calculated based on this, denoted as , is the time corresponding to the lower boundary of the mutation interval, is the time corresponding to the upper boundary of the mutation interval.

4. The method of claim 3, wherein the method further comprises: Step C specifically comprises: 1) down-sampling data in the stray current abrupt change interval at a down-sampling rate of ; 2) performing K-Means clustering analysis on the down-sampled stray current mutation interval data, calculating the silhouette coefficient of clustering, and taking the number of clusters with the maximum silhouette coefficient as the optimal number of clusters; 3) adopting different down-sampling rates, respectively calculating the optimal number of clusters under different down-sampling rates, counting the frequency of the optimal number of clusters under different down-sampling rates, and taking the highest frequency cluster number as the type number of the stray current in the mutation interval of earth's stray current; 4) if the highest frequency cluster number is greater than or equal to 2, it is determined that there is a direct current asymmetric operation condition.

Images