Method for identifying waveform anomaly of high-voltage direct-current capacitive load system based on template matching

Abstract

The invention relates to the technical field of image processing, in particular to a high-voltage direct-current capacitive load system waveform anomaly identification method based on template matching, and the method comprises the steps: obtaining a multi-scale waveform of a voltage signal, carrying out the preliminary matching of a to-be-identified waveform and a template waveform under a coarse scale, and judging the overall trend. Then, under the mesoscale, waveform segmentation is carried out, abnormal segments of the waveform to be identified are screened out, and sampling points corresponding to the abnormal segments and sampling points corresponding to the template waveform are respectively mapped and marked under the original scale; and finally, performing dynamic self-adaptive time warping matching according to the to-be-identified waveform under the original scale and the marked sampling points in the template waveform, and judging whether the waveform is abnormal or not according to a matching result. According to the method, local distortion interference in the waveform matching process is avoided, and whether the original to-be-identified waveform is abnormal or not can be determined according to an accurate matching result. The method improves the accuracy of waveform anomaly recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing. Specifically, it relates to a method for abnormal waveform recognition of a high-voltage DC capacitive load system based on template matching. Background Art

[0002] High-voltage DC capacitive load systems are usually used in important scenarios such as power transmission, industrial power supply, and new energy access. Their operating status is directly related to the safety and stability of the power supply system. The voltage and current waveform signals generated by high-voltage DC capacitive load systems often exhibit complex dynamic characteristics of non-stationarity, multi-scale, and superposition of multi-frequency components due to multiple factors such as the charge and discharge characteristics of the load, equipment aging, electromagnetic interference, and switching operations. These waveforms, as visual representations of time-series data, not only contain the overall picture of system operation but also hide fault trends. Accurate abnormal recognition of such waveforms is the key to realizing system health assessment, fault warning, and intelligent operation and maintenance.

[0003] In recent years, with the development of image processing and pattern recognition technologies, some studies have proposed converting waveform signals into structured images and performing abnormal recognition through template matching. This type of method usually uses a preset standard waveform as the template waveform and performs image-level matching and comparison between the real-time collected waveform to be recognized and the template waveform to analyze whether there are waveform abnormalities.

[0004] The dynamic time warping algorithm is one of the most commonly used alignment algorithms in existing template matching technologies. However, in high-voltage DC capacitive load systems, waveform signals are often affected by local disturbances such as spikes and glitches, which are prone to local distortion. Traditional dynamic time warping algorithms are highly sensitive to such local distortions, easily leading to anchoring deviations in the matching path, causing the overall minimum cost path to be distorted or lagged, and thus resulting in false matching and misjudgment in recognition, reducing the accuracy of waveform abnormality detection.

[0005] Therefore, aiming at the problem of local disturbances existing in the waveforms of high-voltage DC capacitive load systems, it is urgent to design a template matching method with global structure perception ability and precision, which can effectively suppress the interference of local distortions on the matching process, and thus improve the accuracy and stability of waveform abnormality recognition. Summary of the Invention

[0006] To solve the problem that traditional template matching methods are sensitive to local distortions in the waveform abnormality recognition of high-voltage DC capacitive load systems, easily causing false matching and affecting the accuracy of waveform abnormality recognition, the present invention proposes a method for abnormal waveform recognition of a high-voltage DC capacitive load system based on template matching. This method includes: Construct the original waveform to be recognized and the original template waveform of the voltage signal of the high-voltage DC capacitive load system, and obtain the waveforms to be recognized and the template waveforms at medium scale and coarse scale through multi-scale downsampling; Perform a preliminary match between the rough-scale waveform to be recognized and the template waveform to determine whether the overall waveform trends are consistent; if not, determine that the original waveform to be recognized is abnormal; if so, map and mark the corresponding sampling points of the rough-scale waveform to be recognized and the template waveform to the medium-scale waveform to be recognized and the template waveform. In the medium-scale waveform to be recognized and the template waveform, each segment is between two adjacent marked sampling points, and screen the abnormal segments of the medium-scale waveform to be recognized, map and mark the corresponding sampling points of the abnormal segments to the original waveform to be recognized, and at the same time, map and mark the sampling points corresponding to the abnormal segments in the medium-scale template waveform to the original template waveform. Perform dynamic time warping matching according to the marked sampling points in the original waveform to be recognized and the marked sampling points in the original template waveform, and determine whether the original waveform to be recognized is abnormal in response to the judgment result of whether the error of the matching result exceeds the normal error range.

[0007] This technical solution first realizes the extraction of waveform features at different frequencies and different structural scales from the perspective of signal processing by constructing the original waveform and introducing a multi-scale downsampling strategy, which can effectively separate high-frequency noise and low-frequency trends, making the subsequent analysis more hierarchical. Then, by performing a preliminary match on the rough-scale waveform, it can quickly judge whether the overall waveform trends are consistent to determine whether there are obvious trend abnormalities in the waveform to be recognized. At the same time, it propagates the structural features of the overall trend to the medium-scale waveform, realizes the transfer of structural anchor points, and maintains the logical continuity of the matching. Subsequently, a segmented anomaly recognition mechanism is introduced in the medium-scale waveform, and the anomaly segment features are mapped to the original-scale waveform, so that the dynamic time warping is only executed in the abnormal area. Through this waveform matching path convergence mechanism guided by a multi-level structure, the probability of the matching path being interfered by local distortions is reduced, thereby improving the accuracy of subsequent waveform anomaly recognition.

[0008] Further, the dynamic time warping matching includes: constructing a direct distance matrix according to the direct distance between the marked sampling points in the original waveform to be recognized and the marked sampling points in the original template waveform; on the basis of the direct distance matrix, using the dynamic programming method to calculate the cumulative distance matrix, starting from the lower right corner of the cumulative distance matrix and performing path backtracking to the upper left corner to obtain the minimum cumulative cost path to complete the dynamic time warping matching; each element of the cumulative distance matrix corresponds to a combination of marked sampling points, and the value of the element is the cumulative distance of the combination of marked sampling points; the combination of marked sampling points contains two marked sampling points from the original waveform to be recognized and the original template waveform respectively; the minimum cumulative cost path is composed of multiple path points, and each path point corresponds to an element.

[0009] Further, the method for screening abnormal segments of the medium-scale waveform to be recognized is as follows: Obtain the corresponding segments of each segment of the medium-scale waveform to be recognized in the medium-scale template waveform; Determine the degree of abnormality of each segment according to the ratio of the cumulative energy of each segment and its corresponding segment and the phase consistency; Obtain the degrees of abnormality of all segments of the medium-scale waveform to be recognized, and use the quartile method to take the segments with the degree of abnormality greater than the third quartile as the abnormal segments in the medium-scale waveform to be recognized.

[0010] This technical solution measures the degree of energy deviation of each segment through the ratio of the cumulative energy of each segment and its corresponding segment, and can reflect whether there are abnormal situations such as signal attenuation or amplification in this segment. Introducing phase consistency as an auxiliary index can further analyze the phase synchronization between segments, ensuring that potential abnormalities can be identified when the waveform phase changes. This method comprehensively considers energy deviation and phase synchronization, making the identification of abnormal segments more comprehensive.

[0011] Further, the method for determining whether there is an abnormality in the original waveform to be recognized is as follows: If the error of the matching result exceeds the normal error range, it is determined that the original waveform to be recognized has an abnormality; if the error of the matching result does not exceed the normal error range, it is determined that the original waveform to be recognized has no abnormality.

[0012] Further, in the process of calculating the cumulative distance matrix by using the dynamic programming method, it also includes introducing a distance weight coefficient to reconstruct the recurrence formula of the cumulative distance as:

[0013] In the formula, and are the direct distance and cumulative distance of the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform respectively, is the distance weight coefficient of the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform, is to take the minimum value.

[0014] This technical solution effectively adjusts the matching process of the dynamic time warping algorithm by introducing a distance weight coefficient into the recurrence formula of the cumulative distance, making the path matching more flexible and accurate. Specifically, the cumulative distance coefficient can reflect the importance of the waveform carried by each sampling point in the matching process. During the matching process, sampling points with large fluctuations or abnormalities are given higher weights, which helps to suppress the influence of noise or unimportant disturbances on the matching path, ensuring that the algorithm can still preferentially select the matching path that best conforms to the global trend when local distortions exist.

[0015] Further, the abnormality degree of each segment is determined based on the following formula: ; In this formula, is the abnormality degree of the th segment of the mesoscale waveform to be recognized, is the ratio of the cumulative energy of the th segment of the mesoscale waveform to be recognized to the th segment of the mesoscale template waveform, is the phase consistency of the th segment of the mesoscale waveform to be recognized and the th segment of the mesoscale template waveform, is the anti-0 parameter.

[0016] Further, the distance weight coefficient is determined based on the following formula: ; In the formula, is the distance weight coefficient between the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform, is the normalized value of the local signal variance of the th marked sampling point of the original waveform to be recognized, is the normalized value of the local signal variance of the th marked sampling point of the original template waveform and the th marked sampling point.

[0017] This technical solution realizes the dynamic quantitative evaluation of the importance of the corresponding sampling points in the waveform to be recognized and the template waveform by introducing the comprehensive weight based on the normalized value of the local signal variance, quantifies the local waveform of the waveform through the local variance, and forms a confidence measure in the sense of probability by combining the product of the normalized values of the local variances of the two sampling points and its complementary term, so that the weight avoids the misjudgment risk that may be brought by a single variance index and improves the prominent attention to the key feature area.

[0018] Further, the ratio of the cumulative energy of each segment to that of its corresponding segment is determined based on the following method: The sum of the squares of the amplitudes of all sampling points included in each segment of the waveform to be recognized at the mesoscale is calculated as the cumulative energy of this segment; the sum of the squares of the amplitudes of all sampling points included in the corresponding segment of this segment in the template waveform at the mesoscale is calculated as the reference cumulative energy of this segment; the ratio of the cumulative energy of this segment to the reference cumulative energy of this segment is used as the ratio of the cumulative energy of the segment to that of its corresponding segment, which is used to measure the degree of energy deviation of the waveform to be recognized on this segment.

[0019] Further, the phase consistency is determined based on the following method: The Hilbert transform is respectively performed on the waveform to be recognized and the template waveform at the mesoscale to obtain the instantaneous phase of each marked sampling point of the waveform to be recognized and the template waveform at the mesoscale; the phase consistency between each segment of the waveform to be recognized at the mesoscale and the corresponding segment of the template waveform at the mesoscale is determined by the difference between the instantaneous phases of the marked sampling points of each segment of the waveform to be recognized at the mesoscale and the instantaneous phases of the marked sampling points of the corresponding segment in the template waveform at the mesoscale.

[0020] This technical solution provides a more accurate recognition means for waveform anomaly recognition in the HVDC capacitive load system by introducing phase consistency analysis. Phase consistency can effectively reflect the synchronism of the waveform on the time axis. Especially when the waveform is disturbed or distorted, the phase consistency will be significantly affected. Through this method, abnormal waveforms with obvious deviations in time synchronism can be identified, thereby improving the accuracy of waveform matching and anomaly recognition.

[0021] Further, the local signal variance is determined according to the following method: Taking each marked sampling point as the center, the variance of the amplitudes of all sampling points between this marked sampling point and the nearest marked sampling point on its left is denoted as the first signal variance, and the variance of the amplitudes of all sampling points between this marked sampling point and the nearest marked sampling point on its right is denoted as the second signal variance. The mean of the first variance and the second variance is used as the local signal variance of this marked sampling point.

[0022] This technical solution can effectively capture the changes in local fluctuations by introducing the calculation method of local signal variance, enabling higher weights to be given to local abnormal points and signal change regions during the waveform matching process, thereby more accurately reflecting the true characteristics of the waveform and providing more accurate local feature recognition for waveform matching.

[0023] The present invention has the following effects: Through a multi-scale collaborative mechanism, a template matching idea, and a local anomaly focusing strategy, the present invention optimizes the multi-scale feature extraction and local anomaly detection of waveforms, avoiding the sensitivity problem of traditional dynamic time warping algorithms to local distortions during waveform anomaly recognition, ensuring the accuracy of the minimum cumulative cost path, and thus improving the accuracy of waveform anomaly recognition. Description of the Drawings

[0024] Figure 1 is a schematic flowchart of the method of the present invention; Figure 2 is a schematic flowchart of step S3 of the present invention; Figure 3 is a schematic flowchart of step S4 of the present invention. Detailed Embodiments

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention.

[0026] The method for recognizing waveform anomalies in a high-voltage DC capacitive load system based on template matching provided by the present invention, as Figure 1 shown in, includes: S1: Obtain multi-scale waveforms based on the high-voltage DC capacitive load system.

[0027] In a high-voltage DC capacitive load system, the waveform signal of the voltage has complex time-varying characteristics. Affected by factors such as the charge and discharge characteristics of the capacitive load, switch actions, electromagnetic interference, and equipment aging, the waveform often exhibits multi-scale and multi-band abnormal changes. To effectively capture these abnormal features and improve the accuracy and robustness of recognition, this step performs multi-scale sampling on the waveform signal, taking into account both the global trend and local details of the signal.

[0028] In a high-voltage DC capacitive load system, to analyze whether the waveform of the voltage is abnormal within a certain time period, it is necessary to obtain the waveform. Specifically, it includes: Let the length of the time period be T, that is, this time period includes T acquisition moments, each moment corresponds to an amplitude, and the magnitude of the amplitude represents the magnitude of the voltage. First, obtain the waveform within this time period as the original waveform to be recognized, denoted as , and in the present invention, T is set to 1 minute (an empirical value that can be adjusted by oneself). At the same time, pre-obtain the waveform of the high-voltage DC capacitive load system within a time period of length T during normal operation as the original template waveform, denoted as . The original waveform to be recognized and the original template waveform contain all the detailed information of the waveform, such as high-frequency noise, minute distortions, and instantaneous jumps.

[0029] Perform multi-scale downsampling on the original waveform to be recognized and the original template waveform respectively, including: performing downsampling with a ratio of 2:1 (empirical value) on the original waveform to be recognized and the original template waveform respectively. While retaining some details of the waveform, reduce the complexity of the original waveform for detecting abnormalities in intermediate-frequency components, such as periodic distortion, regular fluctuations, etc., to obtain the waveform to be recognized at the medium scale and the template waveform at the medium scale, denoted as and . At the same time, perform downsampling with a ratio of 5:1 (empirical value) on the original waveform to be recognized and the original template waveform respectively, discard the high-frequency details and noise in the original waveform to be recognized and the original template waveform, and retain the overall contour and trend characteristics of the waveform, such as rising, falling trends, peak positions, periodic trends, etc., to obtain the waveform to be recognized at the coarse scale and the template waveform at the coarse scale, denoted as and .

[0030] In the field of signal processing, each point on the waveform is usually called a sampling point. Corresponding serial numbers are sequentially assigned to each sampling point of the original template waveform according to the acquisition order. During the downsampling process, although some sampling points are discarded, each remaining sampling point originates from the original waveform. Therefore, the corresponding serial number in the original waveform can be recorded. In this way, the mapping relationship between the sampling points of waveforms at different scales can be established.

[0031] S2: Initially judge the overall trend of the waveform based on the coarse-scale waveform.

[0032] Use the traditional dynamic time warping algorithm to match the waveform to be recognized at the coarse scale and the template waveform at the coarse scale to obtain the minimum cumulative cost path between them. Each path point on the path corresponds to a sampling point combination. The two sampling points included in this sampling point combination come from and respectively. Take the direct distance between these two sampling points (the direct distance is the absolute value of the difference between the amplitudes of these two sampling points) as the matching error of this path point. Normalize the maximum and minimum values of the matching error of each path point to obtain the normalized matching error of each path point. Calculate the coefficient of variation of the normalized matching errors of all path points (the ratio of the standard deviation to the mean of the normalized matching errors) to reflect the relative fluctuation degree of the normalized matching errors.

[0033] Calculate the comprehensive matching error of the matching path, is the mean of the normalized matching errors of all path points on this path, reflecting the average error level, is the coefficient of variation of the normalized matching error for all path points, reflecting the fluctuation level of the normalized matching error. When the mean is larger, it indicates a larger matching error. When the coefficient of variation is smaller, it indicates a higher confidence level for a larger matching error, and vice versa.

[0034] Therefore, the matching error threshold is set to 0.8 (empirical value). If , it indicates that the overall trend of the original waveform to be recognized is consistent with that of the original template waveform and the fluctuation is stable, and further analysis and positioning are required to determine whether there are local anomalies; if (empirical value), it indicates that the overall trend of the original waveform to be recognized deviates significantly from that of the template waveform, and there are obvious trend anomalies in the waveform to be recognized.

[0035] S3: Map the coarse-scale waveform to the medium-scale waveform and segment it, and evaluate the anomaly degree of each waveform segment to locate the abnormal waveform segment of the medium-scale waveform.

[0036] In a high-voltage DC capacitive load system, voltage waveform anomalies may stem from various factors such as global trend deviation, local energy mutation, or phase timing disorder. Although the original waveform to be recognized contains complete details, its high-frequency noise and complex time-varying characteristics make it difficult to directly use for accurate anomaly positioning. The coarse-scale waveform to be recognized can quickly filter out noise and capture the global trend, such as overall voltage offset or abnormal charge and discharge cycles, but due to excessive smoothing, it cannot capture the intermediate-frequency detail features of the waveform.

[0037] Therefore, in this step, by mapping the coarse-scale waveform to be recognized to the medium-scale waveform to be recognized, the sampling points corresponding to the coarse-scale waveform to be recognized can be marked on the medium-scale waveform to be recognized. These sampling points usually correspond to the key nodes of the waveform, such as waveform peaks, zero-crossing points, or turning points in the charge and discharge stages. Segmenting the medium-scale waveform to be recognized with these key nodes as boundaries can disassemble the continuous waveform into independent segments aligned with the system physical process. This operation not only retains the guidance of the coarse-scale waveform for the global trend but also uses the medium resolution of the medium-scale to capture intermediate-frequency details, making each segment an independent segment with both global positioning anchor points and local feature carriers, thus avoiding the masking of local anomalies by global analysis and breaking through the noise interference of the original waveform, and finally achieving cross-scale accurate analysis from coarse-scale trend preliminary screening to medium-scale local positioning.

[0038] Specifically, as Figure 2 shown, it includes: S31: Map and mark the sampling points included in the coarse-scale waveform to the medium-scale waveform.

[0039] Map the sampling points included in the coarse-scale waveform to be recognized and the coarse-scale template waveform to the medium-scale waveform to be recognized and the medium-scale template waveform respectively, and mark them to obtain all the marked sampling points in the medium-scale waveform to be recognized and all the marked sampling points in the medium-scale template waveform.

[0040] S32: Segment the medium-scale waveform.

[0041] In the medium-scale waveform to be recognized, each segment is between two adjacent marked sampling points to complete the segmentation operation of the medium-scale waveform to be recognized. The medium-scale template waveform is also segmented, and the segmentation method is the same as that of the medium-scale waveform to be recognized.

[0042] Since the time lengths of the coarse-scale waveform to be recognized and the coarse-scale template waveform are the same, and their sampling points are in one-to-one correspondence based on the acquisition time, after mapping the coarse-scale waveform to be recognized and the coarse-scale template waveform to the medium-scale waveform and segmenting them, the corresponding segments in the medium-scale template waveform for each segment in the medium-scale waveform to be recognized can be obtained.

[0043] S33: Calculate the ratio of the cumulative energy of each segment to its corresponding segment.

[0044] In a high-voltage DC capacitive load system, the energy distribution of the voltage waveform is closely related to the system operation state. Under normal conditions, the energy characteristics of each stage (such as the energy accumulation rate during the charge and discharge process and the energy mutation amplitude during the switch action) have stability and regularity. However, when the system shows abnormalities (such as load leakage, line impedance change, or equipment aging), the energy in a local interval may deviate significantly, and such abnormalities often hide in the intermediate-frequency components of the waveform.

[0045] Therefore, by calculating the ratio of cumulative energy, the abstract waveform difference is converted into a quantifiable energy deviation index. When the cumulative energy of a segment in the medium-scale waveform to be recognized is significantly higher than that of the corresponding segment in the medium-scale template waveform, it indicates that there is a greater possibility of problems such as a decrease in the energy storage efficiency of the capacitive load, an abnormal increase in the power supply voltage, or high-frequency harmonic injection resulting in energy surplus within this segment; when the cumulative energy of a segment in the medium-scale waveform to be recognized is significantly lower than that of the corresponding segment in the medium-scale template waveform, it indicates that there is a greater possibility of energy loss phenomena such as abnormal energy consumption of the load, poor line contact, or obstruction of the charge and discharge process within this segment.

[0046] In one embodiment, the ratio of the cumulative energy of each segment to its corresponding segment is determined based on the following formula:

[0047] In this formula, The waveform to be recognized at the mesoscale The segment of the waveform to be recognized at the mesoscale and the The segment of the template waveform at the mesoscale, the cumulative energy ratio is or The sequence number of the sampling points included in the segment, is or The total number of sampling points included in the segment, is The The amplitude of the sampling point in the segment, is The The energy of the sampling point in the segment, is The cumulative energy of the segment, is The The amplitude of the sampling point in the segment, is The The ... energy of the sampling point in the segment, is The cumulative energy of the segment,

[0048] If , it means that the energy of the th segment of the waveform to be recognized at the mesoscale is higher than that of the th segment of the template waveform at the mesoscale. There may be an energy anomaly or signal enhancement phenomenon in this segment. If : It indicates that the energy of the th segment of the waveform to be recognized at the mesoscale is lower than that of the th segment of the template waveform at the mesoscale, which may be due to signal attenuation or information loss. If : It indicates that the th segment of the waveform to be recognized at the mesoscale is similar to the th waveform segment of the template waveform at the mesoscale, which may be a normal or well-matched signal.

[0049] S34: Determine the phase consistency of each segment and its corresponding segment.

[0050] In a high-voltage DC capacitive load system, the phase characteristics of the voltage waveform are the core indicators reflecting the synchronism and stability of the system's dynamic process. During normal operation, the phase change of the voltage waveform is strictly synchronized with the charge and discharge cycle of the capacitive load and the action timing of the switching devices, showing a fixed phase difference or periodic pattern. However, when abnormalities occur in the system (such as delayed switching actions, electromagnetic interference coupling, or equipment parameter drift), the phase relationship of the waveform may be disrupted, resulting in the deviation of the timing structure between segments from the normal mode. The necessity of determining phase consistency lies in that it can reveal the abnormalities in the dynamic process hidden in the waveform from the dimension of signal synchronism and make up for the limitations relying solely on energy analysis.

[0051] In one embodiment, the phase consistency is determined as follows: Perform Hilbert transforms on and respectively to obtain the instantaneous phase of each marked sampling point, that is, the phase information corresponding to the waveform signal at each sampling point, which is an important indicator for measuring the phase synchronism of two signals.

[0052] Next, quantify the phase consistency of these two waveform segments by calculating the phase locking value of each waveform segment of and the corresponding waveform segment in :

[0053] In this formula, is the th segment of and the th segment of phase consistency, that is, the phase locking value. The phase locking value is a statistic for measuring signal synchronization; when the phase locking value is 1, the phase difference between the two signals is a constant, that is, they are completely synchronized; when the locking value is 0, the phases are not synchronized. is the th segment of the th sampling point's instantaneous phase, is the th segment of the th sampling point's instantaneous phase, represents the imaginary unit, represents a complex number with a unit length, and the phase of this complex number is , is the imaginary unit, and its role is to convert the angular difference into a complex number representation on the unit circle, and is determined through Euler's formula , is the or No. The total number of sampling points contained in each segment, To find the modulus length symbol, is a natural constant.

[0054] In this formula, represents a complex number of unit length, whose phase is . It will The instantaneous phase and sampling point of each segment The instantaneous phase differences of the sampling points of the corresponding segments in are added up in complex form and the average modulus is calculated to measure the phase synchronization degree of the two waveform segments.

[0055] In short, when the phase consistency of the two waveform signals is greater, the synchronization is better (the phase difference is a constant), the modulus of the sum of the complex numbers is larger and closer to 1; when the phase is randomly distributed, the modulus of the sum of the complex numbers approaches 0 and closer to 0.

[0056] S35: Determine the degree of abnormality by combining the ratio of phase consistency and cumulative energy.

[0057] Single energy or phase analysis often fails to fully characterize the complex characteristics of waveform anomalies. Energy deviations may reflect sudden amplitude changes or power imbalances, while phase disturbances indicate timing desynchronization or dynamic process distortion. In actual anomaly scenarios, both often occur together (for example, load aging can lead to increased energy loss and may also cause charge and discharge phase shifts). Therefore, combining phase consistency and the ratio of accumulated energy to determine the degree of anomaly avoids the one-sidedness of single feature analysis.

[0058] In one embodiment, the abnormality level is determined based on the following formula:

[0059] In this formula, is the first mesoscale waveform to be identified The abnormality level of each segment, is the first mesoscale waveform to be identified The first segment and the mesoscale template waveform The ratio of the cumulative energy of the segments, is the first mesoscale waveform to be identified The first segment and the mesoscale template waveform The phase consistency of each segment, To prevent 0 parameters. The value is .

[0060] In this formula, The first part measures the mesoscale waveform to be identified. The energy deviation degree between the th segment of a sub - segment and the mesoscale template waveform. When is significantly greater than or less than 1, it indicates that the greater the energy deviation degree, the more likely it is the energy mutation characteristic of an abnormal waveform. The purpose of the square operation is to strengthen the energy deviation degree. Whether it is on the high side or the low side, it is considered abnormal (symmetry processing). The part is used as a normalization factor, which can avoid the abnormal degree being infinitely amplified due to too large an energy deviation degree. Introducing is to prevent the denominator from being zero or too small, resulting in numerical instability and maintaining the numerical stability of the abnormal degree. represents the difference in the overall phase trend between the th segment of the mesoscale waveform to be identified and the th segment of the mesoscale template waveform. If is close to 1, the phase trends are almost the same. The closer it is to 0, the more likely this segment is a normal waveform, and the smaller the abnormal degree. The closer it is to 0; if is closer to 0, the more different the phase trends are, and the more likely this segment is an abnormal waveform, and the greater the abnormal degree. The closer it is to 1.

[0061] S36: Screen the abnormal segments of the mesoscale waveform to be identified according to the abnormal degree.

[0062] Obtain the abnormal degrees of all segments of , sort them in ascending order, and then use the quartile method to regard the waveform segments with abnormal degrees greater than the third quartile (greater than 75%) as abnormal waveform segments.

[0063] S4: Map the abnormal waveform segments of the mesoscale waveform to the original - scale waveform, perform adaptive dynamic time warping matching, and determine the optimal matching path.

[0064] Although the abnormal segments of the mesoscale waveform to be identified have been located in the intermediate - frequency region where the energy and phase deviate significantly, this region may still contain detailed features such as high - frequency noise and transient jumps in the original waveform, which directly affect the accurate determination of the nature of the abnormality. For example, differentiating between progressive distortion caused by equipment aging and occasional spikes caused by instantaneous interference.

[0065] Therefore, in this step, the abnormal segments of the mesoscale waveform to be identified are mapped to the original waveform to be identified and adaptive dynamic time warping matching is performed. It can perform refined analysis on the high - frequency details of the abnormal region on the basis of retaining the global constraints at the coarse scale and the positioning accuracy at the mesoscale.

[0066] According to the medium-scale waveform to be recognized the sampling points it contains and the original waveform to be recognized and the mapping relationship of the sampling points it contains, map the medium-scale waveform to be recognized all the sampling points contained in each abnormal segment of and mark them on the original waveform to be recognized to obtain all the marked sampling points of the original waveform to be recognized

[0067] Similarly, map all the sampling points contained in the corresponding segment of each abnormal segment of the medium-scale waveform to be recognized in the medium-scale template waveform and mark them on the original template waveform to obtain all the marked sampling points in the original template waveform

[0068] Specifically, the process of adaptive dynamic time warping matching is as Figure 3 shown, including: S41: Calculate the direct distance matrix of all the marked sampling points of the original waveform to be recognized and all the marked sampling points of the original template waveform

[0069] First, obtain each marked sampling point of and the direct distance (the difference in the amplitudes of the sampling points, and then take the absolute value) of each marked sampling point of Let the number of marked sampling points of be grange from 1 to ; Let the number of marked sampling points of be range from 1 to According to each marked sampling point of to each marked sampling point of form a direct distance matrix

[0070] In the direct distance matrix the element located at has a value of the th marked sampling point of and ​​The direct distance between two marked sampling points. The smaller the direct distance, the smaller the matching cost between these two marked sampling points, and vice versa.

[0071] S42: Based on the direct distance matrix, use the dynamic programming method to calculate the cumulative distance matrix, and introduce a distance weight coefficient to reconstruct the recurrence formula for the cumulative distance in this process.

[0072] The specific steps include: S421: Initialize the boundary conditions.

[0073] When : For the origin (upper left corner) of the cumulative matrix, and and , ; For the first row (advancing horizontally) of the cumulative matrix, and and , , where ; For the first column (descending vertically) of the cumulative matrix, and and , , ; S422: Reconstruct the recurrence formula for the cumulative distance with a dynamic weight coefficient to calculate the cumulative distance matrix.

[0074] When , during the recurrence process of the cumulative distance, the traditional recurrence formula usually assigns a fixed weight to the transfer path of the time step, which easily causes the algorithm to be difficult to adapt to the dynamic changes of the local characteristics of the waveform, such as amplitude mutation, phase jump, or noise interference. Therefore, by introducing a dynamic weight coefficient, the calculation of the cumulative distance can dynamically respond to the differences in the local characteristics of the waveform, thereby improving the adaptability and accuracy of the matching algorithm for complex signals.

[0075] In one embodiment, introducing a distance weight coefficient reconstructs the recurrence formula for the cumulative distance as:

[0076] In this formula, is the direct distance between the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform, is the cumulative distance between the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform. To obtain the minimum value, is the distance weight coefficient between the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform.

[0077] In this formula, is used to dynamically and adaptively adjust the direct distance between the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform and the weight assignment of the historical cumulative distance, and its value range is usually . The larger is, the more dependent on the direct distance when calculating the cumulative distance to reflect the matching cost between sampling points , dominates the calculation of the cumulative distance, and enhances the capture accuracy of abnormal features such as amplitude mutations and phase jumps; The smaller is, the more dependent on the historical cumulative distance when calculating the cumulative distance to reflect the matching cost between sampling points. Based on the smooth trend of the historical cumulative distance, it can suppress noise interference and maintain the continuity of timing matching.

[0078] In summary, this dynamic adjustment mechanism enables the algorithm to flexibly switch between precise alignment of local features and smooth transition of global trends, improving the adaptability to high-voltage DC capacitive load waveforms with noise, time-variation, or multi-scale anomalies. It not only avoids missed detection or mis-matching of abnormal features by traditional methods but also enhances the robustness of the matching process, ultimately providing a more reliable quantitative basis for high-precision waveform anomaly localization and fault diagnosis.

[0079] In one embodiment, the dynamic weight coefficient is calculated based on the following formula:

[0080] In this formula, is the distance weight coefficient between the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform, is the normalized value of the local signal variance of the th marked sampling point of the original waveform to be recognized. This normalized value is equal to the local signal variance of the th marked sampling point of divided by the maximum value of the local signal variances of all marked sampling points of . is the th marked sampling point of the original template waveform [[ID=6I]]is the The normalized value of the local signal variance of a marked sampling point, which is equal to the local signal variance of the th marked sampling point divided by the maximum value of the local signal variances of all marked sampling points of

[0081] In this formula, the numerator term measures the local fluctuation consistency between the sampling points of the original waveform to be recognized and the corresponding sampling points of the original template waveform. If the local signal variances of two sampling points are both high and the product value is large, it indicates that there is more likely to be a common abnormal feature or strong dynamic change, and key matching is required. The greater the distance weight coefficient, the more dependent on the direct distance between the two sampling points during matching.

[0082] In this formula, the denominator term is equal to the accumulated value of the numerator term and the complementary term of the numerator term. The in the denominator term is the complementary term of the numerator term, which reflects the local consistency between the sampling points of the original waveform to be recognized and the corresponding sampling points of the original template waveform.

[0083] If the local signal variance of the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform is smaller, it indicates that the waveform is stable. At this time, the numerator term is smaller, and in the denominator term, the complementary term of the numerator term is larger, and the denominator is dominated by the complementary term of the numerator term. At this time, the distance weight coefficient is smaller, and more dependent on the historical cumulative distance during matching.

[0084] Conversely, if the local signal variance of the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform is larger, it indicates that the waveform fluctuates violently. At this time, the numerator term is larger, and in the denominator term, the complementary term of the numerator term is small, and the denominator is dominated by the numerator term. At this time, the distance weight coefficient is larger, and more dependent on the direct distance during matching.

[0085] In one embodiment, the local signal variance of each marked sampling point is determined based on the following method: for any marked sampling point of the original waveform to be recognized or the original template waveform, taking it as the center, obtain all the sampling points included between it and the nearest marked sampling point on the left (excluding the nearest marked sampling point on the left itself), and calculate the variance of the amplitudes of these sampling points; obtain all the sampling points included between it and the nearest marked sampling point on the right (excluding the nearest marked sampling point on the right itself), and calculate the variance of the signal amplitudes of these sampling points; take the mean of these two variances as the local signal variance of this marked sampling point.

[0086] The local signal variance of a marker sampling point reflects the importance of that sampling point itself, that is, whether it is unusually active or stable. The purpose is to focus the matching on areas of significant change. A larger local signal variance means that the local signal fluctuations at that marker sampling point are more dramatic, which may correspond to the marker sampling point being more likely to be in a key part of the waveform, thus being more important and therefore being given a higher weight. The opposite is true for the same sampling point.

[0087] Finally, the cumulative distance matrix is obtained by calculating each element of the cumulative distance matrix according to the recursive formula of the reconstructed cumulative distance. Each element of the cumulative distance matrix corresponds to a marked sampling point combination, and the value of the element is the cumulative distance of the marked sampling point combination. The two marked sampling points contained in the marked sampling point combination come from the original waveform to be identified and the original template waveform respectively.

[0088] S43: Perform path backtracking based on the cumulative distance matrix to obtain the optimal matching path.

[0089] Specifically, the path is traced back from the lower right corner of the cumulative distance matrix to the upper left corner to obtain the minimum cumulative cost path, which is used as All marked sampling points and The optimal matching path between all marked sampling points is found to complete the adaptive dynamic time warping matching.

[0090] Specifically, the path backtracking process is: S431: Starting from the element at the lower right corner of the accumulation matrix, use this element as the current path point; S432: Select the smallest element among the elements above, to the left, and in the upper left corner of the current path point as the next path point.

[0091] For example, if the current path point is , in its adjacent elements 、 or Select an element with the smallest value as the next path point.

[0092] S433: Update the current path point, take the next path point of the current path point as the new current path point, and execute step S432; until the path backtracks to the element in the upper left corner of the accumulation matrix, the path backtracking is completed.

[0093] All the path points in the backtracking process form a minimum cumulative cost path, which reflects the best alignment of the marked sampling points of the original waveform to be recognized and the marked sampling points of the original template waveform on the time axis. Therefore, it is used as the optimal matching path between the marked sampling points of the original waveform to be recognized and the marked sampling points of the original template waveform. The optimal matching path consists of multiple path points, each path point corresponds to an element, and each element is composed of two marked sampling points.

[0094] S5: Calculate the matching error according to the optimal matching path for waveform anomaly recognition.

[0095] In the dynamic time warping algorithm, the optimal matching path is obtained through the cumulative distance matrix, which reflects the minimum matching cost between two sequences. In the present invention, the optimal matching path represents the best alignment between the marked sampling points of the original waveform to be recognized and the marked sampling points of the original template waveform. Each path point on the optimal matching path corresponds to a combination of marked sampling points. Each combination of marked sampling points includes two marked sampling points, one from the original waveform to be recognized and the other from the original template waveform. The direct distance between these two sampling points is used as the matching error corresponding to this path point, and the comprehensive matching error of this optimal matching path is calculated with reference to the operations in step S2.

[0096] Specifically: Normalize the maximum and minimum values of the matching errors of each path point on the optimal matching path to obtain the normalized matching error of each path point. Then, calculate the comprehensive matching error of this optimal matching path , is the mean of the normalized matching errors of all path points of this optimal matching path, is the coefficient of variation of the normalized matching errors of all path points of this optimal matching path. When the mean is larger, it indicates that the matching error is larger. When the coefficient of variation is smaller, it indicates that the confidence level of the larger matching error is higher, and vice versa.

[0097] Similarly, set the matching error threshold to 0.5 (empirical value). If , it indicates that the matching between the marked sampling points in the original waveform to be recognized and the marked sampling points in the original template waveform is precise, and it is determined that there is no anomaly in the original waveform to be recognized; if (empirical value), it indicates that the matching error between the marked sampling points in the original waveform to be recognized and the marked sampling points in the original template waveform is too large, and it is determined that there is an anomaly in the original waveform to be recognized.

[0098] Although this specification has shown and described multiple embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided by way of example only.

Claims

1. A method for identifying abnormal waveforms in a HVDC capacitive load system based on template matching, characterized in that, Including: Construct the original waveform to be recognized and the original template waveform of the voltage signal of the high-voltage DC capacitive load system, and obtain the waveforms to be recognized and the template waveforms at medium scale and coarse scale through multi-scale downsampling; Perform a preliminary match between the waveform to be recognized at the coarse scale and the template waveform to determine whether the overall waveform trends are consistent; if not, determine that the original waveform to be recognized is abnormal; If they are consistent, map and mark the corresponding sampling points of the waveform to be recognized and the template waveform at the coarse scale to the waveforms to be recognized and the template waveforms at the medium scale; In the waveforms to be recognized and the template waveforms at the medium scale, each segment between two adjacent marked sampling points is a segment, and filter out the abnormal segments of the waveform to be recognized at the medium scale, map and mark the sampling points corresponding to the abnormal segments to the original waveform to be recognized, and at the same time, map and mark the sampling points corresponding to the abnormal segments in the template waveform at the medium scale to the original template waveform; Perform dynamic time warping matching according to the marked sampling points in the original waveform to be recognized and the marked sampling points in the original template waveform, and determine whether the original waveform to be recognized is abnormal in response to the judgment result of whether the error of the matching result exceeds the normal error range.

2. The waveform anomaly recognition method for a HVDC capacitive load system based on template matching according to claim 1, wherein The dynamic time warping matching includes: Construct a direct distance matrix according to the direct distance between the marked sampling points in the original waveform to be recognized and the marked sampling points in the original template waveform; on the basis of the direct distance matrix, use the dynamic programming method to calculate the cumulative distance matrix, start from the lower right corner of the cumulative distance matrix and perform path backtracking to the upper left corner to end, and obtain the minimum cumulative cost path to complete the dynamic time warping matching; Each element of the cumulative distance matrix corresponds to a combination of marked sampling points, and the value of the element is the cumulative distance of the combination of marked sampling points; the combination of marked sampling points contains two marked sampling points from the original waveform to be recognized and the original template waveform respectively; the minimum cumulative cost path consists of multiple path points, and each path point corresponds to an element.

3. The waveform abnormality recognition method for the HVDC capacitive load system based on template matching according to claim 1, wherein The method for filtering out the abnormal segments of the waveform to be recognized at the medium scale is: Obtain the corresponding segments of each segment of the waveform to be recognized at the medium scale in the template waveform at the medium scale; determine the abnormality degree of each segment according to the ratio of the cumulative energy of each segment and its corresponding segment and the phase consistency; Obtain the abnormality degrees of all segments of the waveform to be recognized at the medium scale, and use the quartile method to take the segments with abnormality degrees greater than the third quartile as the abnormal segments in the waveform to be recognized at the medium scale.

4. The waveform abnormality recognition method for the HVDC capacitive load system based on template matching according to claim 2, wherein The method for determining whether the original waveform to be recognized is abnormal is: If the error of the matching result exceeds the normal error range, determine that the original waveform to be recognized is abnormal; If the error of the matching result does not exceed the normal error range, determine that the original waveform to be recognized is not abnormal.

5. The waveform anomaly recognition method for the HVDC capacitive load system based on template matching according to claim 2, characterized in that, In the process of calculating the cumulative distance matrix by using the dynamic programming method, it also includes introducing a distance weight coefficient to reconstruct the recurrence formula of the cumulative distance as: ； In the formula, and are the direct distance and cumulative distance of the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform respectively, is the distance weight coefficient of the th marked sampling point of the original waveform to be recognized and the th marked sampling point of the original template waveform, is to take the minimum value.

6. The waveform abnormality recognition method for a HVDC capacitive load system based on template matching according to claim 3, wherein The abnormality degree of each segment is determined based on the following formula: ； In this formula, is the abnormality degree of the th segment of the mesoscale waveform to be recognized, is the ratio of the cumulative energy of the th segment of the mesoscale waveform to be recognized to that of the th segment of the mesoscale template waveform, is the phase consistency of the th segment of the mesoscale waveform to be recognized and the th segment of the mesoscale template waveform, is the anti-0 parameter.

7. The waveform anomaly recognition method for a HVDC capacitive load system based on template matching according to claim 5, characterized in that The distance weight coefficient is determined based on the following formula: ； In the formula, is the distance weight coefficient between the -th marked sampling point of the original waveform to be recognized and the -th marked sampling point of the original template waveform, is the normalized value of the local signal variance of the -th marked sampling point of the original waveform to be recognized, and is the normalized value of the local signal variance of the -th marked sampling point of the original template waveform. ​​ 8. The method for identifying abnormal waveforms of a high-voltage DC capacitive load system based on template matching according to claim 6, characterized in that, The ratio of the cumulative energy of each segment and its corresponding segment is determined based on the following method: Sum the squares of the amplitudes of all the sampling points included in each segment of the mesoscale waveform to be recognized as the cumulative energy of this segment; Sum the squares of the amplitudes of all the sampling points included in the corresponding segment of this segment in the mesoscale template waveform as the reference cumulative energy of this segment; use the ratio of the cumulative energy of this segment to the reference cumulative energy of this segment as the ratio of the cumulative energies of this segment and its corresponding segment, which is used to measure the energy deviation degree of the waveform to be recognized on this segment.

9. The waveform anomaly recognition method for a HVDC capacitive load system based on template matching according to claim 6, characterized in that The phase consistency is determined based on the following method: Perform Hilbert transforms on the mesoscale waveform to be recognized and the template waveform respectively to obtain the instantaneous phases of each marked sampling point of the mesoscale waveform to be recognized and the template waveform respectively; Determine the phase consistency between each segment of the mesoscale waveform to be recognized and its corresponding segment by the difference between the instantaneous phases of the marked sampling points of each segment of the mesoscale waveform to be recognized and the instantaneous phases of the marked sampling points of the corresponding segment in the mesoscale template waveform.

10. The method for identifying abnormal waveforms of a high-voltage DC capacitive load system based on template matching according to claim 7, characterized in that, The local signal variance is determined according to the following method: Taking each marked sampling point as the center, record the variance of the amplitudes of all the sampling points between this marked sampling point and the nearest marked sampling point on its left as the first signal variance, and record the variance of the amplitudes of all the sampling points between this marked sampling point and the nearest marked sampling point on its right as the second signal variance; take the mean of the first signal variance and the second signal variance as the local signal variance of this marked sampling point.

Images