Wind power wire harness performance detection method based on data processing

Abstract

The invention belongs to the technical field of electrical variable measurement, and particularly relates to a wind power wire harness performance detection method based on data processing, which comprises the following steps of: segmenting pulse current data of a wind power wire harness; determining a pulse width characterization factor and a structural element length correction coefficient of the pulse current data segment; determining a final length of a structural element of the pulse current data segment, constructing a flat structural element according to the final length of the structural element of the pulse current data segment, performing morphological filtering on the pulse current data segment by using the flat structural element, and determining a waveform extraction threshold value of the pulse current data segment; therefore, partial discharge signals are extracted to detect the performance of the wind power wire harness. According to the method, the frequency domain characteristics of the pulse current data segment are analyzed, the final length of the structural element is self-adapted, the limitation of fixed parameter filtering is overcome, the partial discharge signal is extracted, and the accuracy and robustness of insulation performance detection of the wind power wire harness are improved.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement technology. More specifically, this invention relates to a method for testing the performance of wind power harnesses based on data processing. Background Technology

[0002] Currently, the world is vigorously promoting the development of clean energy. As a core component, wind power generation has seen continuous growth in installed capacity. Medium and high voltage wind power cables, as the core connecting various electrical components inside the wind turbine generator set, undertake the important tasks of power transmission and pulse current control. Their insulation performance directly affects the safe operation of the wind turbine.

[0003] During the production and testing of wind power harnesses, air gaps or impurities often arise due to material defects or process problems, leading to partial discharge phenomena and potentially causing insulation breakdown accidents. Therefore, rigorous partial discharge testing before the finished product leaves the factory is a crucial step in ensuring the quality of wind power harnesses. Traditional testing methods are limited by environmental noise interference and struggle to accurately extract weak partial discharge pulse currents. Morphological filtering algorithms, as a nonlinear pulse current processing method, effectively preserve the geometric characteristics of the pulse current and filter out noise by sliding structuring elements across the pulse current and performing morphological operations.

[0004] However, when using morphological filtering algorithms to extract partial discharge pulse current from wind power harnesses, a fixed-length structural element is usually preset. This fixed-length structural element is difficult to adapt to the pulse current waveform changes of the wind power harness in the coiled test state. When the partial discharge signal pulse propagates along a long distance of the harness, the high-frequency component attenuates faster than the low-frequency component, resulting in a wider pulse current at the far end and a narrower pulse current at the near end. If the structural element is too short, it will cut off the wider pulse at the far end; if the structural element is too long, it will filter out the narrower pulse at the near end, thus reducing the extraction accuracy of the partial discharge signal and affecting the detection of wind power harness performance. Summary of the Invention

[0005] To address the technical problem that traditional morphological filtering algorithms for extracting partial discharge signals suffer from limitations due to the fixed-length pre-set structural elements, which are ill-suited to the varying pulse current waveforms of wind power harnesses during coiling tests, leading to decreased extraction accuracy and impacting wind power harness performance testing, this invention provides a data processing-based wind power harness performance testing method. The method includes: dividing the pulse current data of the wind power harness into several pulse current data segments; determining the pulse width characterization factor for each pulse current data segment, wherein the pulse width characterization factor is negatively correlated with both the frequency centroid and bandwidth of the pulse current data segment; determining a structural element length correction coefficient for each pulse current data segment, wherein the structural element length correction coefficient is positively correlated with both the power spectral entropy and the pulse width characterization factor of the pulse current data segment; correcting the initial length of the structural elements of the pulse current data segment according to the structural element length correction coefficient to obtain the final length of the structural elements; constructing a flat structural element for the pulse current data segment based on the final length of the structural elements; performing morphological filtering on the pulse current data segment using the flat structural element; and determining a feature waveform extraction threshold based on the morphological filtering result to extract the partial discharge signal and achieve wind power harness performance testing.

[0006] This invention effectively identifies waveform broadening characteristics at different transmission distances by calculating the pulse width characterization factor and utilizing frequency domain distribution features. It also identifies high-frequency background noise by calculating the structural element length correction coefficient. By comprehensively considering the signal's pulse width characterization factor and power spectral entropy, it can more accurately distinguish between genuine partial discharge signals and interference from spurious narrow pulses. Furthermore, by constructing flat structural elements based on the corrected final length for morphological filtering, it achieves the wrapping of pulse current waveforms of different shapes by the structural element scale. This avoids excessive erosion of far-end wide pulses by short structural elements and prevents the false filtering of near-end narrow pulses by long structural elements. Simultaneously, it utilizes the smoothing characteristics of long scales to suppress high spectral entropy noise. In the complex environment of wind power harness winding tests, it can more accurately extract partial discharge signals, thereby improving the reliability of harness insulation performance results.

[0007] Preferably, the step of dividing the pulse current data of the wind power harness into several pulse current data segments includes: dividing the pulse current data of the wind power harness into several pulse current data segments of equal length using a sliding window.

[0008] Preferably, the method for obtaining the bandwidth of the pulse current data segment is as follows: the ratio of the frequency standard deviation of the pulse current data segment to the maximum bandwidth of the digital signal system is used as the bandwidth of the pulse current data segment.

[0009] Preferably, the pulse width characterization factor satisfies the expression: In the formula, For the first The pulse width characterization factor for each pulse current data segment. This represents the total number of frequency points in the pulse current data segment spectrum. For the first The pulse current data segment in the first Power spectral density values ​​at each frequency point For the first The pulse current data segment in the first Frequency values ​​at each frequency point For the first The frequency centroid of each pulse current data segment For the first The bandwidth of each pulse current data segment It is a natural exponential function.

[0010] This invention achieves the evaluation of pulse width characterization factors by constructing an inverse proportional function relationship with the product of the frequency centroid and the bandwidth exponent as the denominator. The frequency centroid term reflects the frequency domain distribution position of the signal energy, while the bandwidth term amplifies the influence of bandwidth differences. This allows the pulse current data segment with a high centroid and wide bandwidth to obtain a smaller pulse width characterization factor, thereby mapping the narrow pulse shape of the near-end signal when the waveform has not been widened, and the wide pulse shape of the far-end signal after the waveform has been widened. This provides a more reliable evaluation basis for calculating the final length of the structural element.

[0011] Preferably, the structural element length correction coefficient satisfies the expression: In the formula, For the first The structure element length correction factor for each pulse current data segment For the first The pulse width characterization factor for each pulse current data segment. For the first Power spectral entropy of each pulse current data segment This represents the maximum spectral entropy value within the historical pulse current data segment. The value of the structure element length correction coefficient for each pulse current data segment is adjusted by the coefficient. It is a natural exponential function. It is the maximum-minimum normalization function.

[0012] This invention constructs a composite function that includes a power spectral entropy ratio term and a pulse width characterization factor exponent term, thereby enabling the evaluation of the structural element length correction coefficient. The power spectral entropy term reflects the degree of disorder in the signal frequency domain, while the pulse width characterization factor exponent term amplifies the weight of the pulse width factor in interference determination through nonlinear mapping. This results in a larger correction coefficient for high-frequency noise that is misjudged as a narrow pulse, thus enabling more effective identification and filtering of background noise.

[0013] Preferably, the final length of the structuring element satisfies the expression: In the formula, For the first The final length of the structure element for each pulse current data segment For the first The structure element length correction factor for each pulse current data segment The initial length of the structure element for the pulse current data segment. This represents the function for rounding up.

[0014] This invention determines the final length of a structural element by constructing a product gain function based on the initial length of the structural element and a correction coefficient for the structural element length. The correction coefficient is used as a gain factor to dynamically adjust the initial length. When high-frequency noise interference is detected, the length of the structural element is stretched through multiplication to enhance morphological smoothing capability. When the signal is pure, the length of the structural element is shortened to preserve details, ensuring that high-frequency noise can be effectively filtered out while preserving the details of the real partial discharge pulse.

[0015] Preferably, the method for obtaining the flat structural element is as follows: define a one-dimensional sequence with a length equal to the final length of the structural element, and assign the amplitude height at each point in the one-dimensional sequence to 0 to obtain the flat structural element.

[0016] Preferably, the morphological filtering of the pulse current data segment yields a denoised signal, which satisfies the expression: In the formula, For the first The first pulse current data segment The signal after denoising at each sampling point This is the index of the sampling point in the pulse current data segment. For opening operation, For closing operations, For the first The signal before denoising of each pulse current data segment For the first A flat structure element for each pulse current data segment.

[0017] Preferably, the feature waveform extraction threshold satisfies the expression: In the formula, For the first Threshold for extracting characteristic waveforms within a single pulse current data segment. For the first The first pulse current data segment The signal after denoising at each sampling point This represents the total number of sampling points in the pulse current data segment. This is the index of the sampling point in the pulse current data segment. For the first The arithmetic mean of the absolute values ​​of the signals after denoising each pulse current data segment.

[0018] This invention achieves adaptive evaluation of the feature waveform extraction threshold by constructing a multiple gain function based on the arithmetic mean of the absolute values ​​of the denoised signal. The arithmetic mean term characterizes the residual background noise level of the current data segment after morphological filtering, and the multiple coefficient establishes the statistical boundary that distinguishes the effective signal from the background residual through linear amplification. This allows the extraction threshold to be dynamically adjusted in real time according to the fluctuation of environmental noise, thereby enabling the sharp capture of sudden pulses while shielding random interference, and improving the accuracy of extracting the feature waveform of partial discharge signal in non-stationary noise environments.

[0019] Preferably, the extraction of partial discharge signals includes: traversing all pulse current data segments, marking all continuous time segments within the pulse current data segments whose absolute amplitude exceeds the characteristic waveform extraction threshold as characteristic waveforms, and recording their start and end times, thereby completing the extraction of partial discharge signals from the wind power harness and realizing the performance detection of the wind power harness.

[0020] The beneficial effects of this invention are as follows: This invention solves the technical problem that traditional morphological filtering algorithms cannot adapt to the complex and variable pulse waveforms in wind power harness winding tests due to the inability of preset fixed-length structural elements, thus limiting the accuracy of partial discharge signal extraction. This is achieved by introducing an adaptive structural element length correction mechanism based on multi-dimensional frequency domain feature fusion.

[0021] This invention establishes the intrinsic mapping relationship between the pulse width characterization factor and the frequency centroid and bandwidth by analyzing the frequency domain physical characteristics of the pulse current data segment. Based on this, it further analyzes the structural element length correction coefficient. The structural element length correction coefficient tightly couples the power spectral entropy, which reflects the signal disorder and background noise level, with the pulse width characterization factor, which reflects the time-frequency morphology of the signal. This enables real-time and dynamic adaptive adjustment of the structural element length of morphological filtering under different test environments.

[0022] This invention can calculate the final length of the structural element that precisely matches the signal characteristics of each independent pulse current data segment. By constructing a flat structural element based on the final length and performing morphological filtering, accurate matching between the filter scale and the signal morphology is achieved. For narrow pulses without near-end dispersion, short structural elements are matched to prevent false filtering; for wide pulses with far-end dispersion, long structural elements are matched to prevent waveform fragmentation. Simultaneously, the entropy characteristic is combined to effectively suppress high-frequency background noise. Ultimately, this invention improves the accuracy and completeness of partial discharge signal extraction from wind power harnesses, enhances the robustness of performance testing results, and enables timely and accurate identification of potential defects within the insulation layer, providing a solid technical guarantee for the long-term safe and stable operation of wind power generation equipment. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the data processing-based wind power harness performance testing method of the present invention; Figure 2 A schematic diagram of filtering when the length of the fixed structural element in the existing technology is too long; Figure 3 A schematic diagram of filtering when the length of the fixed structural element in the existing technology is too short; Figure 4 This is a schematic diagram of the adaptive structuring element length filtering of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] This invention discloses a method for testing the performance of wind power harnesses based on data processing, referring to... Figure 1 This includes steps S001 to S005: S001: Acquire the pulse current data of the wind power harness and divide the pulse current data of the wind power harness into several pulse current data segments.

[0027] Specifically, during the withstand voltage test, a high-frequency current transformer clipped to the grounding wire of the wind power harness is used to collect pulse current in real time. After acquiring the pulse current data, the continuous pulse current data is divided into several pulse current data segments. In this embodiment, the sampling frequency of the pulse current is... The pulse current data segment is segmented using the overlap rate method. , length is The pulse current data is segmented using a sliding window with each sampling point. In other embodiments, the implementer can set the sampling frequency of the pulse current and the segmentation method of the pulse current data according to the actual implementation situation.

[0028] S002: Determine the pulse width characterization factor of the pulse current data segment. The pulse width characterization factor is negatively correlated with the frequency centroid of the pulse current data segment and negatively correlated with the bandwidth of the pulse current data segment.

[0029] It should be noted that when using morphological filtering algorithms to extract partial discharge signals, the preset fixed-length structural elements are difficult to adapt to the different pulse current waveforms during wind power harness testing, leading to a decrease in the accuracy of partial discharge signal extraction and inaccurate test results for wind power harness performance. According to transmission line theory, when a partial discharge pulse propagates along the harness, the skin effect of the conductor and the dielectric loss of the insulation medium act as a low-pass filter, causing the signal waveform to be widened. Therefore, this invention combines the frequency centroid and bandwidth of the pulse current data segment to determine the pulse width characterization factor of the pulse current data segment, which is used to characterize the degree of waveform widening of the pulse current data segment.

[0030] Specifically, the pulse width characterization factor satisfies the following expression: ; In the formula, For the first The pulse width characterization factor for each pulse current data segment. This represents the total number of frequency points in the pulse current data segment spectrum. For the first The pulse current data segment in the first Power spectral density values ​​at each frequency point For the first The pulse current data segment in the first Frequency values ​​at each frequency point For the first The frequency centroid of each pulse current data segment For the first The bandwidth of each pulse current data segment , For the first Standard deviation of frequency for each pulse current data segment The maximum bandwidth of the digital signal system, thereby ensuring The value of is in arrive between, It is a natural exponential function.

[0031] in, The larger the value, the more likely it is to be the first. The more concentrated the signal energy of the pulse current data segment is in the high-frequency region, the more it indicates that the signal energy of the pulse current data segment is concentrated in the high-frequency region. The more completely the high-frequency components are preserved in the pulse current data segment, the more it indicates that the... The greater the likelihood that the pulse current data segment belongs to a near-end narrow pulse, the more important it is to match the first pulse. The geometry of the pulse current data segment is determined to prevent excessively long structure elements from filtering out valid signals, and is allocated to the first pulse current data segment. The smaller the length of the structure element of the pulse current data segment, the better. The smaller the pulse width characterization factor of each pulse current data segment. The larger the value, the more likely it is to be the first. The wider the signal bandwidth distribution of the pulse current data segment, the more it indicates that the... The greater the likelihood that the pulse current data segment has not undergone long-distance transmission and that its high-frequency components are preserved intact, the more likely it is that the first pulse current data segment has not undergone long-distance transmission and that its high-frequency components are preserved intact. The greater the probability that a pulse current data segment belongs to a near-end narrow pulse, the higher its reliability. To match the first... The geometry of the pulse current data segment is determined to prevent excessively long structure elements from filtering out valid signals, and is allocated to the first pulse current data segment. The smaller the length of the structure element of the pulse current data segment, the better. The smaller the pulse width characterization factor of each pulse current data segment, the smaller the pulse width characterization factor will be.

[0032] S003: Determine the structural element length correction coefficient for the pulse current data segment. The structural element length correction coefficient is positively correlated with the power spectral entropy and pulse width characterization factor of the pulse current data segment.

[0033] It should be noted that, in the noisy environment of a factory, high-frequency noise such as inverter switches also has the characteristics of high center of gravity and wide bandwidth, which may lead to misjudgment as near-end narrow pulses. If short structural elements are used directly in this case, the noise will not be filtered out. Near-end narrow pulses, as real partial discharge signals, usually have a specific resonant frequency, relatively concentrated and orderly energy distribution, and lower power spectral entropy. On the other hand, the spectrum of background noise is usually chaotic, with uniform and disordered energy distribution and higher power spectral entropy. Therefore, this invention combines the power spectral entropy of the pulse current data segment and the pulse width characterization factor to determine the structural element length correction coefficient of the pulse current data segment.

[0034] Specifically, the structuring element length correction factor satisfies the expression: ; In the formula, For the first The structure element length correction factor for each pulse current data segment For the first The pulse width characterization factor for each pulse current data segment. For the first Power spectral entropy of each pulse current data segment This represents the maximum spectral entropy value within the historical pulse current data segment. This is an adjustment factor for the length correction coefficient of the structure element in the pulse current data segment. It is used to adjust the value range of the structure element length coefficient for the pulse current data segment, ensuring that the final length of the structure element obtained later can be adjusted by either increasing or decreasing the initial length. The empirical value range for this adjustment factor is as follows: In this embodiment, the adjustment coefficient is set to 0.5. In other embodiments, the implementer can set it according to the actual implementation situation. For example, when the maximum spectral entropy value in the pulse current historical data segment is large, the adjustment coefficient can be appropriately increased; when the maximum spectral entropy value in the pulse current historical data segment is small, the adjustment coefficient can be appropriately decreased. It is a natural exponential function. It is the maximum-minimum normalization function.

[0035] in, The larger the value, the higher the value of the pulse current data segment. This indicates that when analyzing the frequency centroid distribution and normalized bandwidth of the pulse current data segment, the first... The higher the probability that the pulse current data segment belongs to the near-end narrow pulse and is a real partial discharge signal, the more likely it is to be preserved. The smaller the length correction coefficient of the corresponding structural element of a pulse current data segment, the better. The larger the value, the more likely it is to be the first. The higher the frequency domain disorder of the signal within the pulse current data segment, the more it indicates that the... The more pronounced the noise characteristics within a pulse current data segment, the larger the structure element length correction coefficient should be for smoothing noise removal.

[0036] S004: Correct the initial length of the structure element of the pulse current data segment according to the structure element length correction coefficient to obtain the final length of the structure element of the pulse current data segment.

[0037] It should be noted that after obtaining the structural element length correction coefficient of the pulse current data segment, this invention will correct the initial length of the structural element of the pulse current data segment based on the structural element length correction coefficient. Traditional morphological filtering algorithms usually set the structural element length of the pulse current data segment to a default value of 10. This invention, however, uses the structural element length correction coefficient to correct the initial length of the structural element to obtain the final length of the structural element, so that the morphological filtering algorithm can extract the partial discharge signal more accurately.

[0038] Specifically, the final length of the structuring element satisfies the expression: ; In the formula, For the first The final length of the structure element for each pulse current data segment For the first The structure element length correction factor for each pulse current data segment The initial length of the structure element for the pulse current data segment, for example, , This represents the function for rounding up.

[0039] Among them, the The larger the structure element length correction coefficient of the pulse current data segment, the more it indicates that the... The greater the likelihood that the pulse current data segment belongs to background noise rather than the actual partial discharge pulse current, the more likely the second pulse current data segment will be. The final length of the structuring element for each pulse current data segment will also be larger, ensuring that the morphological filtering algorithm can filter out the first pulse current data segment. Noise in the pulse current data segment.

[0040] S005: Construct a flat structural element for the pulse current data segment based on the final length of the structural element, perform morphological filtering on the pulse current data segment based on the flat structural element, and determine the feature waveform extraction threshold based on the morphological filtering result in order to extract the partial discharge signal.

[0041] Specifically, extracting the partial discharge signal includes: Based on the final length of the structural element, a corresponding flat structural element is constructed. The method for obtaining the flat structural element is as follows: define a one-dimensional sequence with a length equal to the final length of the structural element, and assign the amplitude height at each point in the one-dimensional sequence to 0 to obtain the flat structural element.

[0042] It should be noted that by assigning the amplitude height of each point in the one-dimensional sequence to 0, this invention ensures that when probing the waveform characteristics of the pulse current data segment, only the geometric width of the window is used, without introducing additional amplitude deviation, thus ensuring that morphological operations can be matched and filtered based on the shape of the pulse current.

[0043] Based on a flat structural element, morphological filtering is performed on the pulse current data segment to obtain a denoised signal, which satisfies the expression: ; In the formula, For the first The first pulse current data segment The signal after denoising at each sampling point This is the index of the sampling point in the pulse current data segment. For opening operation, For closing operations, For the first The signal before denoising of each pulse current data segment For the first A flat structure element for each pulse current data segment.

[0044] It should be noted that this invention uses opening operations to suppress positive spike noise and closing operations to suppress negative spike noise. For the oscillation characteristics of partial discharge signals, taking the average of the two operations can effectively cancel bidirectional high-frequency background noise, avoid waveform distortion caused by single operations, and completely preserve the oscillation characteristics of the signal while smoothing and denoising.

[0045] Furthermore, a feature waveform extraction threshold is determined based on the morphological filtering results. The feature waveform extraction threshold satisfies the following expression: ; In the formula, For the first Threshold for extracting characteristic waveforms within a single pulse current data segment. For the first The first pulse current data segment The signal after denoising at each sampling point This represents the total number of sampling points in the pulse current data segment. This is the index of the sampling point in the pulse current data segment. For the first The arithmetic mean of the absolute values ​​of the signals after denoising each pulse current data segment.

[0046] It should be noted that this invention uses the mean of the denoised signal to characterize the background noise level of the pulse current data segment, and sets 4 times the mean as the feature waveform extraction threshold to ensure that the feature waveform extraction threshold can be adjusted in real time according to the fluctuation of environmental noise. While shielding background noise, it extracts the real partial discharge signal that is higher than the noise floor, thus taking into account both the sensitivity and robustness of detection.

[0047] Furthermore, the partial discharge signal is extracted by comparing the amplitude of each data point in the pulse current data segment after morphological filtering with the characteristic waveform extraction threshold. The extraction method of the partial discharge signal is as follows: traverse all data points in the pulse current data segment, mark all continuous time segments with absolute amplitude values ​​exceeding the characteristic waveform extraction threshold as characteristic waveforms, and record their start and end times to complete the extraction of the partial discharge signal.

[0048] The presence of a partial discharge signal in the pulse current data segment indicates the presence of potential defects such as air gaps or impurities in the insulation layer of the wind power harness, which may pose a risk of insulation breakdown. The product is therefore deemed unqualified and will be reworked.

[0049] like Figure 2 , Figure 3 and Figure 4 As shown, Figure 2 The results of morphological filtering using a fixed long structuring element are shown. Although the long structuring element has a good wrapping and smoothing effect on the wide pulse at the far end, for the narrow pulse at the near end in the first half, the structuring element length is much larger than the pulse span, resulting in severe morphological over-erosion. The amplitude height of the narrow pulse at the near end is greatly reduced and is significantly lower than the original signal peak. This amplitude attenuation can easily lead to the omission of weak early fault signals when setting the threshold. Figure 3 The results of morphological filtering using a fixed short structural element are shown. Although the filtered waveform retains the amplitude height of the near-end narrow pulse for the short structural element, its smoothing and noise reduction capabilities are severely insufficient, with a large number of glitches remaining at the baseline. More seriously, for the far-end wide pulse in the latter half, the structural element length cannot cover the signal width, causing the originally complete wide peak to be incorrectly cut into multiple sawtooth pseudo-peaks, resulting in waveform fragmentation. This phenomenon can mislead the detection system to interpret a single discharge event as multiple pulses, seriously affecting the accuracy of insulation performance assessment. Figure 4 The present invention demonstrates the filtering results achieved by using an adaptive structuring element length. The filtered waveform curve automatically matches a short structuring element in the first half, restoring the geometric characteristics and amplitude of the near-end narrow pulse. In the second half, it automatically matches a long structuring element, smoothly enveloping the far-end wide pulse and eliminating the fragmentation phenomenon. The entire waveform curve closely fits the envelope contour of the original signal while suppressing background noise interference, demonstrating that the present invention has excellent adaptability and robustness when processing partial discharge signals from different wind power line harnesses.

Claims

1. A method for testing the performance of wind power harnesses based on data processing, characterized in that, include: The pulse current data of the wind power harness is divided into several pulse current data segments; Determine the pulse width characterization factor of the pulse current data segment. The pulse width characterization factor is negatively correlated with the frequency centroid of the pulse current data segment and negatively correlated with the bandwidth of the pulse current data segment. Determine the structural element length correction coefficient for the pulse current data segment, wherein the structural element length correction coefficient is positively correlated with the power spectral entropy and pulse width characterization factor of the pulse current data segment; The initial length of the structure element of the pulse current data segment is corrected according to the structure element length correction coefficient to obtain the final length of the structure element of the pulse current data segment. Construct a flat structure element for the pulse current data segment based on the final length of the structure element; Morphological filtering of pulse current data segments is performed based on flat structural elements. The characteristic waveform extraction threshold is determined based on the morphological filtering results to extract partial discharge signals and realize the performance detection of wind power harnesses.

2. The wind power harness performance testing method based on data processing according to claim 1, characterized in that, The step of dividing the pulse current data of the wind power harness into several pulse current data segments includes: dividing the pulse current data of the wind power harness into several pulse current data segments of equal length using a sliding window.

3. The wind power harness performance testing method based on data processing according to claim 1, characterized in that, The method for obtaining the bandwidth of the pulse current data segment is as follows: the ratio of the frequency standard deviation of the pulse current data segment to the maximum bandwidth of the digital signal system is used as the bandwidth of the pulse current data segment.

4. The wind power harness performance testing method based on data processing according to claim 1 or 3, characterized in that, The pulse width characterization factor satisfies the expression: ； In the formula, For the first The pulse width characterization factor for each pulse current data segment. This represents the total number of frequency points in the pulse current data segment spectrum. For the first The pulse current data segment in the first Power spectral density values ​​at each frequency point For the first The pulse current data segment in the first Frequency values ​​at each frequency point For the first The frequency centroid of each pulse current data segment For the first The bandwidth of each pulse current data segment It is a natural exponential function.

5. The wind power harness performance testing method based on data processing according to claim 1, characterized in that, The length correction coefficient of the structuring element satisfies the expression: ； In the formula, For the first The structure element length correction factor for each pulse current data segment For the first The pulse width characterization factor for each pulse current data segment. For the first Power spectral entropy of each pulse current data segment This represents the maximum spectral entropy value within the historical pulse current data segment. The value of the structure element length correction coefficient for each pulse current data segment is adjusted by the coefficient. It is a natural exponential function. It is the maximum-minimum normalization function.

6. The wind power harness performance testing method based on data processing according to claim 1, characterized in that, The final length of the structuring element satisfies the expression: ； In the formula, For the first The final length of the structure element for each pulse current data segment For the first The structure element length correction factor for each pulse current data segment The initial length of the structure element for the pulse current data segment. This represents the function for rounding up.

7. The wind power harness performance testing method based on data processing according to claim 1, characterized in that, The method for obtaining the flat structural element is as follows: define a one-dimensional sequence with a length equal to the final length of the structural element, and assign the amplitude height at each point in the one-dimensional sequence to 0 to obtain the flat structural element.

8. The method for testing the performance of wind power harnesses based on data processing according to claim 1, characterized in that, The pulse current data segment is morphologically filtered to obtain a denoised signal, which satisfies the expression: ； In the formula, For the first The first pulse current data segment The signal after denoising at each sampling point This is the index of the sampling point in the pulse current data segment. For opening operation, For closing operations, For the first The signal before denoising of each pulse current data segment For the first A flat structure element for each pulse current data segment.

9. The method for testing the performance of wind power harnesses based on data processing according to claim 1, characterized in that, The feature waveform extraction threshold satisfies the expression: ； In the formula, For the first Threshold for extracting characteristic waveforms within a single pulse current data segment. For the first The first pulse current data segment The signal after denoising at each sampling point This represents the total number of sampling points in the pulse current data segment. This is the index of the sampling point in the pulse current data segment. For the first The arithmetic mean of the absolute values ​​of the signals after denoising each pulse current data segment.

10. The method for testing the performance of wind power harnesses based on data processing according to claim 1, characterized in that, The extraction of partial discharge signals includes: traversing all pulse current data segments, marking all continuous time segments within the pulse current data segments whose absolute amplitude exceeds the characteristic waveform extraction threshold as characteristic waveforms, and recording their start and end times, thereby completing the extraction of partial discharge signals from the wind power harness and realizing the performance detection of the wind power harness.

Images