A method and system for identifying and calibrating fault traveling waves suitable for medium voltage distribution networks

Abstract

A kind of identification calibration method and system suitable for medium voltage distribution network fault travelling wave, for the problems that existing technology is difficult to adapt to complex and varied wave front shape, calibration precision and robustness is insufficient, the present application first adopts medium voltage distribution network travelling wave signal and pre-processes, constructs five types of wave front shape vectors;Then design multi-dimensional classification criterion to divide wave front into five categories, realize online calibration of classification threshold value through threshold value dynamic updating mechanism;Finally, for different types of wave front, corresponding adaptive calibration method is used to complete accurate calibration.The present application can effectively solve the problems of wave front calibration lag, reflection wave misselection, strong noise interference and other problems, adapt to complex working conditions of medium voltage distribution network, and have excellent engineering landing performance.

Description

Technical Field

[0001] This invention belongs to the field of power system relay protection and fault location technology, and particularly relates to a method and system for identifying and calibrating fault traveling waves applicable to medium-voltage distribution networks. Background Technology

[0002] Currently, in the field of power system relay protection and fault location technology, the traveling wave location method has received widespread attention and application in medium-voltage distribution networks due to its advantages such as being unaffected by line parameters and having high location accuracy. Existing traveling wave location technology mainly relies on capturing the arrival time of the initial traveling wave (i.e., the "wave front") and calculating the transmission time of the traveling wave between the fault point and the monitoring point to achieve fault location.

[0003] For example, patent application CN114966320A constructs a fault simulation model, verifies the simulation model, generates simulation data, adds noise preprocessing, and trains a fault segment judgment model. It leverages the advantage of recurrent neural networks in capturing dependencies in time-series data, inputs fault simulation samples, and trains the neural network model to learn the mapping relationship between the fault traveling wave signal measured at the beginning of the power supply arm and the fault occurrence segment, establishing a fault segment judgment model. The detected fault traveling wave signal is input into the trained fault segment judgment model to determine the fault occurrence segment. Then, the distance between the fault point and the beginning of the fault segment is calculated using the time difference between the initial wavefront and subsequent reflected waves arriving at the measurement point in the fault traveling wave, ultimately determining the location of the fault distance measurement point. Another example is patent application CN108196166A, which performs zero-phase-shift digital filtering on the traveling wave waveform data, selects a reference point before the waveform's starting point is reached, and establishes a straight line equation using the waveform's maximum value point. The point with the largest distance between two points on the waveform reaching this straight line is calculated as the starting point of the traveling wave waveform. The fault location is then calculated based on the principle of double-ended traveling wave fault ranging.

[0004] However, the waveform of traveling waves in engineering practice is affected by multiple factors such as the complexity of the line structure, frequent branch reflections, strong electromagnetic noise, and differences in fault characteristics. It often exhibits diverse and irregular waveforms, including clean, jittery, gradually changing, stepped superposition, and noise submersion. Traditional ranging algorithms are generally based on fixed thresholds, single differential operators, or ideal waveform templates. Their recognition logic relies on uniform parameter settings and typical waveform assumptions, making it difficult to adapt to the highly heterogeneous actual waveform characteristics. This results in insufficient stability of the recognition results under different operating conditions, often leading to premature or delayed waveform calibration, and consequently causing significant ranging errors.

[0005] Current technologies lack a systematic mechanism for summarizing and classifying complex wavefront morphologies, relying heavily on empirical criteria for feature selection and signal shaping. This limitation is particularly pronounced in complex scenarios such as multi-branch networks, weak fault signals, and strong noise interference, often resulting in weakened effective signal leading edges or inappropriate amplification of interference components, further reducing the reliability of identification. With the increasing penetration of distributed power sources in distribution networks, the widespread application of overhead-cable hybrid lines, and the increasingly complex electromagnetic disturbance environment, the shortcomings of traditional traveling wave wavefront extraction methods in terms of universality and adaptability are becoming increasingly apparent.

[0006] In summary, existing traveling wave head processing technologies suffer from three fundamental limitations: inability to effectively distinguish different morphological features, lack of targeted processing strategies, and absence of a dynamic response mechanism. These deficiencies make it difficult to meet the stringent requirements for accurate wave head calibration in fault location under the complex environment of medium-voltage distribution networks, and have become a major bottleneck restricting the reliable engineering application of traveling wave ranging technology. To address this challenge, there is an urgent need to establish an intelligent processing method based on adaptive morphological feature recognition. By constructing a systematic wave head morphological feature library and a dynamic algorithm selection mechanism, the accuracy and robustness of traveling wave head recognition can be fundamentally improved. Summary of the Invention

[0007] To address the challenge that existing single algorithms cannot handle complex and variable fault waveforms, the following core issues are specifically addressed:

[0008] (1) Wavehead calibration lag problem: For slowly changing waveforms under high impedance faults, the traditional modulus maxima method cannot accurately lock the starting point.

[0009] (2) Problem of incorrect selection of reflected waves: In response to near-end faults and oscillating waveforms, solve the problem of misjudgment caused by signal superposition or jitter, and eliminate the dead zone of ranging.

[0010] (3) Strong noise interference problem: For weak signals at the end, solve the problem of not being able to effectively extract fault features under extremely low signal-to-noise ratio.

[0011] This invention provides a method for complex wavefront morphology identification and adaptive calibration of traveling waves in medium-voltage distribution networks. Instead of relying on a single abrupt change feature, this method constructs a refined wavefront morphology representation system, classifying fault traveling waves into five categories: clean, jittery, slowly varying, stepped superposition, and noise-overwhelmed. For different wavefront categories, this invention introduces multi-structure feature set fusion, wavefront dynamic reachability domain construction, adaptive search strategies, and robust time-frequency joint criteria to achieve accurate pinpointing of the initial wavefront time.

[0012] The present invention adopts the following technical solution.

[0013] This invention provides a method for identifying and calibrating fault traveling waves applicable to medium-voltage distribution networks, comprising:

[0014] S1. Acquire traveling wave signals from medium-voltage distribution networks and preprocess the traveling wave signals; construct wavefront morphology vectors based on the preprocessed traveling wave signals; the wavefront morphology vectors include peak amplitude features, instantaneous slope mean features, jitter variance features, local linearity features, and noise ratio features;

[0015] S2, design classification criteria for different wave heads, classify and divide the wave head morphology vector based on the classification criteria, and obtain wave head classification results; wherein, the classification threshold in the classification criteria is updated through a threshold dynamic update mechanism;

[0016] S3. Based on the wavefront division results, perform wavefront calibration using the corresponding adaptive calibration method.

[0017] More preferably, the method for constructing the classification criteria for different types of waveheads includes:

[0018] Strong noise type is characterized by a noise ratio exceeding the noise ratio threshold.

[0019] The clean type is characterized by peak amplitude exceeding the abrupt change amplitude threshold, instantaneous slope mean exceeding the slope threshold, and jitter variance below the jitter variance threshold.

[0020] The jitter type is characterized by peak amplitude exceeding the abrupt change amplitude threshold and jitter variance not lower than the jitter variance threshold;

[0021] The gradual change type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold, instantaneous slope mean characteristics not exceeding the slope threshold, and local linearity characteristics not exceeding the linearity threshold.

[0022] The stepped superposition type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold and local linearity characteristics exceeding the linearity threshold.

[0023] More preferably, the noise ratio threshold is calculated as follows:

[0024] Several noise scenarios are sampled to estimate the noise energy distribution, and a set percentile value of the normal wavefront energy distribution is statistically calculated. Based on the noise energy distribution, the noise energy mean and noise energy standard deviation are calculated, and combined with the set percentile value, the noise ratio threshold is calculated. The background noise energy of the current line is sampled, and a smoothing factor is set. The calibrated noise ratio threshold is obtained by combining the background noise energy and the smoothing factor.

[0025] The mutation amplitude threshold is calculated as follows: the mutation amplitude threshold is obtained based on the inter-class variance function and the set self-adjustment constant.

[0026] The slope threshold is calculated as follows: the slope threshold is obtained based on the standard deviation of derivative noise and a set coefficient.

[0027] The jitter variance threshold is calculated as follows: the jitter variance threshold is calculated based on the estimated steady-state noise variance and jitter amplification factor when there is no fault; when considering the changes in the background noise on site, the corrected jitter variance threshold is obtained based on the current real-time estimated background noise variance.

[0028] The linearity threshold is calculated as follows: Gaussian mixture model is used to divide the model, local linearity features are modeled, and the midpoint of the two Gaussian centers is calculated by the model as the linearity threshold.

[0029] More preferably, the calibration method for the noise ratio threshold is as follows: sampling the current line background noise energy, combining the historical statistical noise energy mean with the set percentile value, and calibrating the noise ratio threshold online;

[0030] The calibration method for the mutation amplitude threshold is as follows: the mutation amplitude threshold is calculated by combining the mutation amplitude threshold with the self-adjustment constant, and the mutation amplitude threshold is calibrated online.

[0031] The calibration method for the slope threshold is as follows: the set coefficient is adjusted over time, and the slope threshold is calibrated online using normalization.

[0032] The calibration method for the jitter variance threshold is as follows: when the background noise changes, the jitter variance threshold is calculated by combining the ratio of the noise standard deviation when the background noise changes to the steady-state noise standard deviation, and the jitter variance threshold is calibrated online.

[0033] The calibration method for the linearity threshold is as follows: the linearity threshold is calibrated online using the local linearity feature model.

[0034] More preferably, the implementation method of the threshold dynamic update mechanism includes:

[0035] The threshold for the next time step is equal to the difference between the current time step threshold multiplied by 1 and the smoothing factor, plus the product of the actual statistical value calculated from the latest wavehead detections and the smoothing factor.

[0036] More preferably, the adaptive calibration method for different types of wavefronts is as follows:

[0037] A rapid calibration method based on the differential-slope joint abrupt change criterion is adopted for clean wavefronts;

[0038] A calibration method combining envelope construction and abrupt change interval contraction is used for jittery wavefronts;

[0039] A calibration method combining relative rate of change and cumulative increment is used for slowly varying wavefronts;

[0040] A calibration method is adopted for stepped superposition wavefronts, which involves finding step abrupt change points, clustering steps, and selecting the earliest strong step.

[0041] A calibration method combining continuous wavelet transform to enhance abrupt changes and extraction of the maximum abrupt change point of scale energy is used for the noise-saturated wavefront.

[0042] This invention also proposes an identification and calibration system for fault traveling waves in medium-voltage distribution networks, comprising a morphological vector construction module, a waveform category classification module, and an adaptive calibration module, characterized in that:

[0043] The morphology vector construction module acquires traveling wave signals from a medium-voltage distribution network and preprocesses the traveling wave signals; it then constructs a wavefront morphology vector based on the preprocessed traveling wave signals; the wavefront morphology vector includes peak amplitude characteristics, instantaneous slope mean characteristics, jitter variance characteristics, local linearity characteristics, and noise ratio characteristics.

[0044] The waveform classification module is designed with classification criteria for different wavefronts. Based on the classification criteria, the wavefront morphology vector is classified to obtain the wavefront classification result. The classification threshold in the classification criteria is updated through a dynamic threshold update mechanism.

[0045] The adaptive calibration module performs wavehead calibration using the corresponding adaptive calibration method based on the wavehead division results.

[0046] More preferably, in the waveform classification module, the method for constructing the classification criteria for different types of wavefronts includes:

[0047] Strong noise type is characterized by a noise ratio exceeding the noise ratio threshold.

[0048] The clean type is characterized by peak amplitude exceeding the abrupt change amplitude threshold, instantaneous slope mean exceeding the slope threshold, and jitter variance below the jitter variance threshold.

[0049] The jitter type is characterized by peak amplitude exceeding the abrupt change amplitude threshold and jitter variance not lower than the jitter variance threshold;

[0050] The gradual change type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold, instantaneous slope mean characteristics not exceeding the slope threshold, and local linearity characteristics not exceeding the linearity threshold.

[0051] The stepped superposition type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold and local linearity characteristics exceeding the linearity threshold.

[0052] The present invention also proposes a terminal, including a processor and a storage medium:

[0053] The storage medium is used to store instructions;

[0054] The processor is configured to operate according to the instructions to perform the steps according to the method described above.

[0055] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0057] 1. The concept of "wavehead morphology vector" is proposed, which can quantitatively describe the multidimensional characteristics of the wavehead leading edge, enabling the classification and analysis of complex morphologies.

[0058] 2. Construct an adaptive calibration strategy library to automatically select the best calibration model based on the morphology, rather than using a fixed threshold or uniform operator.

[0059] 3. A dynamic reachability mechanism for wavefronts is designed to effectively eliminate reflected wave interference and significantly reduce misjudgments in near-end fault scenarios.

[0060] 4. Introducing a robust time-frequency joint criterion to maintain wavefront detection stability even at extremely low signal-to-noise ratios.

[0061] 5. The entire process is feasible for engineering scenarios and can be directly deployed in power distribution automation terminals, station devices, or master station systems. Attached Figure Description

[0062] Figure 1 This is a flowchart of an identification and calibration method for fault diagnosis in medium-voltage power distribution networks according to the present invention;

[0063] Figure 2 This is a flowchart illustrating the overall technical process of the present invention;

[0064] Figure 3 This is a diagram illustrating the construction of wavehead morphology vectors in this invention;

[0065] Figure 4 This is a diagram of the waveform classification method of the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0067] The present invention proposes the following technical solution:

[0068] like Figure 1 As shown, this invention proposes a method for identifying and calibrating fault traveling waves suitable for medium-voltage distribution networks:

[0069] Traveling wave signals from a medium-voltage distribution network are acquired and preprocessed. A wavefront morphology vector is constructed based on the preprocessed traveling wave signals. The wavefront morphology vector includes peak amplitude characteristics, instantaneous slope mean characteristics, jitter variance characteristics, local linearity characteristics, and noise ratio characteristics.

[0070] The design proposes classification criteria for different wave heads, and classifies the wave head morphology vectors based on the classification criteria to obtain wave head classification results; wherein, the classification threshold in the classification criteria is updated through a threshold dynamic update mechanism.

[0071] The methods for constructing the classification criteria for different types of wave heads include:

[0072] Strong noise type is characterized by a noise ratio exceeding the noise ratio threshold.

[0073] The clean type is characterized by peak amplitude exceeding the abrupt change amplitude threshold, instantaneous slope mean exceeding the slope threshold, and jitter variance below the jitter variance threshold.

[0074] The jitter type is characterized by peak amplitude exceeding the abrupt change amplitude threshold and jitter variance not lower than the jitter variance threshold;

[0075] The gradual change type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold, instantaneous slope mean characteristics not exceeding the slope threshold, and local linearity characteristics not exceeding the linearity threshold.

[0076] The stepped superposition type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold and local linearity characteristics exceeding the linearity threshold.

[0077] The method for calculating the noise ratio threshold is as follows:

[0078] Several noise scenarios are sampled to estimate the noise energy distribution, and a set percentile value of the normal wavefront energy distribution is statistically calculated. Based on the noise energy distribution, the noise energy mean and noise energy standard deviation are calculated, and combined with the set percentile value, the noise ratio threshold is calculated. The background noise energy of the current line is sampled, and a smoothing factor is set. The calibrated noise ratio threshold is obtained by combining the background noise energy and the smoothing factor.

[0079] The mutation amplitude threshold is calculated as follows: the mutation amplitude threshold is obtained based on the inter-class variance function and the set self-adjustment constant.

[0080] The slope threshold is calculated as follows: the slope threshold is obtained based on the standard deviation of derivative noise and a set coefficient.

[0081] The jitter variance threshold is calculated as follows: the jitter variance threshold is calculated based on the estimated steady-state noise variance and jitter amplification factor when there is no fault; when considering the changes in the background noise on site, the corrected jitter variance threshold is obtained based on the current real-time estimated background noise variance.

[0082] The linearity threshold is calculated as follows: Gaussian mixture model is used to divide the model, local linearity features are modeled, and the midpoint of the two Gaussian centers is calculated by the model as the linearity threshold.

[0083] The calibration method for the noise ratio threshold is as follows: sample the current line background noise energy, combine it with the historical statistical noise energy mean and the set percentile value, and calibrate the noise ratio threshold online.

[0084] The calibration method for the mutation amplitude threshold is as follows: the mutation amplitude threshold is calculated by combining the mutation amplitude threshold with the self-adjustment constant, and the mutation amplitude threshold is calibrated online.

[0085] The calibration method for the slope threshold is as follows: the set coefficient is adjusted over time, and the slope threshold is calibrated online using normalization.

[0086] The calibration method for the jitter variance threshold is as follows: when the background noise changes, the jitter variance threshold is calculated by combining the ratio of the noise standard deviation when the background noise changes to the steady-state noise standard deviation, and the jitter variance threshold is calibrated online.

[0087] The calibration method for the linearity threshold is as follows: the linearity threshold is calibrated online using the local linearity feature model.

[0088] The implementation method of the threshold dynamic update mechanism includes:

[0089] The threshold for the next time step is equal to the difference between the current time step threshold multiplied by 1 and the smoothing factor, plus the product of the actual statistical value calculated from the latest wavehead detections and the smoothing factor.

[0090] Based on the wavefront division results, wavefront calibration is performed using the corresponding adaptive calibration method.

[0091] The adaptive calibration methods for different types of wavefronts are as follows:

[0092] A rapid calibration method based on the differential-slope joint abrupt change criterion is adopted for clean wavefronts;

[0093] A calibration method combining envelope construction and abrupt change interval contraction is used for jittery wavefronts;

[0094] A calibration method combining relative rate of change and cumulative increment is used for slowly varying wavefronts;

[0095] A calibration method is adopted for stepped superposition wavefronts, which involves finding step abrupt change points, clustering steps, and selecting the earliest strong step.

[0096] A calibration method combining continuous wavelet transform to enhance abrupt changes and extraction of the maximum abrupt change point of scale energy is used for the noise-saturated wavefront.

[0097] This invention also proposes an identification and calibration system for fault traveling waves in medium-voltage distribution networks, including a morphological vector construction module, a waveform category classification module, and an adaptive calibration module.

[0098] The morphology vector construction module acquires traveling wave signals from a medium-voltage distribution network and preprocesses the traveling wave signals; it then constructs a wavefront morphology vector based on the preprocessed traveling wave signals; the wavefront morphology vector includes peak amplitude characteristics, instantaneous slope mean characteristics, jitter variance characteristics, local linearity characteristics, and noise ratio characteristics.

[0099] The waveform classification module is designed with classification criteria for different wavefronts. Based on the classification criteria, the wavefront morphology vector is classified to obtain the wavefront classification result. The classification threshold in the classification criteria is updated through a dynamic threshold update mechanism.

[0100] The adaptive calibration module performs wavehead calibration using the corresponding adaptive calibration method based on the wavehead division results.

[0101] In the waveform classification module, the method for constructing the classification criteria for different types of wavefronts includes:

[0102] Strong noise type is characterized by a noise ratio exceeding the noise ratio threshold.

[0103] The clean type is characterized by peak amplitude exceeding the abrupt change amplitude threshold, instantaneous slope mean exceeding the slope threshold, and jitter variance below the jitter variance threshold.

[0104] The jitter type is characterized by peak amplitude exceeding the abrupt change amplitude threshold and jitter variance not lower than the jitter variance threshold;

[0105] The gradual change type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold, instantaneous slope mean characteristics not exceeding the slope threshold, and local linearity characteristics not exceeding the linearity threshold.

[0106] The stepped superposition type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold and local linearity characteristics exceeding the linearity threshold.

[0107] The present invention also proposes a terminal, including a processor and a storage medium:

[0108] The storage medium is used to store instructions;

[0109] The processor is configured to operate according to the instructions to perform the steps according to the method described above.

[0110] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0111] Example 1

[0112] This invention proposes a method for identifying and calibrating fault traveling waves applicable to medium-voltage distribution networks, such as... Figure 2 As shown, the specific steps are as follows:

[0113] Step 1: Constructing the wave head morphology vector

[0114] This invention first establishes a set of wavefront morphology vectors suitable for traveling wave data in medium-voltage distribution networks. This vector accurately characterizes the local morphology of the initial arrival leading edge of the traveling wave and serves as the basis for subsequent morphology classification and adaptive calibration. The vector consists of five independent, physically meaningful, and engineering-implementable feature quantities, capable of covering typical wavefront morphologies such as clean, jittery, gradually varying, stepped superposition, and noise-overwhelmed patterns. Figure 3 As shown.

[0115] (1) Input and output definitions

[0116] Input: A single-wave traveling voltage or current signal with a sampling frequency ≥ 1MHz (t);

[0117] Output: Wavehead shape vector The vector dimension is fixed, providing a unified input for subsequent classifiers.

[0118] (2) Signal preprocessing

[0119] To avoid noise dominating the characteristic quantity, the input traveling wave signal is subjected to:

[0120] ① Zero bias correction:

[0121]

[0122] In the formula, This is the mean.

[0123] ② Bandpass filtering (typically 20–500kHz): preserves the effective frequency band of the traveling wave.

[0124] ③Amplitude normalization:

[0125]

[0126] Standardized signal obtained after preprocessing .

[0127] (3) Construction methods of five types of core feature quantities

[0128] ① Peak amplitude characteristics

[0129]

[0130] In the formula, These represent the start and end times of the wavefront analysis window, respectively. Used to characterize the intensity of abrupt changes at the leading edge of the wavefront, reflecting whether the waveform is "clean" or masked by noise; x(t) represents the normalized signal after preprocessing.

[0131] ② Instantaneous slope mean characteristics

[0132]

[0133] In the formula, Indicates the sampling time. Indicates the number of sampling points. Indicates the sampling period. Used to characterize the average rate of change of the rising segment of the wavefront; a large slope indicates a steep wavefront; a small slope indicates a gradual change or high resistance characteristics.

[0134] ③ Jitter variance characteristics

[0135]

[0136] In the formula, This represents the mean within the window. Used to identify whether there is obvious oscillation or weak reflection superposition in the wavefront. The larger the value, the more "shaky" the wavefront is. Reflection superposition and high-frequency oscillation of the electric arc are particularly obvious.

[0137] ④ Local linearity characteristics

[0138] Fitting a line using least squares within the window:

[0139]

[0140] The linearity index is defined as the fitting error:

[0141]

[0142] In the formula, , The coefficients are the least squares estimates. Used to distinguish between stepped superposition and clean, steeply changing wavefronts. Note: The waveform is approximately linear (suitable for gradually changing / stepped waveforms). This indicates the presence of strong mutations or noise.

[0143] ⑤ Noise ratio characteristics

[0144]

[0145] In the formula, For indicator functions, This is the noise threshold (estimated statistically from background noise). It reflects the proportion of noise within the wavefront window. A large value indicates that the wavefront signal is overwhelmed by noise, and it is suitable for identifying noisy wavefronts.

[0146] Step 2, Waveform Classification Method

[0147] Based on the wave head morphology vector constructed above This invention proposes a wavefront morphology adaptive classification method that can be directly applied in engineering, used to classify traveling wavefronts of various complex faults in medium-voltage distribution networks, providing a basis for subsequent adaptive calibration strategies. The overall idea is to quantitatively characterize the differences in wavefront abrupt changes, smoothness, jitter, and noise ratios using different features in the morphology vector. By constructing multi-interval, multi-constraint classification criteria, it achieves accurate classification of five typical wavefront types, such as... Figure 4 As shown.

[0148] (1) Classification criterion design

[0149] ①High-noise type

[0150] When the noise ratio characteristic is high, it is directly identified as a strong noise type:

[0151]

[0152] At this point, the true wavefront energy has been masked by noise, and subsequent strong noise suppression and cumulative mutation calibration strategies are required.

[0153] ② Clean type

[0154] It satisfies the following characteristics: strong mutation, large slope, and small jitter.

[0155]

[0156] This waveform exhibits a typical spike structure, with significantly weaker reflection interference and noise, making it the easiest waveform to calibrate.

[0157] ③ Jitter type

[0158] Strong mutations, but accompanied by significant jitter or short-period oscillations:

[0159]

[0160] Such traveling waves are often seen in near-end faults, continuous arc oscillations, or superimposed capacitor reflections, requiring calibration strategies that employ jitter suppression and envelope reconstruction.

[0161] ④ Gradual change type

[0162] Weak mutation rate, small slope, and high linearity:

[0163]

[0164] This type of wavefront typically exists in scenarios such as high-resistance contact and pre-breakdown of cable faults. It presents a smooth transition shape and is the type that is most difficult to extract accurately using traditional methods.

[0165] ⑤ Stepped stacking type

[0166] The mutation is not obvious, the slope is moderate, but the linearity is poor, and it is characterized by the superposition of several gentle step jumps:

[0167]

[0168] Such wavefronts are typically found in overhead-cable hybrid lines, multi-branch reflections, or multi-path propagation superpositions, and must be calibrated using multi-order catastrophe decomposition and main step extraction methods.

[0169] (2) Methods and basis for determining classification thresholds

[0170] ① Noise ratio threshold (Note: Corresponding noise ratio feature) )

[0171] 1. Offline statistics

[0172] Sampling of a large number of noisy scenes to estimate noise energy distribution Simultaneously, the energy distribution of normal wavefronts was statistically analyzed. 5th percentile Set the initial threshold:

[0173]

[0174] In the formula, This represents the average noise energy (historical statistics). This represents the standard deviation of noise energy.

[0175] 2. Online calibration

[0176] Sample the background noise energy of the current line ,but:

[0177]

[0178] in This is a smoothing factor.

[0179] If none of the above processes can be performed, the typical range of values ​​in engineering is as follows: .

[0180] ② Mutation amplitude threshold (Note: Corresponding to peak amplitude characteristics) )

[0181] Mismatched reflections result in a bimodal distribution of wavefront amplitude abrupt changes. This invention employs Otsu's method to locate the segmentation point.

[0182]

[0183] In the formula, This represents the inter-class variance function in Otsu's method, used to find the optimal split point. Based on the seasonal variation of the fundamental amplitude of the current line, the following is adopted:

[0184]

[0185] in , which is a self-adjusting constant. If the above process cannot be performed, the typical range of values ​​in engineering is: .

[0186] ③ Slope threshold (Note: Corresponding to the instantaneous slope mean feature) )

[0187] The baseline scale for slope is the standard deviation of derivative noise. Given, therefore take

[0188]

[0189] Among them, the set coefficient It is recommended to set the initial value to 5.

[0190] Preferably, the lower limit is set to 3 (to prevent false alarms): if <3, the threshold is too low, and a large number of random background noise fluctuations will be misjudged as the wavefront of a fault traveling wave; the upper limit is set to 8 (to prevent missed detections): there are a large number of high-resistance faults or remote faults in medium-voltage distribution networks. The traveling waves generated by these faults attenuate over long distances, and by the time they reach the detection point, the wavefronts have become gentler. If the value is set too high (e.g., exceeding 8), the threshold will be too high, causing these weak but real fault signals to fail to trigger detection. A recommended initial value of 5 is a robust trade-off between "extremely low false alarm rate" and "sufficient sensitivity".

[0191] If the sampling rate is very high, the derivative noise will be amplified, therefore Should follow Adjustments can be made using a normalized form:

[0192]

[0193] ④ Jitter variance threshold (Note: Corresponding to jitter variance feature) )

[0194] Estimating steady-state noise variance under fault-free conditions ,but:

[0195]

[0196] In the formula, This is the jitter amplification factor. When considering changes in ambient background noise:

[0197]

[0198] in, This represents the current real-time estimated variance of the background noise.

[0199] ⑤ Linearity threshold (Note: Corresponding to local linearity feature) )

[0200] Partitioning was performed using a Gaussian mixture model. Modeling:

[0201]

[0202] The midpoint of the two Gaussian centers is used as the threshold:

[0203]

[0204] The typical range of values ​​in engineering is as follows: .

[0205] (3) Threshold dynamic update mechanism

[0206] To ensure the long-term stability of the threshold in actual field conditions, this invention adopts the following online update rules:

[0207]

[0208] In the formula, This is based on the actual statistical values ​​calculated from the latest 5 to 10 wavefront tests. As a smoothing factor, The specific value depends on the stability of the line environment. For lines with stable background noise, the value is biased towards 0.05 to enhance anti-interference capability; for lines with frequent load fluctuations or complex environments, the value is biased towards 0.15 to improve the tracking speed of environmental changes and ensure stable updates without oscillation.

[0209] in, This represents the "current true level estimate" obtained based on the most recent 5-10 wavefront detection data. To avoid drastic fluctuations in the threshold due to random errors in a single detection (such as random impulse interference or bad data points), the arithmetic mean cannot be simply used. This invention uses a truncated average method for comprehensive calculation:

[0210] In the most recent N (e.g., N=10) detections, remove the maximum and minimum values, and then average the remaining values. Let the feature value of the most recent N detected values ​​be... Sort them and then remove them. and ,but

[0211]

[0212] After the above process, extreme outliers can be effectively eliminated, preventing the threshold from being incorrectly raised or lowered due to a single false detection.

[0213] Based on the above steps, the waveform category is determined, and the corresponding traveling wave signal with the waveform shape is obtained, including the continuous signal x. f (t) and discrete signal x f (n).

[0214] Step 3: Traveling Wave Wavehead Adaptive Calibration Method

[0215] (1) Calibration method of clean wave head

[0216] Clean wavefronts are characterized by large abrupt changes, steep slopes, and concentrated energy. These wavefronts can be rapidly calibrated using a combined difference-slope abrupt change criterion. The wavefront position satisfies:

[0217]

[0218] In the formula:

[0219] , which is a first-order difference used to characterize the degree of abrupt change. For discrete traveling wave signals corresponding to waveform shapes;

[0220] , where is the sliding slope, and the window length is . ,recommend ,like If the slope is too small (e.g., less than 3), the slope calculation becomes overly sensitive to high-frequency noise, easily leading to false alarms; if... If the value is too large (e.g., greater than 8), a smoothing effect will occur, blurring the wavefront features and introducing a significant detection delay. Therefore, at typical sampling frequencies (1MHz~5MHz), taking 3~8 points can achieve the best balance between noise immunity and detection accuracy.

[0221] , For root mean square operations, generally ;

[0222] This is used to balance mutation intensity and slope.

[0223] To avoid a 1–2 point deviation caused by the sampling interval:

[0224]

[0225] (2) Diverging wave head calibration method

[0226] The abrupt change position of the jittery wavefront is disturbed by high-frequency oscillations; directly using the differential extremum would result in the wavefront falling at the oscillation peak. Therefore, the "envelope construction + abrupt change interval contraction" method is adopted. First, the envelope energy is constructed:

[0227]

[0228] In the formula, n represents the current discrete sampling index (time point), x f [n] represents the discrete sampled value (normalized amplitude of voltage or current) of the traveling wave signal judged as "jittery" at time n, and M represents the sliding window length of the energy integral class. Small window size suppresses the effects of jitter. Then, the starting point of the energy rise is located.

[0229]

[0230] in, Finally, the oscillation range narrows, and within the range... Point with maximum slope in the internal search:

[0231]

[0232] Since the energy window length is M, the theoretical maximum lag caused by the energy criterion is approximately M sampling points. Therefore, L must be greater than or equal to M. Preferably, in this invention, the value range of L is set to [5, 10].

[0233] (3) Gradually changing wave head calibration method

[0234] Gradually varying wavefronts are one of the key challenges addressed by this invention. These signals exhibit low rates of change and subtle abrupt changes, making it difficult for traditional differential methods to capture their true starting point. This method is based on the criterion of "relative rate of change + cumulative increment".

[0235] ① Define relative rate of change

[0236]

[0237] It is a constant used to prevent the denominator from being too small, and is generally taken as 10⁻⁶.

[0238] ② Criterion for Gradually Changing Waveheads

[0239]

[0240] Where the threshold is:

[0241]

[0242] : The interval between experience and simulation verification.

[0243] ③ Cumulative Incremental Constraints

[0244] To avoid random disturbances from satisfying the conditions, the following must be met:

[0245]

[0246] In the formula, For cumulative windows, a value of 5 to 10 is typically used; It is a constant, determined statistically based on the wavehead morphology vector characteristics.

[0247] (4) Stepped superposition wavehead calibration method

[0248] The superimposed wavefronts of the stepped wavefronts are characterized by multiple ascending steps. The first step often corresponds to the true wavefront, while the others are branch reflections. Traditional methods are prone to mistakenly selecting subsequent steps with larger amplitudes.

[0249] ① Find all step transition points

[0250] Using the sliding slope:

[0251]

[0252] ② Step clustering

[0253] Calculate the local amplitude for all abrupt change points:

[0254]

[0255] Using simple threshold clustering, mutation points are categorized into "strong steps" and "weak steps" based on their magnitude:

[0256]

[0257] in For all of the median.

[0258] ③ Select the earliest appearing strong step as the wave head.

[0259]

[0260] (5) Noise-saturated wavefront calibration method

[0261] ① Continuous wavelet transform enhances abrupt change

[0262] Using mother wavelet :

[0263]

[0264] In the formula, 'a' represents a small-scale enhancement of high-frequency mutations. , The shift factor corresponds to a moment on the time axis and is used to determine the position of the wavelet function in time. b = 1, 2, ..., N (N represents the total number of signal sampling points). This represents the energy distribution in the scale domain.

[0265] ② Extract the maximum energy mutation point at scale

[0266] Define scale energy:

[0267]

[0268] ③ The wave head position is the point where energy first rises rapidly.

[0269]

[0270] In the formula, the threshold .

[0271] Example 2

[0272] This invention also proposes an identification and calibration system for fault traveling waves in medium-voltage distribution networks, including a morphological vector construction module, a waveform category classification module, and an adaptive calibration module.

[0273] The morphology vector construction module acquires traveling wave signals from a medium-voltage distribution network and preprocesses the traveling wave signals; it then constructs a wavefront morphology vector based on the preprocessed traveling wave signals; the wavefront morphology vector includes peak amplitude characteristics, instantaneous slope mean characteristics, jitter variance characteristics, local linearity characteristics, and noise ratio characteristics.

[0274] The waveform classification module is designed with classification criteria for different wavefronts. Based on the classification criteria, the wavefront morphology vector is classified to obtain the wavefront classification result. The classification threshold in the classification criteria is updated through a dynamic threshold update mechanism.

[0275] The adaptive calibration module performs wavehead calibration using the corresponding adaptive calibration method based on the wavehead division results.

[0276] Example 3

[0277] The present invention also proposes a terminal, including a processor and a storage medium:

[0278] The storage medium is used to store instructions;

[0279] The processor is configured to operate according to the instructions to perform the steps of the method according to Embodiment 1.

[0280] Example 4

[0281] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in Embodiment 1.

[0282] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0283] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0284] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0285] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0286] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for identifying and calibrating fault traveling waves applicable to medium-voltage distribution networks, characterized in that, include: S1. Acquire traveling wave signals from medium-voltage distribution networks and preprocess the traveling wave signals; construct wavefront morphology vectors based on the preprocessed traveling wave signals; the wavefront morphology vectors include peak amplitude features, instantaneous slope mean features, jitter variance features, local linearity features, and noise ratio features; S2, design classification criteria for different wave heads, classify and divide the wave head morphology vector based on the classification criteria, and obtain wave head classification results; wherein, the classification threshold in the classification criteria is updated through a threshold dynamic update mechanism; The methods for constructing the classification criteria for different types of wave heads include: Strong noise type is characterized by a noise ratio exceeding the noise ratio threshold. The clean type is characterized by peak amplitude exceeding the abrupt change amplitude threshold, instantaneous slope mean exceeding the slope threshold, and jitter variance below the jitter variance threshold. The jitter type is characterized by peak amplitude exceeding the abrupt change amplitude threshold and jitter variance not lower than the jitter variance threshold; The gradual change type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold, instantaneous slope mean characteristics not exceeding the slope threshold, and local linearity characteristics not exceeding the linearity threshold. The stepped superposition type is characterized by peak amplitude features not exceeding the abrupt change amplitude threshold and local linearity features exceeding the linearity threshold; S3. Based on the wavefront division results, perform wavefront calibration using the corresponding adaptive calibration method.

2. The method for identifying and calibrating fault traveling waves in medium-voltage distribution networks according to claim 1, characterized in that: The method for calculating the noise ratio threshold is as follows: Sample several noise scenarios, estimate the noise energy distribution, and statistically analyze the set percentile value of the normal wavefront energy distribution; The noise ratio threshold is calculated based on the noise energy mean and noise energy standard deviation obtained from the noise energy distribution, combined with the set percentile value. Sample the background noise energy of the current line and set a smoothing factor. Combine the background noise energy and the smoothing factor to obtain the calibrated noise ratio threshold. The mutation amplitude threshold is calculated as follows: the mutation amplitude threshold is obtained based on the inter-class variance function and the set self-adjustment constant. The slope threshold is calculated as follows: the slope threshold is obtained based on the standard deviation of derivative noise and a set coefficient. The jitter variance threshold is calculated as follows: the jitter variance threshold is obtained by estimating the steady-state noise variance and the jitter amplification factor under fault-free conditions. When considering changes in ambient background noise, the corrected jitter variance threshold is obtained based on the current real-time estimated background noise variance. The linearity threshold is calculated as follows: the local linearity features are divided using a Gaussian mixture model, and the midpoint of the two Gaussian centers obtained after the division is taken as the linearity threshold.

3. The method for identifying and calibrating fault traveling waves in medium-voltage distribution networks according to claim 2, characterized in that: The calibration method for the noise ratio threshold is as follows: sample the current line background noise energy, combine it with the historical statistical noise energy mean and the set percentile value, and calibrate the noise ratio threshold online. The calibration method for the mutation amplitude threshold is as follows: the mutation amplitude threshold is calculated by combining the mutation amplitude threshold with the self-adjustment constant, and the mutation amplitude threshold is calibrated online. The calibration method for the slope threshold is as follows: the set coefficient is adjusted over time, and the slope threshold is calibrated online using normalization. The calibration method for the jitter variance threshold is as follows: when the background noise changes, the jitter variance threshold is calculated by combining the ratio of the noise standard deviation when the background noise changes to the steady-state noise standard deviation, and the jitter variance threshold is calibrated online. The calibration method for the linearity threshold is as follows: the linearity threshold is calibrated online using the Gaussian mixture model.

4. The method for identifying and calibrating fault traveling waves in medium-voltage distribution networks according to claim 1, characterized in that: In S2, the implementation method of the threshold dynamic update mechanism includes: The threshold for the next time step is equal to the difference between the current time step threshold multiplied by 1 and the smoothing factor, plus the product of the actual statistical value calculated from the latest wavehead detections and the smoothing factor.

5. The method for identifying and calibrating fault traveling waves in medium-voltage distribution networks according to claim 1, characterized in that: In S3, the adaptive calibration method for different types of wavefronts is as follows: A rapid calibration method based on the differential-slope joint abrupt change criterion is adopted for clean wavefronts; A calibration method combining envelope construction and abrupt change interval contraction is used for jittery wavefronts; A calibration method combining relative rate of change and cumulative increment is used for slowly varying wavefronts; A calibration method is adopted for stepped superposition wavefronts, which involves finding step abrupt change points, clustering steps, and selecting the earliest strong step. A calibration method is adopted for the noisy wavefront, which combines continuous wavelet transform to enhance abrupt changes with the extraction of the maximum abrupt change point of scale energy.

6. A fault traveling wave identification and calibration system suitable for medium-voltage distribution networks, comprising a morphological vector construction module, a waveform category classification module, and an adaptive calibration module, characterized in that: The morphology vector construction module acquires traveling wave signals from a medium-voltage distribution network and preprocesses the traveling wave signals; it then constructs a wavefront morphology vector based on the preprocessed traveling wave signals; the wavefront morphology vector includes peak amplitude characteristics, instantaneous slope mean characteristics, jitter variance characteristics, local linearity characteristics, and noise ratio characteristics. The waveform classification module is designed with classification criteria for different wavefronts. Based on the classification criteria, the wavefront morphology vector is classified to obtain the wavefront classification result. The classification threshold in the classification criteria is updated through a dynamic threshold update mechanism. The methods for constructing the classification criteria for different types of wave heads include: Strong noise type is characterized by a noise ratio exceeding the noise ratio threshold. The clean type is characterized by peak amplitude exceeding the abrupt change amplitude threshold, instantaneous slope mean exceeding the slope threshold, and jitter variance below the jitter variance threshold. The jitter type is characterized by peak amplitude exceeding the abrupt change amplitude threshold and jitter variance not lower than the jitter variance threshold; The gradual change type is characterized by peak amplitude characteristics not exceeding the abrupt change amplitude threshold, instantaneous slope mean characteristics not exceeding the slope threshold, and local linearity characteristics not exceeding the linearity threshold. The stepped superposition type is characterized by peak amplitude features not exceeding the abrupt change amplitude threshold and local linearity features exceeding the linearity threshold; The adaptive calibration module performs wavehead calibration using the corresponding adaptive calibration method based on the wavehead division results.

7. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-5.

Images