High resistance ground fault perception method and system based on resistance multi-dimensional physical characteristics

Abstract

The application discloses a high-resistance grounding fault sensing method and system based on resistance multi-dimensional physical characteristics, and belongs to the technical field of novel power distribution system fault detection. The method fully utilizes the physical nature difference between transient normal disturbance and HIF fault mechanism of new energy, and first proposes to extract multi-dimensional characteristic quantities from the resistance waveform. The variance characteristic is used for quantifying the resistance fluctuation degree, the interval slope change frequency characteristic is used for capturing the local change mode of the resistance waveform, and the frequency domain energy distribution characteristic is used for analyzing the harmonic characteristic of the resistance waveform. A multi-criterion fusion fault sensing mechanism is constructed based on the three types of characteristics, which is used for judging whether a high-resistance grounding fault occurs. The application can effectively distinguish the real high-resistance grounding fault from normal disturbances such as load switching, capacitor switching, photovoltaic switching and the like, has stronger robustness and engineering applicability, has high fault sensing success rate, has good adaptability to noise interference and new energy access, and is obviously superior to traditional methods.

Description

Technical Field

[0001] This invention belongs to the field of power system relay protection technology, specifically relating to a method for detecting high-impedance faults (HIF) in new distribution networks, and particularly a high-impedance fault detection method and system based on multidimensional physical characteristics of resistance. Background Technology

[0002] As the "last mile" of power transmission and distribution, ensuring the safe and stable operation of the distribution network is essential. In particular, the large-scale integration of distributed renewable energy sources, such as photovoltaics and wind power, into the distribution network is driving the evolution of traditional distribution networks into new types with complex structures and flexible operating modes. This transformation has profoundly changed the fault characteristics of the distribution network, posing greater challenges to relay protection technology. Among these challenges, the accurate detection of high-impedance faults (HIFs) is particularly prominent. HIFs typically originate from distribution lines forming high-impedance grounding paths due to severe weather, mechanical forces, or insulation aging, resulting in contact with non-metallic media. The fault current is extremely weak (usually less than 50A, sometimes as low as 1A), and its amplitude highly overlaps in the frequency spectrum with the unbalanced current during normal system operation, load switching current, and harmonic currents introduced by renewable energy sources. This makes traditional staged overcurrent protection based on power frequency current amplitude thresholds ineffective. Therefore, HIFs often remain "invisible" within the power grid, increasing the risks of forest fires, electric shock injuries, and equipment insulation degradation. When a low-current grounding system needs to achieve the goal of "selecting the line and section of a permanent single-phase grounding fault and quickly isolating it nearby", new requirements are put forward for the rapid and reliable detection of HIF.

[0003] Existing methods for detecting High-Intensity Fault (HIF) can be broadly categorized into three types: frequency-domain based methods, time-domain based methods, and transient quantity-based methods. Frequency-domain based methods utilize the harmonic characteristics of HIF arcs for detection, such as algorithms based on the third harmonic, fifth harmonic, or specific frequency band energy. However, normal operations in distribution networks, such as capacitor bank switching, transformer inrush current, and power electronic loads, also generate significant harmonics, leading to high false trip rates. This is especially true after the integration of new energy sources, where harmonic interference near the inverter switching frequency becomes significant, further limiting the practicality of frequency-domain detection methods. Time-domain based methods directly analyze the shape of zero-sequence current or voltage waveforms, for example, by using mathematical morphology filters to extract distortion features or analyzing the volt-ampere characteristic curve. However, these methods rely on pre-defined fixed time windows or mathematical models, making them insufficiently adaptable to HIFs where distortion patterns dynamically change with the grounding medium. Transient quantity-based methods utilize tools such as wavelet transform and S-transform to extract high-frequency components in the initial stage of the fault. However, transient signals decay rapidly, are easily affected by noise, and their characteristic stability is difficult to guarantee.

[0004] In recent years, machine learning and deep learning technologies have been introduced into the field of HIF sensing, such as support vector machines, random forests, and convolutional neural networks. These technologies improve detection performance through data-driven approaches. However, these methods are based on probability statistics and high-dimensional parameter fitting, which ignores the physical mechanism of the fault, resulting in poor model interpretability. Furthermore, their performance is highly dependent on a large number of high-quality samples, and their generalization ability is challenged when HIF samples are scarce in actual power distribution networks.

[0005] At its root, the main bottleneck of existing HIF detection methods lies in the lack of a mechanism-level distinction between transient normal disturbances such as those from new energy sources and the essential differences between HIF faults. Most existing methods remain at the "signal level" of differentiation, that is, comparing harmonic content, energy distribution, or waveform similarity. However, the current changes caused by disturbances such as the switching of new energy sources and loads originate from the switching operation of power electronic devices, while the current distortion of HIF originates from external nonlinear physical processes such as arc breakdown and dielectric ionization. The signals generated by the two mechanisms may be similar in the time and frequency domains, but their physical nature is completely different. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing high-resistance grounding fault detection methods, namely, their failure to fully utilize the essential physical differences between transient normal disturbances such as those from new energy sources and the HIF fault mechanism. This invention provides a high-resistance grounding fault sensing method and system based on the multidimensional physical characteristics of resistance. The technical solution of this invention starts with the resistance waveform, which reflects the essence of the fault, and extracts three types of physically meaningful features: variance, the number of interval slope changes, and frequency domain energy distribution. A multi-criteria fusion fault sensing mechanism is constructed to effectively distinguish between real faults and normal disturbances, achieving reliable detection of high-resistance grounding faults.

[0007] To achieve the above technical objectives, the present invention adopts the following technical solution:

[0008] A high-resistance grounding fault detection method based on multidimensional physical characteristics of resistance includes:

[0009] Step 1, Resistance Waveform Data Acquisition: Acquire / obtain the fault resistance waveform sequence R of the distribution network. F ;

[0010] Step 2, Multidimensional physical feature extraction of resistance waveform: Using a preset window length, extract the fault resistance waveform sequence R... F The data is divided into continuous segments, and then multi-dimensional physical features are extracted for each segment.

[0011] The multidimensional physical features are: variance feature Var(m), and the number of times the slope changes in the interval C. t (m), spectral flatness Fla(m), where m is the marker of the data segment;

[0012] Step 3, Multi-criteria fusion fault perception: Utilize the multi-dimensional physical features to perceive whether a high-resistance grounding fault has occurred.

[0013] This invention proposes for the first time to comprehensively track and judge resistance waveforms from three specific dimensions (wave intensity, local variation mode, and harmonic energy distribution) to achieve reliable HIF sensing. The combined application of these three specific dimensions represents a significant improvement in the multi-dimensional, mechanistic-level feature application for HIF detection.

[0014] Furthermore, step 3 involves first constructing a fixed fault criterion based on multi-dimensional physical characteristics, and then sensing whether a high-resistance grounding fault action has occurred based on the fixed fault criterion.

[0015] The fixed fault criterion is:

[0016]

[0017] In the formula, V is the variance threshold, and C is the threshold for the number of times the slope changes in the interval.

[0018] Wherein, if the variance characteristic Var(m) of the data segment is C, the number of times the slope changes in the interval is C. t (m), the spectral flatness Fla(m) both satisfy the above inequality, and the corresponding data segment is the fault characteristic data segment; there are multiple consecutive fault characteristic data segments, and a high-impedance grounding fault action signal is issued.

[0019] It should be understood that the above criteria and threshold values ​​are empirical values ​​adopted based on the unique physical mechanism of HIF and through certain reasonable deductions. The specific reasoning basis is explained as follows:

[0020] Regarding the variance characteristic Var(m): The variance threshold V is set with the goal of achieving a significant difference between the variance of the resistance fluctuation magnitude of a high-resistance ground fault (HIF) and that under normal operating conditions. Due to the intermittent extinguishing of the arc, the resistance fluctuation of an HIF can reach tens to hundreds of kΩ, while the resistance fluctuation of most disturbances (except for severe load switching) is less than 1 kΩ². Therefore, in some implementations, setting the variance threshold V to 1 kΩ² is sufficient to identify most HIFs. In other implementations, adjusting the value according to accuracy or fluctuating around 1 kΩ² is also feasible.

[0021] Regarding the direction and number of changes in the interval slope C t (m): The threshold C for the number of slope changes in the interval is based on a physical threshold, not a statistical threshold, of the HIF resistor waveform shape. Normal disturbances exhibit limited monotonicity changes within a quarter-cycle; however, the unique "M"-shaped waveform of HIF inevitably leads to multiple slope sign changes within half a cycle. Theoretical analysis shows that as long as a significant "zero-rest-reignition" process exists, C... tIt must be greater than 2; therefore, in some implementations, C t Setting the threshold to 2 already provides strong applicability. In other feasible embodiments, adjustments can be made adaptively based on actual application needs and environments.

[0022] Regarding the spectral flatness Fla(m): This criterion is set based on the physical difference between the broadband characteristics of HIF and the narrow-band characteristics of switching disturbances. The HIF arc reignition / extinguishing process excites broadband harmonics, with energy evenly distributed in low and high frequencies, so the Fla of each sub-band waveform is definitely not 0; while normal switching disturbances usually have a switching frequency at the moment of switching, and their energy distribution is concentrated on the switching frequency, causing the Fla of the waveform in that sub-band to be 0. Therefore, setting the min[Fla] threshold ≠ 0 can ensure that the influence of energy-concentrated disturbances, such as those mainly caused by new energy switching, on HIF sensing is excluded.

[0023] Furthermore, in step 2, the preset window length is half the period length of the resistance waveform, and the mathematical model of the variance feature Var(m) is:

[0024]

[0025] In the formula, Var(m) is the variance of the m-th segment of the resistance waveform data, and Rm is the variance of the resistance waveform data. F (j) represents the resistance waveform data of the j-th segment, R FA (m) represents the average value of N resistance waveform data, including the m-th resistance waveform data, where N is a positive integer.

[0026] Furthermore, in step 2, the preset window length is half the period length of the resistance waveform, and the number of times the slope changes in the interval is C. t The mathematical model for (m) is as follows:

[0027] For each sampling point m in the m-th segment of the resistance waveform data s With sampling point m s Define the length as L=N at the center. t The calculation window for / P is defined, where P is a preset value. Linear regression of the curve within this window is performed using the least squares method, and the regression slope is defined as the value at the sampling point m. s The interval slope Slo(m) s The calculation formula is:

[0028]

[0029] In the formula, R F (j s ) represents the j-th fault resistance in this interval. s Each sampling point signal;

[0030] Then count the number of times the slope changes direction within half a period to obtain C. t (m).

[0031] Furthermore, in step 2, the preset window length is half the period length of the resistance waveform, and the extraction process of the spectral flatness Fla(m) is as follows:

[0032] The Pisarenko harmonic decomposition method was used to perform spectral analysis on the m-th segment of the resistance waveform. The obtained spectrum was divided into k sub-bands, and the spectral flatness Fla(m) of each sub-band was calculated. s ), and thus the spectral flatness Fla(m) is obtained:

[0033]

[0034] In the formula, R F (f ms ) is R F (f m At frequency f ms The calculated energy amplitude is given by s = 1, 2, ..., k; R F (f m Let f be the frequency domain waveform of the fault resistor in the m-th segment. m f represents the frequency range covered by the entire spectrum of the m-th signal segment. ms Let be the center frequency of the s-th sub-band after the spectrum of the m-th signal is divided into k sub-bands.

[0035] Furthermore, in step 3, the detection of whether a high-resistance grounding fault has occurred using the multidimensional physical characteristics is based on the variance feature Var(m) and the number of times the slope changes in the interval, C. t (m), the spectrum flatness Fla(m) corresponds to the high resistance grounding fault characteristics and sets the fault criterion or introduces deep learning or machine learning to construct a multi-dimensional physical feature matrix as the model input of deep learning or machine learning;

[0036] The variance characteristic Var(m) corresponds to the high-resistance grounding fault characteristics as follows: the resistance waveform of a high-resistance grounding fault exhibits high variance.

[0037] Number of times the slope changes in the interval C t The characteristics of the high-resistance grounding fault corresponding to (m) are as follows: Within half a cycle, the interval slope curve of the high-resistance grounding fault under zero rest duration has multiple extreme points, and the number of times the interval slope changes is C. t (m) more than 2 times;

[0038] The high-resistance grounding fault characteristics corresponding to the spectral flatness Fla(m) are: the values ​​of all elements in the spectral flatness Fla(m) under the high-resistance grounding fault are close in magnitude and not equal to 0.

[0039] Furthermore, step 1 also includes performing the following: processing the fault resistor waveform sequence R. F Preprocessing is performed, namely, using a low-pass filter to remove high-frequency noise and using mathematical morphology filtering to eliminate short-term irregular distortion.

[0040] Furthermore, the method also includes an anti-interference performance verification step: by building a 10kV multi-control power grid simulation model in PSCAD, setting different fault locations, arc resistance, noise interference and new energy power output conditions, the verification method has a 100% success rate in fault perception under 550 sets of simulation data, which is obviously better than the traditional method.

[0041] Furthermore, the present invention also provides a system based on the above method, comprising:

[0042] The resistance waveform data acquisition module is used to acquire / obtain the fault resistance waveform sequence R of the distribution network. F ;

[0043] A multi-dimensional physical feature extraction module for resistance waveforms is used to extract the fault resistance waveform sequence R within a preset window length. F The data is divided into continuous segments, and then multi-dimensional physical features are extracted for each segment.

[0044] The multidimensional physical features are: variance feature Var(m), and the number of times the slope changes in the interval C. t (m), spectral flatness Fla(m), where m is the marker of the data segment;

[0045] The fault detection module is used to detect whether a high-resistance grounding fault has occurred by utilizing the multi-dimensional physical characteristics.

[0046] Furthermore, the present invention also provides a computer terminal, comprising:

[0047] One or more processors;

[0048] A memory that stores one or more computer programs;

[0049] The processor invokes a computer program to implement the steps of a high-resistance grounding fault detection method based on multidimensional physical characteristics of resistance.

[0050] Finally, the present invention also provides a computer-readable storage medium storing a computer program that is invoked by a processor to implement the steps of a high-resistance grounding fault detection method based on multidimensional physical characteristics of resistance.

[0051] Compared with the prior art, the present invention achieves the following progress and effects:

[0052] This invention, starting from the physical mechanism level, effectively distinguishes HIF from various normal disturbances through the coordinated judgment of three types of features: variance, slope change, and frequency domain energy, achieving a 100% fault detection success rate. The three types of features target different physical characteristics of HIF, resulting in a highly interpretable model that overcomes the shortcomings of black-box models. The multi-criteria fusion mechanism requires all three conditions to be met simultaneously for a diagnosis of HIF; noise and disturbances cannot simultaneously mimic all features, thus exhibiting high robustness. It has low computational complexity, requires no large number of samples for training, and can be embedded in existing protection devices. Attached Figure Description

[0053] Figure 1 This is a flowchart of the method described in the embodiments of this application.

[0054] Figure 2 This is the waveform of the fault resistor.

[0055] Figure 3 It shows the fault resistance waveform and its variance curve.

[0056] Figure 4 shows the slope variation characteristics of the resistance range under non-fault and fault conditions. Figure 4a shows the range slope curve of the HIF resistor, and Figure 4b shows the range slope curve of the CS resistor.

[0057] Figure 5 shows the resistance energy distribution characteristics under non-fault and fault conditions. Figure 5a shows the spectrum and flatness of the fault resistance; Figure 5b shows the spectrum and flatness of the photovoltaic switching PS resistance.

[0058] Figure 6 It is a simulation model of a 10kV multi-control power grid.

[0059] Figure 7 These are the calculation results of three fault characteristic quantities from 550 sets of simulation data segments.

[0060] Figure 8 shows the resistance waveform and characteristic curves of the LS operating condition during the start-up of a large motor. In Figure 8a, the LS switching resistance is shown, 8b is the variance, 8c is the interval slope, and 8d is the spectral flatness. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. The technical features involved in the various embodiments of the invention described below can be combined with each other as long as they do not conflict with each other.

[0062] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0064] This invention provides a high-resistance grounding fault detection method based on multi-dimensional physical characteristics of resistance. Starting from the physical mechanism level, it effectively distinguishes HIFs from various normal disturbances through the coordinated judgment of three types of features: variance, slope change, and frequency domain energy. The following detailed description of embodiments of this invention is based on the technical solution of this invention, providing detailed implementation methods and specific operating procedures to further explain the technical solution of this invention.

[0065] This embodiment provides a high-resistance grounding fault detection method based on multi-dimensional physical characteristics of resistance, taking a 10kV new distribution network high-resistance grounding fault detection as an example, referring to... Figure 1 As shown, the main steps include:

[0066] Step 1, Resistance waveform data acquisition and preprocessing. This includes acquiring the fault resistor waveform sequence R. F The fault resistor waveform is obtained by preprocessing using a low-pass filter and mathematical morphology filtering methods, as shown below. Figure 3 As shown, the fault resistor (HIF resistor) is the total fault resistor.

[0067] The "zero-sequence" characteristic of the current is a hallmark physical phenomenon of high-resistivity ground faults (HIFs). Although the amplitude and waveform distortion of the fault zero-sequence current differ under different grounding media, there is always a nonlinear distortion near the current zero-crossing point originating from the "zero-sequence" characteristic. This type of nonlinear distortion has a clear physical meaning, causing the HIF resistance to exhibit a special "M"-shaped trend within one cycle, such as... Figure 2As shown, the resistance waveform is significantly different from that under transient normal disturbance conditions such as those from new energy sources. Therefore, if the resistance sequence is dynamically tracked from multiple perspectives, that is, if feature quantities that can directly reflect the essential process of the fault are constructed from the physical mechanism level, sensitive detection of high-resistance ground faults (HIFs) can be achieved. It is worth mentioning that due to the continuous arcing of high-resistance ground faults (HIFs), the fault signal will exhibit irregular distortion, and the random occurrence of actual noise in the distribution network is not conducive to the multi-dimensional feature extraction of the resistance waveform. To address this, a low-pass filter (e.g., setting a cutoff frequency of 1kHz) is used to filter out high-frequency noise components; and mathematical morphology filtering is employed to eliminate outliers caused by short-term irregular distortions, thereby reducing the impact of interference information on fault resistance feature extraction to a certain extent.

[0068] Step 2, Analysis of Resistor Dynamic Characteristics:

[0069] (1) Variance

[0070] To quantify the fluctuation of the resistance waveform, we first consider the variance characteristic Var. Variance is an indicator that describes the stability of a data sequence. Due to the special "M"-shaped fluctuation of the resistance waveform when a high-resistance ground fault (HIF) occurs (i.e., it fluctuates every half cycle), the fault resistance exhibits high variance characteristics during its change.

[0071] Therefore, the data in half a cycle of the resistance waveform is considered as a data segment, and the interval length of this data segment is N. t Next, calculate the variance of the N data segments as the variance feature Var. The mathematical model is as follows:

[0072] Taking the m-th data segment as an example, the corresponding variance feature Var(m) is the variance of the adjacent N segments of resistance waveform data containing the m-th data segment, and the expression is shown in Equation (1).

[0073] (1)

[0074] In the formula, Var(m) is the variance of the m-th segment of the resistance waveform data, and Rm is the variance of the resistance waveform data. F (j) represents the j-th segment of the resistance waveform, which is of length N. t The sequence; R FA (m) represents the average value of N resistance waveform data segments including the m-th segment. This invention is not limited to the selection rules for the N resistance waveform data segments including the m-th segment; for example, the N-1 adjacent data segments preceding the m-th segment can be selected for calculation, i.e., the subscripts of the summation symbol in Formula 1. For example, if the m-th segment is an intermediate data segment, the summation subscript j = m − k, the superscript j = m + l, and the constraint k + l + 1 = N, where k and l represent the number of data segments before and after the current segment, respectively.

[0075] Figure 3 The figure shows the variance curve of the HIF resistance waveform. Due to the intermittent arc extinction, the fault resistance exhibits a high variance (>1kΩ²), which significantly amplifies the fluctuation of resistance under fault conditions. Obviously, the variance under non-fault conditions is unlikely to reach the kΩ² level, thus failing to distinguish whether a high-resistance ground fault (HIF) has occurred.

[0076] However, relying solely on variance has certain limitations: First, some drastic load fluctuations or specific switching operations may momentarily lead to high variance values, which can easily be confused with true high-resistance ground faults (HIFs) and increase the probability of misjudgment; Second, and more importantly, variance only characterizes the overall amplitude of HIF resistance fluctuations and fails to reveal the differences in its inherent change patterns. Typical HIF resistance fluctuations are not only large in amplitude, but their core characteristic lies in the arc extinction / reignition mechanism, especially during the "zero-down" period. Due to thermal equilibrium, they exhibit a unique periodic, slowly rising and falling "M"-shaped pattern (the combustion and extinction of the arc are constrained by the balance of local energy accumulation and dissipation at the fault point. When the current increases, heat accumulation leads to gas ionization, resistance decreases, and the arc reignites; when the current decreases to zero, heat dissipation dominates, the dielectric deionizes, resistance increases sharply, and the arc extinguishes. This physical mechanism, driven by the periodic change of the power supply voltage and closely coupled with the thermal process, makes the "increase-decrease" change of resistance synchronous with the power frequency, and the rate of change is relatively slow due to thermal inertia, thus forming a periodic, slowly rising and falling "M"-shaped pattern). This pattern is its essential characteristic that distinguishes it from transient disturbances (such as switching operations). That is, the fluctuation of HIF is the result of the continuous and periodic evolution of the physical state (degree of ionization, temperature) of the fault point driven by the system voltage cycle; while normal disturbances are the process of rapidly entering a new steady state after a sudden change in circuit topology or parameters, and do not have periodic repeatability. Therefore, to more accurately capture and quantify the specific morphological characteristics dominated by the thermal balance principle, especially the significant increase and variation of resistance during the "zero rest" period, and to effectively distinguish between capacitor switching (CS) / load switching (LS) disturbances that only cause short-term, minor changes, the interval slope is introduced as a key characteristic quantity. The number of interval slope changes is an effective tool for quantifying the unique local variation pattern of the HIF resistance waveform (i.e., the specific manifestation of the periodic, slow rise and fall pattern). Its effectiveness has been demonstrated through simulation comparisons in the document. Figure 3 The results (and analysis) were verified.

[0077] (2) Interval slope

[0078] The slope of the interval directly reflects the rate and direction of change in the resistance waveform, thus highlighting the essential difference between high-resistance grounding fault (HIF) and normal disturbance conditions. To eliminate the influence of outliers on the waveform slope, further analysis is performed on each sampling point m in the m-th segment of the resistance waveform data. sCentered on this point, define a boundary with length L=N. t The calculation window for / P is set, where P is a preset value, and in this embodiment, P is set to 4. Then, according to equation (2), the least squares method is used to perform linear regression on the curve within the window, and the slope after regression is defined as the interval slope Slo(m) of the central sampling point. s By sliding this window, you can obtain the slope curve of the interval that matches the sampling frequency of the resistor waveform.

[0079] (2)

[0080] In the formula, R F (j s ) is the j-th resistance of the HIF resistor within this window interval. s Each sampling point signal.

[0081] The interval slope characteristic can effectively identify transient disturbances such as load switching (LS) and capacitor switching (CS). Taking CS as an example, Figure 4 shows the interval slope curves of the resistance waveforms for HIF and CS conditions, respectively. Figure 4a shows the interval slope curve of the HIF resistance, and Figure 4b shows the interval slope curve of the CS resistance. It can be seen that within half a cycle, the interval slope curve of HIF under zero rest time has multiple extreme points due to high resistance grounding. The interval slope changes direction many times within this time window, and this number is represented by C. t (m) represents the phenomenon. The physical essence of this phenomenon still lies in the arc extinction and reignition characteristics. Specifically, even after the current crosses zero and the arc temporarily extinguishes, the residual plasma in the arc gap has not completely dissipated. Through a very small amplitude current, a convex-concave resistance waveform is formed within a short period after the current crosses zero, thus causing multiple interval slope minimum points. In contrast, during CS, the slope only changes direction twice—falling and then rising—at the switching moment, and its C... t The value stabilizes at 2. Therefore, in this embodiment, the threshold C for the number of times the slope changes in the interval is 2. t When (m) is greater than 2, a fault can be determined to have occurred.

[0082] (3) Energy distribution

[0083] The aforementioned resistance-related characteristics all determine whether a high-resistance ground fault (HIF) has occurred from the perspective of time-domain waveform. Considering that new energy sources connecting to the distribution network generate a large number of harmonics, exhibiting a concentration of specific harmonic energy, and that the arc reignition / extinguishing process of a HIF will excite broadband harmonics with energy evenly distributed in low and high frequencies, a spectral analysis is performed on the m-th segment of the resistance waveform to extract the energy proportion in different frequency bands. This embodiment preferentially uses the Pisarenko harmonic decomposition method for spectral analysis, effectively avoiding the spectral distortion problem caused by short-time data windows in new energy scenarios due to its high-resolution characteristics. The specific steps are as follows:

[0084] 1) Sample the m-th segment of the time-domain resistance waveform to obtain the discrete signal R. F (p), where p is the number of sampling points for the discrete signal, L is the total number of sampling points, and n is the number of sampling points for the autocorrelation function. Then, the discrete signal R is calculated. F The autocorrelation function R of (p) F [q]:

[0085] (3)

[0086] 2) Construct the autocorrelation matrix R F :

[0087] (4)

[0088] 3) For the autocorrelation matrix R F Perform eigenvalue decomposition (which is the process of finding the eigenvalues ​​of a matrix in linear algebra):

[0089] (5)

[0090] In the formula, V F It is the eigenvector matrix, Λ F H is the eigenvalue matrix, and H represents the conjugate transpose of the matrix.

[0091] 4) Select the n largest eigenvalues ​​λ1, λ2, ..., λn from the eigenvalue matrix. n and their corresponding eigenvectors v1, v2, ..., v n The m-th segment of the resistance frequency domain waveform R is constructed using the selected eigenvalues ​​and eigenvectors. F (f m ), can be represented as:

[0092] (6)

[0093] In the formula, R F (f ms ) is R F (f m At frequency fms The calculated energy amplitude is given by s = 1, 2, ..., n; R F (f m () represents the frequency domain waveform of the m-th resistor segment, where f is the frequency, generally referring to the frequency; f m f represents the frequency range covered by the entire spectrum of the m-th signal segment. ms Let be the center frequency of the s-th sub-band after the spectrum of the m-th signal is divided into k sub-bands.

[0094] Equation (6) optimizes the signal energy ratio within a short-time data window through a weighting mechanism based on the reciprocals of eigenvalues, and enhances the signal's anti-interference capability by utilizing the orthogonal projection characteristics of eigenvectors. Based on the above analysis, spectral flatness is introduced to measure the uniformity of the resistor's frequency domain waveform energy. Similarly, R... F (f m If the frequency band is divided into k sub-bands, then the spectral flatness Fla(m) of the s-th sub-band is... s ) and R F (f m The spectral flatness Fla(m) of the signal is shown in Equation (7). It is worth noting that spectral flatness only requires calculation of the average value of the data, making the calculation simple and intuitive, and the result is independent of the overall signal amplitude (only focusing on the relative distribution), thus exhibiting high robustness. When the magnitude difference between all element values ​​in Fla(m) is small, it indicates that R... F (f m The energy is evenly distributed across the entire analyzed frequency range, closely resembling the characteristics of a high-impedance ground fault (HIF); when an element in Fla(m) has a value close to 0, it indicates that R... F (f m The energy is highly concentrated on a few single frequencies, approaching the normal disturbance characteristics of a switch.

[0095] (7)

[0096] Figure 5 shows the energy distribution characteristics of the resistance waveforms under high-resistance ground fault (HIF) and photovoltaic (PS) switching at an inverter switching frequency of 2.5kHz, respectively. Figure 5a shows the fault resistance spectrum and its flatness; Figure 5b shows the photovoltaic-switched PS resistance spectrum and its flatness. It can be seen that within a certain frequency range, the energy of HIF is distributed across all frequency bands, with a relatively large overall spectral flatness; while the energy of the photovoltaic-switched PS is mainly concentrated at the switching frequency, with spectral flatness near that frequency being zero. Therefore, using 10 sub-bands as a window period, if at least one of the 10 sub-bands has a calculated Fla(m) value... s If ) is 0, that is, min[Fla(m)]=0, then it is determined that no high-resistance ground fault (HIF) has occurred in the distribution network.

[0097] Step 3: At this point, multidimensional features of the fault resistance have been extracted from the fault mechanism level (covering multiple physical processes including dynamic changes in grounding resistance and energy changes). Judgment conditions have been set for the extracted features as fault criteria, and a fixed fault criterion has been proposed: Based on the preceding analysis, when a high-resistance ground fault (HIF) occurs, the difference between the variance Var of the fault resistance and that of the resistance without a fault can reach the order of kΩ². This is further supported by the number of times the slope changes in the interval, C. t The spectral flatness Fla can eliminate the influence of normal disturbances on fault identification, thus achieving accurate HIF sensing. In this embodiment, the preferred fault criterion is shown in Equation (8).

[0098] (8)

[0099] In the formula, Var is in kΩ². If the m-th data segment satisfies formula (8) simultaneously, then the data segment is judged as a "fault characteristic data segment". To avoid misjudgment, considering the persistence of fault characteristics, when multiple "fault characteristic data segments (FCDS)" appear consecutively (such as 4 in this embodiment, and can be adaptively adjusted in other feasible embodiments), it is considered that a high-resistance grounding fault (HIF) has occurred.

[0100] It should be understood that the core innovation of this invention is not to pursue absolute accuracy of a single threshold, but rather a fault-tolerant mechanism based on multi-criteria fusion. This design makes the method relatively lenient in its requirements for the accuracy of a single threshold, because the system reliability is built on the synergistic effect of multi-dimensional features. Compared with traditional single-criteria methods, multi-criteria fusion has significant advantages: ① Strong anti-interference ability: transient disturbances are difficult to simultaneously mimic HIF features in all dimensions; ② Strong adaptability: insensitive to changes in a single threshold, resulting in high system stability; ③ High reliability: misjudgment requires the simultaneous fulfillment of three conditions, with an extremely low probability.

[0101] Furthermore, without considering the "best results," using the three-dimensional features proposed in this invention as input to a deep learning or machine learning model to achieve ground fault detection is also technically feasible and represents another application of the feature set proposed in this invention. The selected deep learning network and machine learning model can refer to existing technologies or be further optimized based on existing technologies; this invention does not impose specific limitations on them, only limiting the model input to: a feature vector / matrix constructed based on the above three-dimensional features; and the model output to: the high-resistance ground fault detection result. The above embodiments, based on a fusion method with fixed criteria, have advantages in reliability, interpretability, and engineering simplicity.

[0102] To further evaluate the performance of this invention, it is necessary to conduct tests and verifications under various conditions, including different power distribution network structures, fault locations, arc resistance, noise interference, and renewable energy output. A system was built in PSCAD as follows: Figure 6 The simulation model of a 10kV multi-branch power grid shown in the figure is illustrated. E1 is the waveform measurement point with a sampling frequency of 10kHz. E2, E3, E5, E7, and E8 are branch nodes, and E4, E6, E9, and E... 10 The faults are set at the endpoints of the distribution lines. Fault F1 is set 1km from the E1 endpoint of line E1E2, fault F2 is set 1km from the E3 endpoint of line E2E3, fault F3 is set 2km from the E5 endpoint of line E5E6, and faults F4 and F5 are set at the E8 and E9 endpoints respectively. Photovoltaic power supplies are connected to the E2, E3, and E8 endpoints respectively. Data on four types of HIF (phase A grounding) and three types of normal disturbances are acquired. The simulation data sets are expanded by changing the location of the fault or disturbance, the magnitude of the arc resistance, and the degree of noise interference, as shown in Table 1. The calculation results of the three fault characteristic quantities of the 550 sets of simulation data segments are as follows. Figure 7 As shown.

[0103] Table 1 Simulation data parameter settings

[0104] Depend on Figure 7 It can be seen that all three fault characteristic quantities of the 400 sets of HIFs satisfy equation (8), while none of the 150 sets of disturbance data simultaneously satisfy equation (8), resulting in a 100% fault detection success rate. The reason is:

[0105] (1) Regarding the changes in fault location and arc resistance, in extreme cases (faults at the end of the line and extremely high arc resistance), the amplitude of the resistance waveform will be higher and the fluctuations will be more severe, resulting in a variance characteristic Var greater than 1kΩ. 2 The number of times C is related to the direction of change of the slope of the interval t Conditions greater than 2 are more easily satisfied. Furthermore, regardless of the fault location or resistance value, the fluctuations in the resistance waveform originate from the nonlinear, random physical process of the arc, and its frequency domain characteristics are fundamentally different from the periodic oscillations or harmonics generated by normal disturbances. The spectral flatness Fla can still effectively distinguish real fault fluctuations from other disturbances. Therefore, this method has inherent adaptability to changes in fault location and arc resistance, and its effectiveness remains unaffected.

[0106] (2) For noise interference of varying degrees, since the present invention uses low-pass filters and mathematical morphology to preprocess the data in the early stage, most noise interference can be avoided to accurately extract the simulation data. However, under strong noise interference of 20dB, the variance characteristics Var and the number of interval slope changes C of the resistance waveform are significantly affected. tThe noise level will increase, and the spectral flatness Fla of all sub-bands may approach 1. This can easily cause confusion between HIF and normal disturbances. However, subsequent spectral analysis using the Pisarenko harmonic decomposition method can effectively suppress the influence of noise on the calculation of spectral flatness Fla. Therefore, noise interference cannot simultaneously satisfy the three conditions of equation (8), and the system will not malfunction.

[0107] (3) For normal disturbance switching positions and capacity changes, only the amplitude and frequency of the disturbance will be affected. The essence of various switching operations is instantaneous disturbance, not a continuous state, making it difficult to satisfy C. t The condition is greater than 2. Therefore, it does not affect the judgment result of this method.

[0108] Next, consider the extreme case. Figure 6 Position E4 in the simulation represents a typical high-variance, severe disturbance condition (LS) during the startup of a large motor. The load capacity is set to 1MW, and the switching process lasts 6ms. The resistance waveform within 10ms during this period is as follows: Figure 8a As shown, its variance, slope, and spectral flatness curves representing the energy distribution are respectively as follows: Figure 8b , 8c As shown in Figure 8d, the disturbance waveform lacks the regular "M"-shaped structure of HIF, which is strictly synchronized with the zero-crossing point of the supply voltage and repeats repeatedly. Even though the oscillation causes Var to rise, and C... t The value fluctuates drastically by more than 2 during this period, but its Fla also exhibits energy concentration due to the fixed oscillation frequency. Furthermore, from a mechanistic perspective, C... t It is difficult to consistently exceed 2 within each half-cycle of analysis. The proposed algorithm requires four consecutive "Fault Feature Data Segments (FCDS)" to be considered a fault. This design is specifically designed to handle such transient disturbances. A continuous switching process may generate 1-2 suspicious data segments, but it is difficult to continuously generate data segments that simultaneously satisfy all three conditions and have a regular shape for four power frequency cycles. The physical essence of HIF is a continuous nonlinear physical process. Its "M"-shaped waveform originates from the repetitive arc reignition and extinction inevitably caused by the zero crossing of the power supply voltage and the nonlinear characteristics of the arc-grounding medium within each power frequency cycle. This fluctuation is periodically forced; as long as a fault exists, the waveform will stably present an "M" shape in each cycle. The core of this invention is to capture this essential difference through multi-dimensional features. Therefore, the three characteristic quantities of HIF will simultaneously satisfy the criteria in each cycle during the fault duration. For disturbances, even if the transient process causes one or two characteristic quantities to instantaneously satisfy the conditions, the probability of all three characteristic quantities being simultaneously and continuously satisfied is extremely low.

[0109] The above analysis shows that HIF is the only physical phenomenon that can simultaneously induce significant amplitude fluctuations, continuous random changes, and non-flat spectrum characteristics across the entire frequency band. Various disturbances during normal system operation can at most mimic one or two of these characteristics, but cannot simultaneously satisfy all conditions. The synergistic effect of multiple criteria in Equation (8) ensures the method's strong robustness to the aforementioned influencing factors at the mechanistic level, effectively avoiding misjudgments and improving the reliability of HIF sensing.

[0110] To further highlight the advantages of this invention, three HIF sensing methods were selected for comparison: one based on fault voltage variation (Method 1), one based on fault current time-frequency characteristics (Method 2), and one based on common fault current low-order harmonics (Method 3). These three methods utilize time-frequency domain, time-domain, and frequency-domain analysis techniques, respectively, covering representative high-frequency and low-frequency components of the signal. In Method 1, since mathematical morphology methods can detect minute changes in waveforms, HIF sensing can be performed based on the degree of distortion of the fault voltage in the time domain. Method 2 uses wavelet transform to extract the time-frequency characteristics of the fault current, and then uses the calculated standard deviation features to construct a fault detection criterion (FDC). Method 3 proposes a HIF sensing method by extracting the energy amplitudes (H2, H3, and H5) of the second, third, and fifth harmonics in the frequency domain.

[0111] For ease of comparison, Figure 6 The E2 position was set with HIF, LS, CS, and PS respectively, and the comparison results are shown in Table 2. This invention can accurately distinguish between HIF, LS, CS, and PS. In contrast, while Method 1 can detect HIF using Closing-Opening Difference Operation (CODO), it cannot distinguish between HIF and LS. Although Method 2 can successfully distinguish between HIF and LS, it still faces challenges in accurately detecting HIF. Method 3 struggles to accurately identify normal system operating states (such as PS and CS). Furthermore, the difficulty in setting appropriate thresholds limits the application scope of existing methods. The multi-criteria fusion mechanism of this invention does not require absolute accuracy of a single threshold. Compared with traditional single-criteria methods, multi-criteria fusion has significant advantages: ① Strong anti-interference capability: transient disturbances are difficult to simultaneously mimic HIF characteristics across all dimensions; ② Strong adaptability: insensitive to changes in a single threshold, resulting in high system stability; ③ High reliability: misjudgment requires simultaneous fulfillment of three conditions, with an extremely low probability. Of course, a threshold fine-tuning interface can be reserved in future hardware implementations, allowing for minor optimization based on actual operating data. In summary, the comparative results confirm the effectiveness and advantages of the present invention.

[0112] Table 2 Fault detection results of different methods

[0113] In other embodiments, the present invention also provides a system based on the above method, including a resistance waveform data acquisition module, a multi-dimensional physical feature extraction module for resistance waveforms, and a fault perception module connected sequentially or interconnected. The resistance waveform data acquisition module is used to acquire / obtain the fault resistance waveform sequence R of the distribution network. F The multi-dimensional physical feature extraction module for the resistance waveform is used to extract the fault resistance waveform sequence R within a preset window length. F The data is divided into continuous segments, and multi-dimensional physical features are extracted for each segment. The fault perception module is used to detect whether a high-resistance grounding fault has occurred using the multi-dimensional physical features.

[0114] It should also be understood that the specific implementation process of each module is described in the above method. This invention will not repeat it here. The above division of functional modules is only for illustrative purposes. In some embodiments, some functional modules can be combined and some functional modules can be separated. Each functional module can be implemented in software, hardware, or a combination of software and hardware. The software and hardware devices include, but are not limited to, general-purpose computer equipment, programmable gate arrays, digital signal processors, microprocessors and their corresponding programming or burning software.

[0115] In other embodiments, the present invention also provides a computer terminal, including: one or more processors and a memory storing one or more computer programs;

[0116] The processor invokes a computer program to implement the steps of a high-resistance grounding fault detection method based on multidimensional physical characteristics of resistance.

[0117] Please refer to the explanation of the method above for the specific implementation process of each step.

[0118] The processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention.

[0119] The memory can be implemented in the form of read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory, and the processor calls the algorithm program of the methods described above in the embodiments of this invention.

[0120] In other embodiments, the present invention also provides a computer-readable storage medium storing a computer program that is invoked by a processor to implement the steps of a high-resistance grounding fault detection method based on multidimensional physical characteristics of resistance.

[0121] Please refer to the explanation of the method above for the specific implementation process of each step.

[0122] The aforementioned computer-readable storage medium can be an internal storage unit of the hardware and software device described in any of the foregoing embodiments, such as the hard drive or memory of the controller. The readable storage medium can also be an external storage device of the controller, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the controller. Further, the readable storage medium can include both internal storage units and external storage devices of the controller. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0123] Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application refers to flowchart illustrations and / or instructions executed by a processor of a method, apparatus (system), and computer program product according to embodiments of this application to create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams. These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowchart illustrations and / or one or more block diagrams. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more blocks of a block diagram.

[0125] The above embodiments are preferred embodiments of this application. Those skilled in the art can make various changes or improvements based on them. Without departing from the overall concept of this application, these changes or improvements should fall within the scope of protection claimed in this application.

Claims

1. A high-resistance grounding fault detection method based on multi-dimensional physical characteristics of resistance, characterized in that, Includes the following steps: Step 1, Resistance Waveform Data Acquisition: Acquire / obtain the fault resistance waveform sequence R of the distribution network. F ; Step 2, Multidimensional physical feature extraction of resistance waveform: Using a preset window length, extract the fault resistance waveform sequence R... F The data is divided into continuous segments, and then multi-dimensional physical features are extracted for each segment. The multidimensional physical features are: variance feature Var(m), number of times the slope changes in the interval C. t (m), spectral flatness Fla(m), where m is the marker of the data segment; Step 3, Multi-criteria fusion fault perception: Utilize the multi-dimensional physical features to perceive whether a high-resistance grounding fault has occurred.

2. The method according to claim 1, characterized in that: Step 3 is to first construct a fixed fault criterion based on multi-dimensional physical characteristics, and then detect whether a high-resistance grounding fault action occurs based on the fixed fault criterion. The fixed fault criterion is: In the formula, V is the variance threshold, and C is the threshold for the number of times the slope changes in the interval. Wherein, if the variance characteristic Var(m) of the data segment is C, the number of times the slope changes in the interval is C. t (m), the spectral flatness Fla(m) both satisfy the above inequality, and the corresponding data segment is the fault characteristic data segment; there are multiple consecutive fault characteristic data segments, and a high-impedance grounding fault action signal is issued.

3. The method according to claim 1, characterized in that: In step 2, the preset window length is half the period length of the resistance waveform, and the mathematical model of the variance feature Var(m) is: In the formula, Var(m) is the variance of the m-th segment of the resistance waveform data, and Rm is the variance of the resistance waveform data. F (j) represents the resistance waveform data of the j-th segment, R FA (m) represents the average value of N resistance waveform data, including the m-th resistance waveform data, where N is a positive integer.

4. The method according to claim 1, characterized in that: In step 2, the preset window length is half the cycle length of the resistance waveform, and the number of times the slope changes in the interval is C. t The mathematical model for (m) is as follows: For each sampling point m in the m-th segment of the resistance waveform data s , with sampling point m s Define the length as L=N at the center. t The calculation window for / P is defined, where P is a preset value. Linear regression of the curve within this window is performed using the least squares method, and the regression slope is defined as the value at the sampling point m. s The interval slope Slo(m) s The calculation formula is: In the formula, R F (j s ) represents the j-th fault resistance in this interval. s Each sampling point signal; Then count the number of times the slope changes direction within half a period to obtain C. t (m).

5. The method according to claim 1, characterized in that: In step 2, the preset window length is half the period length of the resistance waveform. The extraction process of the spectral flatness Fla(m) is as follows: The Pisarenko harmonic decomposition method was used to perform spectral analysis on the m-th segment of the resistance waveform. The obtained spectrum was divided into k sub-bands, and the spectral flatness Fla(m) of each sub-band was calculated. s ), and thus the spectral flatness Fla(m) is obtained: In the formula, R F (f ms ) is R F (f m At frequency f ms The calculated energy amplitude is given by s = 1, 2, ..., k; R F (f m Let f be the frequency domain waveform of the fault resistor in the m-th segment. m f represents the frequency range covered by the entire spectrum of the m-th signal segment. ms Let be the center frequency of the s-th sub-band after the spectrum of the m-th signal is divided into k sub-bands.

6. The method according to claim 1, characterized in that, Step 3 utilizes the multidimensional physical characteristics to detect whether a high-resistance grounding fault has occurred. This is based on the variance characteristic Var(m) and the number of interval slope changes C. t (m), the spectrum flatness Fla(m) corresponds to the high resistance grounding fault characteristics and the fault criterion is set, or deep learning or machine learning is introduced to construct a multi-dimensional physical feature matrix as the model input of deep learning or machine learning; The variance characteristic Var(m) corresponds to the high-resistance grounding fault characteristics as follows: the resistance waveform of a high-resistance grounding fault exhibits high variance. Number of times the slope changes in the interval C t The characteristics of the high-resistance grounding fault corresponding to (m) are as follows: Within half a cycle, the interval slope curve of the high-resistance grounding fault under zero rest duration has multiple extreme points, and the number of times the interval slope changes is C. t (m) more than 2 times; The high-resistance grounding fault characteristics corresponding to the spectral flatness Fla(m) are: the values ​​of all elements in the spectral flatness Fla(m) under the high-resistance grounding fault are close in magnitude and not equal to 0.

7. The method according to claim 1, characterized in that, Step 1 also includes: processing the fault resistor waveform sequence R. F Preprocessing is performed, namely, using a low-pass filter to remove high-frequency noise and using mathematical morphology filtering to eliminate short-term irregular distortion.

8. A system based on the method of any one of claims 1-7, characterized in that: include: The resistance waveform data acquisition module is used to acquire / obtain the fault resistance waveform sequence R of the distribution network. F ; A multi-dimensional physical feature extraction module for resistance waveforms is used to extract the fault resistance waveform sequence R within a preset window length. F The data is divided into continuous segments, and then multi-dimensional physical features are extracted for each segment. The multidimensional physical features are: variance feature Var(m), number of times the slope changes in the interval C. t (m), spectral flatness Fla(m), where m is the marker of the data segment; The fault detection module is used to detect whether a high-resistance grounding fault has occurred by utilizing the multi-dimensional physical characteristics.

9. A computer terminal, characterized in that: include: One or more processors; A memory that stores one or more computer programs; The processor invokes a computer program to achieve the following: The steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer program is stored and is invoked by the processor to implement: The steps of the method according to any one of claims 1-7.

Images