A power distribution network voltage sag feature evaluation method, system and device

Abstract

The application discloses a new energy distribution network voltage sag feature extraction method, system and device in the technical field of power quality detection, first calculates the total harmonic distortion rate THD of the voltage sag signal, and then adjusts and improves the S transform Gaussian window width according to λ = β / ( α +THD) online The application realizes the window width self-adaption of the whole signal level through the THD reciprocal feedback, does not need offline parameter optimization or window type replacement, solves the problems that the existing improved S transform cannot dynamically trade off the time-frequency resolution with the harmonic distortion degree and cannot balance the harmonic separation and short voltage sag positioning in the new energy inverter harmonic superposition voltage sag scene, and significantly improves the feature extraction precision in the new energy high harmonic distortion scene.

Description

Technical Field

[0001] This invention relates to the field of power quality detection technology, and in particular to a method, system and medium for evaluating voltage sag characteristics of distribution networks with the involvement of new energy sources and sensitive loads. Background Technology

[0002] With the deepening implementation of the "dual-carbon" strategy, my country's power system is accelerating its transformation into a new type of power system dominated by new energy sources. Distributed power sources such as photovoltaic and wind power, with their advantages of being clean, flexible, and readily absorbed locally, are seeing a continuous increase in their penetration rate in medium- and low-voltage distribution networks. At the same time, the scale and proportion of sensitive loads such as semiconductor manufacturing, precision machining, and data centers are also growing rapidly. Accurately assessing the characteristic quantities of voltage sags is a key prerequisite for identifying sag sources, analyzing propagation characteristics, and making mitigation decisions, and is of great significance for ensuring the safe and stable operation of sensitive loads.

[0003] Regarding methods for evaluating voltage sag characteristics, national standards such as GB / T 30137, in different years and time-domain monitoring routes using the root mean square algorithm with added thresholds and hysteresis, tend to underestimate the sag amplitude while overestimating the residual high voltage. When harmonic distortion from new energy sources is present, this method itself is insensitive to harmonics, cannot extract fine phase transition values, and is not suitable for high-precision evaluation scenarios. In the time-frequency analysis route, traditional Discrete Fourier Transform (DFT) and Fast Fourier Transform (FFT) can obtain global frequency domain information of the signal, but cannot effectively detect abrupt changes in the signal, making it difficult to locate the start and end times and transient characteristics of sags. Although Short Time Fourier Transform (STFT) can simultaneously preserve time and frequency domain information, its window function is fixed, and the time-frequency resolution cannot be dynamically adjusted according to signal characteristics. Wavelet transform has multi-resolution analysis capabilities, but the feature evaluation effect heavily depends on the selection of the mother wavelet and lacks adaptability. The HHT transform proposed by Huang E in 1998, combined with Hilbert transform, is suitable for non-stationary signal analysis, but suffers from problems such as mode mixing and endpoint effects. The S-transform, proposed by Stockwell et al. in 1996, combines the advantages of the short-time Fourier transform and the wavelet transform. It employs a Gaussian window function that automatically adjusts with frequency—wide window and high frequency domain resolution at low frequencies, and narrow window and high time domain resolution at high frequencies—capable of balancing the analytical needs of both the time and frequency domains while preserving the absolute phase information of the signal. Therefore, it has been widely used in voltage sag signal analysis. Dash et al. were among the first to introduce the S-transform into power quality disturbance analysis, demonstrating its advantages in disturbance detection, location, and classification, laying the foundation for extracting voltage sag features based on the complex time-frequency matrix of the S-transform.Subsequently, several improved schemes were proposed to address the problems of large computational load and insufficient flexibility of the standard S-transform and Gaussian window function. For example, the paper "Power Quality Disturbance Detection Method Using Incomplete S-Transform" published in *High Voltage Engineering*, Vol. 35, No. 10, 2009, and the patent document with authorization announcement number CN106548013B, which discloses an improved voltage sag source identification method using incomplete S-transform based on FFT power spectrum, are both based on incomplete S-transform of FFT power spectrum. The patent document with authorization announcement number CN104391207B discloses a voltage sag detection method using fundamental frequency single-vector S-transform, which shortens the computation time by solving only the fundamental frequency vector and improves the low-frequency time domain resolution by using a fixed empirical window width coefficient to increase the Gaussian window. (This is from *Power System Protection and Control*, 201...) The following papers, including "Voltage Sag Identification Based on Improved S-Transform" published in Volume 41, Issue 17, 2013; "Improved S-Transform Method and Its Application in Multi-Feature Collaborative Analysis of Power Quality" published in Yanshan University in 2017; "Study on Voltage Sag Characteristics Based on Improved S-Transform-TT Transform" published in Volume 42, Issue 4, 2019 of "Electrical Transmission"; "Improved S-Transform Voltage Sag Source Identification by Introducing Adjustment Factors" published in Volume 51, Issue 18, 2021 of "Electrical Transmission"; and a voltage sag detection method based on multi-attribute decision-making and improved S-transform method disclosed in patent document CN111257619B, which improves the accuracy of sag feature extraction to a certain extent by introducing multiple adjustment factors to optimize the Gaussian window shape, etc.

[0004] However, existing inventions and research still have some shortcomings: (1) The incomplete S-transform window function lacks adaptive adjustment capability. The Gaussian window shape of the standard S-transform is fixed, and the window width is only inversely proportional to the frequency. It fails to establish an adaptive feedback relationship between the window width and the real-time distortion degree of the signal. When there is harmonic distortion, a narrow frequency domain window is needed to extract the fundamental amplitude to separate harmonics, while a narrow time domain window is needed to improve the time resolution to locate the start and end times of sag. The standard S-transform cannot meet these contradictory requirements, resulting in a significant increase in the evaluation error of duration or sag depth. Some use the frequency range as the feedback variable, while others use offline global constants. They cannot be dynamically adjusted according to the real-time distortion degree of each sag event. The optimized adjustment factors are mostly artificially preset constants or artificially switched based on the local energy of the signal. In addition, the computational load of the two-factor optimization or the scheme that requires superimposed TT transform is large, which is not conducive to deployment. Some solutions employ generalized S-transform followed by parameter optimization, windowing, or standard S-transform results for multi-dimensional feature mining, which only improves the mathematical performance of time-frequency aggregation. This relies on the mathematical and statistical characteristics of the signal itself and is a static or pre-set optimization. The adjustment is blind and cannot be dynamically fed back by power quality indicators in the actual operating environment of the power grid. Without verification and closed-loop correction, such open-loop adaptive algorithms are susceptible to noise or harmonic interference, leading to adjustment failure and inability to accurately capture the true sag characteristics.

[0005] (2) None of them have made special optimizations for waveform distortion caused by the access of new energy and sensitive loads. New energy is connected to the grid through power electronic inverters. The high-frequency switching of switching devices generates a large amount of harmonic current, which causes complex waveform distortion to be superimposed on the sag signal. The inverter harmonic spectrum and sag coexist. The fixed window width cannot effectively separate each harmonic. The fundamental amplitude extraction accuracy is insufficient. In addition, short-term fault ride-through, reclosing, etc. can cause the sag duration to be as short as a few milliseconds. Traditional evaluation methods are not capable of processing such complex signals with distortion and low signal-to-noise ratio. The feature extraction accuracy is difficult to meet the engineering requirements. Summary of the Invention

[0006] To address the technical problem of insufficient feature extraction accuracy of traditional methods under waveform distortion caused by the integration of new energy sources and sensitive loads, this application provides a method and apparatus for evaluating voltage sag characteristics of distribution networks containing new energy sources and sensitive loads. On one hand, a method for evaluating voltage sag characteristics of distribution networks containing new energy sources and sensitive loads is provided, comprising the following steps: S1 sets the sampling frequency and continuously acquires voltage signals to establish a pre-event loop buffer; S2 calculates the root mean square value of the power frequency cycle according to the update cycle to determine whether a voltage sag event is triggered. Before the S3 freeze event, the steady-state window is used to obtain the total harmonic distortion rate index, which characterizes the degree of harmonic distortion in the power grid, and the window width adjustment factor is determined. l The total harmonic distortion rate is determined by the ratio of the root sum of squares of the amplitude at each frequency point within the frequency band to the root mean square of the fundamental frequency. The window width adjustment factor... l According to the preset proportional coefficient β With total harmonic distortion and preset correction factors α The ratio of the sums is fixed and remains constant throughout this event; S4 continuously records until the voltage recovers to form a complete event window containing data before and after the event. The complete event window includes at least one power frequency cycle before the start of the voltage dip, the voltage dip process, and one power frequency cycle after the end of the voltage dip. S5 uses a discrete improved S-transform based on harmonic distortion rate adaptive adjustment of the window width to extract the complex time-frequency coefficients at the fundamental frequency of the complete event window; S6 calculates the residual voltage, voltage sag depth, duration, and phase jump of the voltage sag. The residual voltage is determined by an aggregation function based on the normalized per-unit sequence of the fundamental amplitude during the sag period. The voltage sag depth is determined based on the difference between the reference voltage per-unit value and the residual voltage per-unit value. The duration event is determined by first calculating the fundamental amplitude difference sequence, then locating the sampling point indices corresponding to the minimum and maximum values ​​of the difference sequence to obtain the start and end times, and then calculating the difference between the start and end times. The phase jump is determined by first performing phase dewinding processing on the phase sequence to obtain continuous phases, and then calculating the maximum absolute value of the difference between the phase at each point during the sag and the initial phase.

[0007] S7 outputs a record of temporary drop event characteristics.

[0008] As a further aspect of the present invention, the correction coefficient α The value ranges from 0.08 to 0.25, and the proportionality coefficient is... β The value range is 0.14 to 0.50.

[0009] As a further aspect of the present invention, the α The value is 0.15. β The value is 0.24.

[0010] As a further aspect of the present invention, the residual voltage is taken as the arithmetic mean of the minimum value of the normalized per-unit value sequence of the fundamental amplitude during the sag period, within one fundamental frequency cycle before and after the corresponding time.

[0011] As a further aspect of the present invention, the residual voltage is taken as the minimum value of the normalized per-unit value sequence of the fundamental amplitude during the sag period.

[0012] As a further aspect of the present invention, the total harmonic distortion rate is represented by the symbol THD, and is expressed using the formula... The calculation involves substituting THD into the window width adjustment factor calculation formula in decimal form, where... This is the root mean square value of the fundamental voltage. h For the harmonic component order, For the first h Root mean square value of the second harmonic component H To achieve the highest harmonic order in the analysis, the window width adjustment factor is calculated using the formula... l = β / ( α +THD) calculation.

[0013] As a further aspect of the present invention, the discrete improved S-transform form is as follows: In the formula i For time sampling index value, T The sampling time interval, n For frequency sampling index value,N The number of sampling points. k It is a cyclic shift index in the frequency dimension.

[0014] As a further aspect of the present invention, the discrete improved S-transform only calculates one row of S-transform coefficients corresponding to the fundamental frequency.

[0015] On the other hand, a distribution network voltage sag characteristic assessment system is provided, comprising a voltage sampling module, a sag event detection module, a harmonic distortion rate calculation and window width adjustment factor determination module, a complete event window determination module, an improved S-transform module, a feature calculation module, and a result output module connected in sequence. The voltage sampling module is used to perform step S1 of the aforementioned distribution network voltage sag characteristic assessment method; the event detection module is used to perform step S2 of the aforementioned distribution network voltage sag characteristic assessment method; the harmonic distortion rate calculation and window width adjustment factor determination module is used to perform step S3 of the aforementioned distribution network voltage sag characteristic assessment method; the complete event window determination module is used to perform step S4 of the aforementioned distribution network voltage sag characteristic assessment method; the improved S-transform module is used to perform step S5 of the aforementioned distribution network voltage sag characteristic assessment method; the feature calculation module is used to perform step S6 of the aforementioned distribution network voltage sag characteristic assessment method; and the result output module is used to perform step S7 of the aforementioned distribution network voltage sag characteristic assessment method.

[0016] On the other hand, a distribution network voltage sag characteristic assessment device is provided, including a memory and a processor; the memory is used to store a computer program; the processor is used to implement the distribution network voltage sag characteristic assessment method as described above when the computer program is executed.

[0017] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: This paper presents a method for evaluating voltage sag characteristics in distribution networks containing new energy sources and sensitive loads. The method collects voltage sag signals and calculates harmonic distortion rate. Based on this, the window width adjustment factor of the improved S-transform is adaptively adjusted so that the time-frequency resolution dynamically matches the waveform distortion degree. The adjusted improved S-transform is used to extract sag depth, duration, and phase jump characteristics, which are then input into a propagation characteristic analysis model. The propagation trajectory and influence range of the sag are output through association rules, thereby improving the accuracy of sag characteristic evaluation.

[0018] The improved S-transform proposed in this application, based on adaptive window width adjustment according to harmonic distortion rate, achieves dynamic optimization of time-frequency resolution while maintaining the advantages of the standard S-transform, thus improving the accuracy and robustness of voltage sag feature extraction. It not only significantly improves the accuracy of locating the start and end times of sag signals, solving the problems of long time-domain transitions and large detection errors in short sag durations in the standard S-transform, but also enhances the accuracy of sag depth assessment for composite sag signals with waveform distortion, making it suitable for sag scenarios with power electronic harmonics prevalent in distribution networks with renewable energy and sensitive loads. Furthermore, it maintains high feature extraction accuracy even under strong noise environments, with all feature evaluation errors within acceptable engineering ranges.

[0019] This application firstly achieves adaptive dynamic adjustment of the S-transform window width. Using harmonic distortion rate as the core indicator, this scheme establishes an adaptive mapping relationship between the window width adjustment factor and the degree of waveform distortion, enabling the time-frequency resolution to dynamically match the signal characteristics, thus balancing the dual requirements of accurate extraction of the fundamental amplitude and precise location of abrupt changes.

[0020] Meanwhile, this application effectively solves the challenge of characteristic evaluation of complex sag signals in distribution networks with renewable energy sources and sensitive loads. It is specifically optimized to address harmonic distortion and noise interference introduced by inverters, resulting in minimal increase in computational load. Even in complex scenarios with waveform distortion and low signal-to-noise ratios, it maintains high extraction accuracy, making it suitable for sag evaluation needs in distribution networks with a high proportion of renewable energy sources and sensitive loads.

[0021] This scheme does not improve the S-transform. Instead, it uses adaptive feedback variables. Instead of selecting statistical mathematical indicators such as kurtosis, which are sensitive to spikes but insensitive to steady-state harmonic energy, it chooses the Total Harmonic Distortion (THD), a physically perceptible parameter that aligns with the requirement for frequency domain resolution harmonic separation. Furthermore, it incorporates... l The reciprocal feedback relationship, unlike linear or other nonlinear relationships, does not overshoot when THD is high, leading to an excessively wide window, excessively wide frequency domain, and loss of time domain. The feedback relationship of this scheme satisfies the condition when THD is low. l No explosion ，α To prevent division by zero, narrow windows, THD approaches 0 when it is high, wide windows. l A two-way trade-off is achieved, thus enabling event-level adaptive dynamic adjustment of the S-transform window width. For different sag events with varying total harmonic distortion (THD), the THD is calculated based on the steady-state THD. lThe value is adjusted so that the time-frequency resolution dynamically matches the degree of waveform distortion of the event, while λ remains constant within the same event to avoid contaminating the total harmonic distortion (THD) estimation due to transient voltage sags. This achieves event-level online adaptiveness and resolves the contradiction in the improved S-transform for scenarios involving new energy sources and sensitive loads connected to the distribution network. The improved S-transform requires both high-resolution frequency domain separation of harmonics, resulting in a wide window that smooths the start and end of transient voltage sags, and high-resolution time domain positioning of transient voltage sags, resulting in a narrow window that obscures harmonic separation. This significantly improves the accuracy and efficiency of voltage sag feature extraction, particularly in scenarios involving new energy sources and sensitive loads connected to the distribution network. For different mechanism methods such as wavelet transform and HHT, this application further compares them at the processing mechanism level, as shown in Table 1.

[0022] Table 1. Comparison of mechanisms of different transient descent characteristic analysis methods Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating one embodiment of the method of this solution; Figure 2 This is a schematic diagram of the fundamental amplitude curves before and after a voltage sag; Figure 3 This is a schematic diagram of the fundamental amplitude difference curve before and after a voltage sag; Figure 4 A schematic diagram of unwinding the phase curve of the fundamental signal; Figure 5 This is a schematic diagram of the time-domain waveform of the test signal 1, which is a harmonic-free distortion voltage sag used in the effectiveness verification of this scheme. Figure 6 This is a schematic diagram of the time-domain waveform of test signal 2 containing the 3rd and 5th harmonic voltage dips used in the effectiveness verification of this scheme. Detailed Implementation

[0025] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0026] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0027] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0028] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0029] This invention provides a method, system, and medium for evaluating voltage sag characteristics in distribution networks, such as... Figure 1 As shown, it includes the following steps: Step S1: Set the sampling frequency and continuously acquire voltage signals. The sampling frequency is determined according to the following formula: , , .

[0030] In the formula, The sampling period is Sampling frequency, The system's rated frequency, for example, 50Hz. H To analyze the highest harmonic order, To allow for acceptable positioning errors at start and end times, a target time resolution can be set. H satisfy ; The update cycle for event detection results; M It represents the number of sampling points included in an event detection update cycle. It is a positive integer and can be used to reduce computational load.

[0031] Given the requirements for computational speed and power quality analysis, this embodiment adopts... H =50, based on the requirement of at least 50 measurements for Class A measurement methods in the national standard GB / T17626.30—2023, preferred. f s The range is 6.4–20 kHz. Ts The range is 0.05–0.20 ms. T u The range is 0.5 to 5 ms. M Range 10–25.

[0032] Step S2: Continuously collect the voltage of the monitoring points at a fixed sampling period and write it to the loop buffer. Each time... T u Only the RMS value of the power frequency cycle and the event status are updated once; a complete S-transform is not performed. When the RMS value of the power frequency cycle is continuous... p The second time below the descent trigger threshold U th At that time, it was determined that a temporary descent had begun. U th According to national standards, the range can be 0.85–0.9 pu, with 0.9 pu being preferred. p The value range is 1 to 3.

[0033] Step S3: After temporary descent is triggered, before the freeze event. L A steady-state data window of a power frequency frequency is used to calculate the harmonic distortion rate and determine the corresponding window width adjustment factor introduced based on the standard S-transform Gaussian window function for this event. l ,and l The data window remains unchanged throughout the event to avoid contaminating the harmonic distortion rate estimation due to transient sags. A continuous, seamless rectangular window is used; the harmonic distortion rate is calculated only once after each sag event, without window sliding. When the highest analyzed harmonic order H is 50, this is the preferred setting. L Ten power frequency cycles were selected to match the 10-cycle seamless harmonic measurement window.

[0034] This scheme does not assume that the THD before and during a voltage dip is necessarily continuous. The THD obtained in the pre-event window is defined as the background harmonic distortion rate, used to characterize the grid harmonic environment and inverter operating state that persist before the voltage dip occurs. For a brief voltage dip lasting only a few power frequency cycles, the voltage dip process itself is insufficient to form a long data window that meets the requirements for stable harmonic measurement, and the non-stationary components generated by the sudden start and end of the fault, DC bias, and control switching will contaminate the THD estimation. Therefore, the first... l Using the background THD before the event, i.e., the total harmonic distortion before the sag. THD pre The event is identified and frozen within the event period. As an optional verification method, when the sag process includes a relatively stable intermediate segment of no less than 5 power frequency cycles, the total harmonic distortion rate during the sag process can be calculated after the event ends. THD sag If | THD sag -THDpre |Exceeding the preset difference threshold e THD , e THD A value of 0.02 to 0.05 can be used, based on... THD sag Recalculate l An improved S-transform is then performed on the complete event window; when the duration of the sag is less than 5 power frequency cycles, the THD determined before the event is maintained. l .

[0035] Due to the window width adjustment factor l When the value is large, a narrower time-domain window results in higher time-domain resolution, shorter transition processes, and improved accuracy in extracting the duration of the sag. However, this reduces frequency-domain resolution, making the fundamental voltage more susceptible to interference. l When the voltage is small, its frequency domain window is narrow, which can accurately separate harmonic components. However, the time domain transition process is relatively long, and it cannot accurately extract the characteristic quantities of some short-duration sags. Given that the harmonic content of distribution networks containing new energy sources and sensitive loads is higher, and the voltage waveform distortion is more severe, it is even more difficult to accurately extract and evaluate sag characteristics. Therefore, this application introduces the voltage harmonic distortion rate (THD for common harmonics). u (represented as total harmonic distortion and as a window width adjustment factor) l The parameters to be determined. Voltage harmonic distortion rate (THD) u In this scheme, it will be referred to as harmonic distortion rate (THD). Harmonic distortion rate is a core indicator for measuring the harmonic content of a signal. It is defined as the ratio of the square root of the sum of the squares of the root mean square values ​​of all harmonic components in the signal to the root mean square value of the fundamental voltage. Its formula can be obtained by Fourier transform according to the appendix of the national standard GB / T 14549—93 Power Quality - Harmonics in Public Power Grids. In the formula, U 1 This is the root mean square value of the fundamental voltage; h For the harmonic component order, U h For the first h Root mean square value of the second harmonic component; H To analyze the highest harmonic order, the numerator of the entire formula is the root mean square value of each harmonic component (i.e., the root of the sum of the squares of the effective values ​​of each harmonic). In this embodiment, THD is substituted in decimal form. l formula.

[0036] When the THD is low, the signal is dominated by the fundamental frequency with low harmonic content. In this case, it is necessary to improve the time domain resolution to accurately extract the start and end times and amplitude changes of the transient sag. A larger resolution should be selected. lWhen the THD is high, the signal contains a large number of harmonic components. In this case, it is necessary to improve the frequency domain resolution and separate the harmonics. Therefore, a smaller value should be selected. l Value. Therefore, this application proposes the following formula: In the formula: α To correct the coefficient and prevent it from being zero when THD is zero. l Infinity, and adjust the sensitivity of the window width adjustment factor to changes in harmonic distortion rate; β This is a proportionality coefficient used for adjustment. l The overall range of values. α The preferred value range is 0.08 to 0.25. β The preferred value range is 0.14 to 0.50. Extensive simulation testing has shown that... α =0.15, β At a value of 0.24, it exhibits good responsiveness to various voltage sag signals. Furthermore, since Fourier transform is a necessary step in performing S-transform, the harmonic distortion rate (THD) can be directly obtained from the Fourier transform and then substituted into the above equation to obtain the final value. l Therefore, the increase in computation is not significant, thus balancing computation speed.

[0037] For ultra-high harmonics, the harmonic distortion rate is represented by the symbol TD. The TD formula in "T / CPSS 1010—2023 Requirements for Measurement of Ultra-High Harmonics in Power Systems" can be used to replace the THD formula in this application.

[0038] Step S4: Continue collecting data until the effective value of the power frequency frequency is continuous. q The recovery will eventually reach the recovery threshold. U rec The above determines the temporary descent to be over. The event analysis window should include at least one power frequency cycle before the temporary descent begins, the temporary descent process, and one power frequency cycle after the temporary descent ends. U rec 0.9 pu is also preferred. q The value range is also 1 to 3.

[0039] Step S5: Extract the complex time-frequency coefficients at the fundamental frequency using an improved S-transform with adaptive window width adjustment based on harmonic distortion rate. After the sag is determined, use the event analysis window determined in step S4 as input and the window width adjustment factor determined in step S3. l An improved S-transform is performed on the voltage sampling sequence within the complete event window.

[0040] Based on the standard S-transform Gaussian window function, a window width adjustment factor is introduced. l This makes the rate at which the time window width in the Gaussian window changes with the signal frequency more controllable. The improved S-transform expression is as follows: S( t , f () represents the elements of the complex matrix obtained after the improved S-transformation. t For continuous time variables, x ( t () represents the continuous voltage signal to be analyzed. This is the time shift factor, used to control the position of the Gaussian window in the time domain, corresponding to the time axis of time-frequency analysis; f For frequency analysis, 0 < f≤f s / 2. Used to filter signals with frequencies of f The components correspond to the frequency axis in time-frequency analysis; their discrete form is: In the formula i For time sampling index value, T The sampling time interval, n For frequency sampling index value, N The number of sampling points. k For frequency-dimensional cyclic shift index, NT Total duration of the event window. The frequency domain complex sequence obtained after performing a discrete Fourier transform on the original voltage signal. Represents at discrete time points IT Discrete frequency Improved S-transform complex value at the location.

[0041] In practical engineering implementation, the aforementioned discrete improved S-transform employs a fast algorithm based on FFT: first, FFT is performed on the voltage sampling sequence within the event window to obtain the frequency domain sequence; then, for each frequency index, a frequency domain Gaussian window is constructed and compared with... Multiply the results; finally, perform an IFFT on the product to obtain the complex S-transform coefficients of the frequency component at each time point. Repeat the above process to traverse all frequency indices to obtain the complete complex time-frequency matrix. This implementation utilizes the computational efficiency of FFT, keeping the computational complexity of a single improved S-transform lightweight.

[0042] Compared to the standard S-transform, the improved S-transform, through adjustment factors... l The ratio of Gaussian window width to frequency was changed: when l When the value is greater than 1, the Gaussian window width narrows, and the time-domain resolution improves; when... l When the value is less than 1, the Gaussian window width widens, improving the frequency domain resolution. This flexible adjustment mechanism allows the improved S-transform to match the optimal time-frequency resolution based on the signal characteristics.

[0043] To reduce the computational burden, this scheme employs FFT / IFFT to implement the discrete improved S-transform. The pre-event THD calculation requires only one N-point FFT; since the sag depth, duration, and phase transition are all derived from the fundamental complex time-frequency coefficients, the preferred implementation only calculates the fundamental frequency. f The row of S-transform coefficients corresponding to 0, i.e., after one spectral windowing and one N-point IFFT, yields the fundamental complex time-frequency coefficient sequence S[ IT , f [0], This sequence is a complex sequence, with its real and imaginary parts containing the amplitude and phase information of the fundamental component, respectively. Generating a complete N×N time-frequency matrix requires performing IFFT frequency-by-frequency, significantly increasing computational and storage requirements; therefore, it is not a preferred method for online evaluation. The update cycle of event detection results... T u Internally, only the sliding effective value and state judgment are performed; the complete improved S-transform is executed once after the event ends; during online scrolling output, to ensure that the calculation is timely, the single window calculation time is consumed. t cal Should meet t cal ≤ T roll ,in T roll For the update cycle of the scrolling results, T roll It ranges from 10 to 100 ms, but is typically 20 ms.

[0044] From the complex time-frequency matrix output by the improved S-transform, the row vector corresponding to the fundamental frequency is extracted to obtain the fundamental complex time-frequency coefficient sequence.

[0045] Step S6: Calculate the residual voltage, voltage sag depth, start and end times, duration, and phase jump of the voltage sag. The residual voltage is determined by an aggregation function based on the normalized per-unit sequence of the fundamental amplitude during the sag period. The voltage sag depth is determined based on the difference between the reference voltage per-unit value and the residual voltage per-unit value. The duration event is determined by first calculating the fundamental amplitude difference sequence, then locating the sampling point numbers corresponding to the minimum and maximum values ​​of the difference sequence to obtain the start and end times, and then calculating the difference between the start and end times. The phase jump is determined by first performing phase dewinding processing on the phase sequence to obtain continuous phases, and then calculating the maximum absolute value of the difference between the phase at each point during the sag and the initial phase.

[0046] The three most important parameters to be evaluated for voltage sags are: the depth of the voltage sag, the duration of the voltage sag, and the phase transition.

[0047] The voltage sag depth (hereinafter referred to as sag depth) is the difference between the reference voltage and the residual voltage. The amplitude of the S-transform at a certain frequency band is always half of the peak voltage corresponding to that frequency band, and the magnitude of the window width adjustment factor does not change this characteristic of the S-transform. Therefore, multiplying the fundamental amplitude matrix... Then divide by the reference voltage U B This yields the per-unit value of the fundamental voltage at each moment, so the fundamental voltage amplitude-time curve reflects the relationship between the per-unit value of the fundamental component and time. However, since the Gaussian window of the S-transform essentially calculates the moving weighted average of the data within the window, the fundamental amplitude curve will exhibit a transitional change at the start and end of the sag, rather than an actual instantaneous abrupt change. Therefore, when the sag duration is short, the reflection of the sag depth may be underestimated, and the residual voltage will be overestimated. Improving the S-transform can be achieved by using a larger window width adjustment factor. l This improves time-domain accuracy and solves the above problems.

[0048] Based on the fundamental amplitude after S-transform S ( IT , f 0), the per-unit value of the residual voltage during the sag is In the formula: T s and T e These represent the start and end times of the voltage sag, respectively. S ( IT , f 0)|Reflection Moment IT The fundamental voltage amplitude at that location, U B This is the root mean square value of the reference voltage. That is, the voltage amplitude at the fundamental frequency is extracted at each sampling point, converted and normalized to a per-unit value, and then the minimum value during the sag period is taken as the residual voltage. The final calculated per-unit value of the residual voltage satisfies 0 ≤ U res ≤1. The above formula is applicable to noise-free or low-noise scenarios.

[0049] To further improve the noise resistance of feature extraction, the time corresponding to the minimum voltage during the sag period can also be used. t min The residual voltage is represented by the average amplitude of the fundamental voltage within one power frequency cycle before and after, to adapt to noisy scenarios. The number of sampling points corresponding to one power frequency cycle is: f s / f If the value is 0, then the residual voltage is: The descent depth is then: U sag =1- U resThat is, the sag depth is determined based on the difference between the per-unit value of the reference voltage and the per-unit value of the residual voltage.

[0050] By combining a large window width adjustment factor, the detection error caused by insufficient amplitude attenuation during short durations of decline can be effectively solved.

[0051] The duration of a voltage sag is determined by the difference between the start and end times of the voltage sag. Considering that the fundamental voltage amplitude-time curve has high energy and strong noise immunity, although its change is relatively slow, the abrupt change time can be located using the rate of change of the fundamental amplitude. In this embodiment, the fundamental amplitude difference curve is obtained by performing a differential operation on the fundamental voltage amplitude:

[0052] In the formula, ΔS(i) The differential characteristic quantity representing the fundamental amplitude sequence is used to characterize the sampling index. i The rate of change of voltage amplitude i For sampling index value, and The sampling index is respectively i and sampling index is i The fundamental voltage of two adjacent sampling points of +1 T The sampling time interval, f 0 The fundamental frequency, N The number of sampling points. i The range of values ​​can prevent array out-of-bounds errors.

[0053] When a voltage sag occurs, the fundamental voltage amplitude drops rapidly, and the differential curve shows a negative peak; when the sag ends, the fundamental voltage amplitude recovers, and the differential curve shows a positive peak, as shown below. Figure 2 and Figure 3 As shown. Therefore, the maximum negative peak of the difference curve corresponds to the fault initiation time. T s The maximum positive peak value of the difference curve corresponds to the time when the fault ends. T e ,Right now: in, This indicates the sampling point number corresponding to the maximum negative peak value of the difference curve. This indicates the sampling point number corresponding to the maximum positive peak value of the difference curve.

[0054] The duration of the temporary descent T c for: Phase jump value Df The formula for calculating the difference between the fundamental voltage phase and the initial phase before and after a voltage dip is as follows: In the formula f [ IT , f 0 ] represents discrete time. IT Fundamental frequency f 0 Real-time phase at that location, f 0 represents the initial phase. In this application, the initial phase refers to the average phase intercept of the entire steady-state window before the event, after deducting the linear phase term at the rated frequency. However, because the range of the arctangent atan2 function in the four quadrants is (- π,π When the phase change exceeds this range, a jump will occur; for example, the phase changes from... π Directly becomes - π This is not an actual voltage phase change. Therefore, the original phase sequence must first be dewound to obtain the true continuous phase. A slope of 2 is drawn at each point on the dewound curve. πf 0 / f s The intercept of this straight line on the vertical axis is the initial phase at time 0 at the corresponding point. Therefore, the maximum and minimum values ​​of the tangent intercepts at each point on the unwinding curve during the temporary sag are the maximum and minimum values ​​of the initial phase after the fault, and the change in the intercept is the magnitude of the phase jump, such as... Figure 2 As shown.

[0055] Step S7: Output voltage sag event characteristic records for propagation analysis and sensitive load impact assessment. These records may include event number, start time, end time, duration, residual voltage, sag depth, phase jump, harmonic distortion rate, window width adjustment factor, etc. Event data is aggregated according to standards such as "T / CPSS 1011—2025 Voltage Sag Event Data Aggregation Method," input into the propagation characteristic analysis model, and output the sag propagation trajectory and impact range using a voltage sag severity assessment method and equipment based on association rules, as disclosed in patent document CN112052887B. This provides standardized input for subsequent sag propagation analysis, severity assessment, and sensitive load impact assessment conforming to "GB / T 39270—2020 Voltage Sag Index and Severity Assessment Method," providing a reference for power grid companies in making governance decisions. It can also serve as input parameters for sensitive load withstand capability assessment, guiding the development of sensitive load governance plans. The characteristic records also meet the data requirements for monitoring and predictive assessment in the power quality assessment technical guidelines.

[0056] Steps S5-S7 perform an improved S-transform on the complete event window to extract the slump depth, duration, and phase transition, forming an event feature record. Thus, as the original sampling points continuously arrive, only buffering and lightweight event detection are performed. The complete S-transform is executed once per slump event by default, instead of being re-executed for each incoming sampling point, significantly reducing computational overhead while maintaining feature extraction accuracy.

[0057] The second embodiment of this application provides an embodiment of a distribution network voltage sag characteristic assessment system, including a voltage sampling module, a sag event detection module, a harmonic distortion rate calculation and window width adjustment factor determination module, a complete event window determination module, an improved S-transform module, a feature calculation module, and a result output module connected in sequence. The voltage sampling module is used to perform step S1 of the aforementioned distribution network voltage sag characteristic assessment method. The event detection module is used to perform step S2 of the aforementioned distribution network voltage sag characteristic assessment method. The harmonic distortion rate calculation and window width adjustment factor determination module is used to perform step S3 of the aforementioned distribution network voltage sag characteristic assessment method. The complete event window determination module is used to perform step S4 of the aforementioned distribution network voltage sag characteristic assessment method. The improved S-transform module is used to perform step S5 of the aforementioned distribution network voltage sag characteristic assessment method. The feature calculation module is used to perform step S6 of the aforementioned distribution network voltage sag characteristic assessment method. The result output module is used to perform step S7 of the aforementioned distribution network voltage sag characteristic assessment method.

[0058] The third embodiment of this application provides an embodiment of a distribution network voltage sag characteristic assessment device, including a memory and a processor; the memory is used to store a computer program; the processor is used to implement the distribution network voltage sag characteristic assessment method as described above when the computer program is executed. The memory can be an erasable programmable read-only memory (EPROM), a flash memory, a hard disk, etc., and the processor can be a central processing unit (CPU), an ARM architecture microprocessor (MCU), etc.

[0059] To verify the effectiveness of the improved S-transform, this scheme uses test signal 1 without waveform distortion and test signal 2 with harmonic distortion, respectively, in a noise-free environment and in a noise environment with a signal-to-noise ratio of 40dB, 30dB, and 20dB, to compare the standard S-transform and the improved S-transform with the same parameters and to analyze the error of each characteristic quantity.

[0060] Rated frequency f 0 is taken as 50Hz, sampling frequency f s Take 10kHz, sampling period T s The time interval is set to 0.1 ms, and the highest harmonic order H is set to 50. α=0.15、 β =0.24、 l min =0.4、 l max =1.6. Both test signals exhibit a residual voltage dip of 0.5 pu between 0.6 s and 1.2 s, therefore their theoretical dip depth is 0.5 pu, theoretical duration is 600 ms, and theoretical phase transition is 0 rad. The test signals are as follows: Test signal 1: Test signal 2: Test signal 1 contains only a 50Hz fundamental frequency; test signal 2 superimposes harmonic terms with amplitudes of 10% and 5% of the fundamental frequency onto the fundamental frequency. In the original formula... and These correspond to the 3rd and 5th harmonics, respectively, instead of the 6th and 10th harmonics. Therefore, the theoretical basis for testing signal 2 is... THD for That is, 11.18%, corresponding to λ≈ 0.917; Test signal 1 THD =0, corresponding to l =1.6. In the formula, 0.6s and 1.2s, if written as 600 and 1200, have the time unit of ms. When substituting them into the sine term, they should be converted to s.

[0061] Based on simulation analysis, the average error of the characteristic quantities of each evaluation method is shown in Table 2. The values ​​in the table are all absolute values ​​of the difference between the estimated value and the theoretical value mentioned above.

[0062] Table 2. Characteristic errors under different conditions

[0063] It can be seen that: for test signal 1 under noise-free conditions, the absolute error of the duration of the standard S-transform is 2.00 ms, while that of the improved S-transform is 0.50 ms, a reduction of 75%; under 20 dB noise, the duration error decreases from 2.15 ms to 0.67 ms. For test signal 2 containing 3rd and 5th harmonics, the sag depth error decreases from 0.003 pu to 0.002 pu under 30 dB noise and from 0.009 pu to 0.006 pu under 20 dB noise, both reductions of approximately 33.3%. The duration error of signal 2 remains approximately 2 ms, because the increased THD... l As the frequency resolution is reduced, the algorithm prioritizes improving it, thus weakening its time-based positioning advantage. The phase jump error is generally similar between the two S-transforms, indicating that the main improvements of this scheme are concentrated on the location of start and end times in low-distortion scenarios and the separation of fundamental amplitude in harmonic scenarios. This scheme achieves adaptive switching between signal segments with different distortion levels, rather than simultaneously achieving optimal time and frequency domains within the same signal segment.

[0064] In summary, this scheme uses THD adaptive determination. l The method improves time-domain resolution when harmonic content is low and frequency-domain resolution when harmonic content is high. It employs event-level computation, single-row FFT / IFFT implementation at the fundamental frequency, and structured result output, providing a clear path for engineering application. The quantitative results in Table 2 demonstrate its improvement over the standard S-transform under the aforementioned specific test conditions.

[0065] The above description is merely a specific implementation of this solution, but the scope of protection of this invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this invention should be included within the scope of protection of this invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A method for evaluating voltage sag characteristics in a distribution network, characterized in that, Includes the following steps: S1 sets the sampling frequency and continuously acquires voltage signals to establish a pre-event loop buffer; S2 calculates the root mean square value of the power frequency cycle according to the update cycle to determine whether a voltage sag event is triggered. Before the S3 freeze event, the steady-state window is used to obtain the total harmonic distortion rate index, which characterizes the degree of harmonic distortion in the power grid, and the window width adjustment factor is determined. λ The total harmonic distortion rate is determined by the ratio of the root sum of squares of the amplitude at each frequency point within the frequency band to the root mean square of the fundamental frequency. The window width adjustment factor... λ According to the preset proportional coefficient β With total harmonic distortion and preset correction factors α The ratio of the sums is fixed and remains constant throughout this event; S4 continuously records until the voltage recovers to form a complete event window containing data before and after the event. The complete event window includes at least one power frequency cycle before the start of the voltage dip, the voltage dip process, and one power frequency cycle after the end of the voltage dip. S5 uses a discrete improved S-transform based on harmonic distortion rate adaptive adjustment of window width to extract the fundamental complex time-frequency coefficients of the complete event window; S6 calculates the residual voltage, voltage sag depth, duration, and phase jump of the voltage sag. The residual voltage is determined by an aggregation function based on the normalized per-unit sequence of the fundamental amplitude during the sag. The voltage sag depth is determined based on the difference between the reference voltage per-unit value and the residual voltage per-unit value. The duration event is determined by first calculating the fundamental amplitude difference sequence, then locating the sampling point indices corresponding to the minimum and maximum values ​​of the difference sequence to obtain the start and end times, and then calculating the difference between the start and end times. The phase jump is determined by first performing phase dewinding processing on the phase sequence to obtain continuous phases, and then calculating the maximum absolute value of the difference between the phase at each point during the sag and the initial phase. S7 outputs a record of temporary drop event characteristics.

2. The method for evaluating voltage sag characteristics in distribution networks according to claim 1, characterized in that: The correction coefficient α The value ranges from 0.08 to 0.25, and the proportionality coefficient is... β The value range is 0.14 to 0.

50.

3. The method for evaluating voltage sag characteristics in distribution networks according to claim 2, characterized in that: The α The value is 0.

15. β The value is 0.

24.

4. The method for evaluating voltage sag characteristics in distribution networks according to claim 1, characterized in that: The residual voltage is the arithmetic mean of the minimum value of the normalized per-unit value sequence of the fundamental amplitude during the sag period, within one fundamental frequency cycle before and after the corresponding time.

5. The method for evaluating voltage sag characteristics in distribution networks according to claim 1, characterized in that: The residual voltage is taken as the minimum value of the normalized per-unit value sequence of the fundamental amplitude during the sag period.

6. The method for evaluating voltage sag characteristics in a distribution network according to claim 2, characterized in that: The total harmonic distortion rate is represented by the symbol THD and is expressed using the formula... The calculation involves substituting THD into the window width adjustment factor calculation formula in decimal form, where... This is the root mean square value of the fundamental voltage. For the harmonic component order, For the first Root mean square value of the second harmonic component To achieve the highest harmonic order in the analysis, the window width adjustment factor is calculated using the formula... λ = β / ( α +THD) calculation.

7. The method for evaluating voltage sag characteristics in distribution networks according to claim 1, characterized in that: The discrete improved S-transform form is as follows: In the formula i For time sampling index value, T The sampling time interval, n For frequency sampling index value, N The number of sampling points. k It is a cyclic shift index in the frequency dimension.

8. The method for evaluating voltage sag characteristics in distribution networks according to claim 7, characterized in that: The discrete improved S-transform only calculates one row of S-transform coefficients corresponding to the fundamental frequency.

9. A distribution network voltage sag characteristic assessment system, characterized in that: The method comprises, in sequence, a voltage sampling module, a sag event detection module, a harmonic distortion rate calculation and window width adjustment factor determination module, a complete event window determination module, an improved S-transform module, a feature calculation module, and a result output module. The voltage sampling module performs step S1 of the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8. The event detection module performs step S2 of the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8. The harmonic distortion rate calculation and window width adjustment factor determination module performs step S3 of the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8. The complete event window determination module performs step S4 of the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8. The improved S-transform module performs step S5 of the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8. The feature calculation module performs step S6 of the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8. The result output module performs step S7 of the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8.

10. A device for evaluating voltage sag characteristics in a distribution network, characterized in that: It includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the distribution network voltage sag characteristic evaluation method as described in any one of claims 1 to 8 when the computer program is executed.

Images