A method for accurate and rapid extraction of power quality disturbance characteristics
Through the optimal Bohman window improvement S conversion method, the problem of insufficient time-frequency resolution of power quality disturbance feature extraction in the power system is solved, and high-precision and fast power quality disturbance signal feature extraction is achieved, meeting the detection requirements in the on-site environment .
Patent Information
- Application Number
- CN202210334763.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-31
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2042-03-31
AI Technical Summary
The existing time-frequency analysis methods are difficult to accurately and quickly extract the time-frequency characteristics of power quality disturbances in the power system, especially when the frequency resolution and the time resolution at high frequencies are insufficient, which cannot meet the accurate detection of the power quality disturbance characteristics. need.
The optimal Bohman window improves the S transformation method, and the dynamic measurement fast algorithm based on the maximum envelope is calculated to extract the time and frequency domain features, and the time frequency resolution is improved. Reduced calculation amount.
It realizes high-precision extraction of the power quality disturbance signal, with amplitude error close to 0, a fast calculation speed, and a single operation time is less than 25ms, meeting the requirements for fast extraction of the characteristics of the power quality disturbance signal.
Smart Images

Figure CN114638270B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of smart grids, and in particular to a method for accurately and quickly extracting power quality disturbance features. Background Art
[0002] With the continuous advancement of power grid construction, the increase of various nonlinear electrical equipment and the widespread use of thyristor rectifiers and voltage regulators have led to the appearance of a large number of harmonics and ultra-high harmonics in the power system; short-term voltage disturbances within 20ms caused by no-load excitation of large transformers and large load switching are issues worthy of attention. Power quality disturbances not only cause waste of electric energy, but also may cause major accidents when serious failures of power equipment occur. For this reason, some users have put forward stricter standards for power quality, which also puts higher requirements on the detection of power quality disturbances. Therefore, the accurate and rapid extraction of power quality disturbance characteristics will become an important basis for the effective management and evaluation of power quality issues.
[0003] The characteristics of power quality disturbances include time domain characteristics and frequency domain characteristics. Fourier transform is an important tool for spectrum analysis. It can successfully extract the frequency domain characteristics of the signal but cannot obtain the time domain characteristics. Therefore, it is only applicable to the analysis of stationary signals. In order to obtain the time and frequency domain characteristics of the signal at the same time, the following time and frequency analysis methods have gradually been developed: short time Fourier transform (STFT), wavelet transform (WT), S-transform (ST), generalized S-transform (GST), multiresolution generalized S-transform (MGST), double resolution S-transform (DRST), etc. Among them, STFT obtains time domain features by introducing a window function related to time and frequency. According to the Heisenberg uncertainty principle, a fixed window function cannot obtain ideal time resolution and frequency resolution at the same time. Therefore, STFT does not have the characteristics of variable time and frequency resolution and cannot meet the requirements for accurate extraction of power quality disturbance features. WT is widely used in the analysis of non-stationary signals. When using wavelet transform or its extended wavelet packet transform, different mother wavelets and decomposition levels need to be selected to detect different frequency components of the signal, which affects the applicability of wavelet transform in detecting real-time non-stationary signals. ST can be regarded as an improvement of STFT and wavelet transform, using Gaussian window to adjust time and frequency resolution. For the extraction of time and frequency features of disturbance signals, the change of fundamental frequency amplitude is very important for detecting voltage swell, sag, flicker and interruption. At the same time, it is also necessary to accurately detect the frequency of transient oscillation and harmonics at high frequency. That is, in most cases, disturbance detection requires better frequency resolution at high frequency and better time resolution at fundamental frequency.However, ST is contrary to actual needs. ST provides better time resolution at high frequencies and better frequency resolution at low frequencies. In order to improve this situation, the following improved STs have been proposed: Generalized S transform GST adds an adjustment factor on the basis of S transform, which improves the time-frequency resolution to a certain extent, but when the frequency span of the disturbance signal is large, it still cannot meet the requirements of time-frequency resolution at different frequency points; MGST introduces a window width scale adjustment factor and pre-analyzes the disturbance signal, divides the entire frequency band into several detection intervals, and achieves better time-frequency resolution, but there is an uneven transition between different frequency bands, which may affect feature extraction; DRST also introduces an adjustment factor to improve the window function, and the time-frequency resolution is further improved compared with MGST, but in essence their window width is still inversely proportional to the frequency, and has not fundamentally solved the shortcomings of poor fundamental frequency amplitude detection capability and decreased frequency resolution with increasing detection frequency. When encountering short-term voltage disturbances and rich high-frequency components, it is still impossible to complete the accurate extraction of features. Summary of the invention
[0004] The purpose of the present invention is to provide a method for accurately and quickly extracting power quality disturbance characteristics with fast extraction speed and high precision.
[0005] To achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a method for accurately and quickly extracting power quality disturbance features, the method comprising: transmitting the power quality disturbance signal to a host computer, the host computer performing an optimal Bohman window improved S transform on the received disturbance signal, obtaining a modulus time-frequency matrix by determining an optimal length parameter, and then performing time domain feature extraction and frequency domain feature extraction through the modulus time-frequency matrix.
[0006] The optimal Bohman window improved S transform specifically comprises the following steps:
[0007] (1) Calculate the discrete Fourier transform sequence of the disturbance signal x(nT):
[0008]
[0009] Where: k = 0, 1, ..., N-1; T is the sampling time interval; N is the total number of sampling points;
[0010] (2) X(k) is the discrete Fourier transform result of the disturbance signal. The maximum sequence formed by all the maximum points in |X(k)| is denoted as |X1(i)|, where i = 0, 1, 2, ..., P-1. Suppose there are H1, H2, ..., H between adjacent maximum points of |X(k)|. i …, H P-1 frequency points, the maximum envelope of |X(k)| is:
[0011]
[0012] Where j = 0, 1, 2, ..., H i-1 ; i = 1, 2, ..., P-1; P is the number of maximum points;
[0013] (3) A fast dynamic measurement algorithm based on the maximum envelope is used to reduce the amount of calculation, and the main characteristic frequency points m, m∈h are screened out by setting a threshold;
[0014] (4) Determine the window width adjustment factor r according to the frequency band where the characteristic frequency point m is located;
[0015] (5) X(m) is translated to obtain X(m+l), where l is the translation factor of the discrete Fourier transform translation, l = 0, 1, ..., N-1;
[0016] (6) Calculate the discrete Fourier spectrum of the Bohman window function. First, the expression of the Bohman window function is given as follows:
[0017]
[0018] Where: t is time;
[0019] The discrete Fourier spectrum of the optimal Bohman window function is obtained from formula (3):
[0020]
[0021] Where: n = 0, 1, ..., N-1;
[0022] (7) Calculate the optimal Bohman window function to improve the S-transform modulus time-frequency matrix FBST:
[0023]
[0024] Where: n=0,1,…,N-1.
[0025] Determining the optimal length parameter refers to setting the window length parameter L. In the low frequency band of the frequency range of 0 to 100 Hz, L=96 is taken; in the high frequency band of the frequency range of 100 to 1600 Hz, L=346 is taken; the time domain feature extraction refers to calculating the time domain amplitude vector |FBST(m, n)| to obtain the time domain feature, that is, the fundamental frequency amplitude curve; the frequency domain feature extraction refers to calculating the frequency domain amplitude vector Obtain frequency domain characteristics and frequency domain amplitude envelope curve.
[0026] It can be seen from the above technical scheme that the beneficial effects of the present invention are: first, the present invention has a higher recognition accuracy when extracting the characteristics of power quality disturbance signals. The results of building a power quality disturbance signal detection platform show that the amplitude error of the present invention is almost close to 0 when it is the lowest, which meets the requirements for accurately extracting the characteristics of power quality disturbance signals in the field environment; second, the present invention has a faster computing speed when extracting the characteristics of power quality disturbance signals, and the single computing time is less than 25ms, which meets the requirements for quickly extracting the characteristics of power quality disturbance signals. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 The figure is a flow chart of the method of the present invention.
[0028] Figure 2 This is a flow chart of the optimal Bohman window improved S transform algorithm in the present invention.
[0029] Figure 3 Schematic diagram of the user interface of the present invention. DETAILED DESCRIPTION
[0030] like Figure 1 As shown, a method for accurately and quickly extracting power quality disturbance features includes: transmitting the power quality disturbance signal to a host computer, the host computer performing an optimal Bohman window improved S transform on the received disturbance signal, obtaining a modulus time-frequency matrix by determining an optimal length parameter, and then performing time domain feature extraction and frequency domain feature extraction through the modulus time-frequency matrix.
[0031] The optimal Bohman window improved S transform specifically comprises the following steps:
[0032] (1) Calculate the discrete Fourier transform sequence of the disturbance signal x(nT):
[0033]
[0034] Where: k = 0, 1, ..., N-1; T is the sampling time interval; N is the total number of sampling points;
[0035] (2) X(k) is the discrete Fourier transform result of the disturbance signal. The maximum sequence formed by all the maximum points in |X(k)| is denoted as |X1(i)|, where i = 0, 1, 2, ..., P-1. Suppose there are H1, H2, ..., H between adjacent maximum points of |X(k)|. i …, H P-1 frequency points, the maximum envelope of |X(k)| is:
[0036]
[0037] Where j = 0, 1, 2, ..., H i-1; i = 1, 2, ..., P-1; P is the number of maximum points;
[0038] (3) A fast dynamic measurement algorithm based on the maximum envelope is used to reduce the amount of calculation, and the main characteristic frequency points m, m∈h are screened out by setting a threshold;
[0039] (4) Determine the window width adjustment factor r according to the frequency band where the characteristic frequency point m is located;
[0040] (5) X(m) is translated to obtain X(m+l), where l is the translation factor of the discrete Fourier transform translation, l = 0, 1, ..., N-1;
[0041] (6) Calculate the discrete Fourier spectrum of the Bohman window function. First, the expression of the Bohman window function is given as follows:
[0042]
[0043] Where: t is time;
[0044] The discrete Fourier spectrum of the optimal Bohman window function is obtained from formula (3):
[0045]
[0046] Where: n = 0, 1, ..., N-1;
[0047] (7) Calculate the optimal Bohman window function to improve the S-transform modulus time-frequency matrix FBST:
[0048]
[0049] Where: n=0,1,…,N-1.
[0050] Determining the optimal length parameter refers to setting the window length parameter L. In the low frequency band of the frequency range of 0 to 100 Hz, L=96 is taken; in the high frequency band of the frequency range of 100 to 1600 Hz, L=346 is taken; the time domain feature extraction refers to calculating the time domain amplitude vector |FBST(m, n)| to obtain the time domain feature, that is, the fundamental frequency amplitude curve; the frequency domain feature extraction refers to calculating the frequency domain amplitude vector Obtain frequency domain characteristics and frequency domain amplitude envelope curve.
[0051] The following combination Figures 1 to 3 The present invention is further described.
[0052] In order to verify the effectiveness of the present invention, MATLAB is used to simulate three time-domain disturbance signals of voltage interruption, temporary rise and temporary drop, and the corresponding parameters are set to control the duration of the disturbance. Then, different algorithms are used to extract the features of the disturbance signal. The relative amplitude errors under different feature extraction methods are shown in Table 1. It can be seen from Table 1 that the error of the present invention is significantly smaller than that of the other two methods under the same disturbance time. As the disturbance time becomes shorter, this advantage will become more obvious. Therefore, the feature extraction accuracy of the present invention is higher, which meets the requirements of the feature extraction accuracy of the power quality disturbance signal.
[0053] Table 1 Relative amplitude error under different disturbance time (10 -5 )
[0054]
[0055] In order to further verify the practicality of the present invention, the system is used to conduct actual test analysis on the time-frequency domain disturbance signals: voltage surge, sag, interruption, and harmonic signals. The running time of the present invention and other methods in processing 1600 disturbance signals is compared. The built Simulink circuit model is loaded into the real-time simulator through the StarSim software to simulate the input of the disturbance signal. The disturbance signal is sampled by the A / D converter in the data acquisition device, and the collected data is downsampled using the DSP processor. The data sampled and processed by the data acquisition device is transmitted to the host computer, and then the time-frequency feature is extracted by the algorithm. The relative amplitude error and operation time under different feature extraction methods are compared, as shown in Table 2. As can be seen from Table 2, the relative amplitude error and required operation time of the feature extraction of the present invention are both ahead of the other two algorithms, indicating that the present invention can also meet the requirements of fast and high-precision disturbance feature extraction of power quality in actual scenarios.
[0056] Table 2 Relative error and operation time of different disturbance signal amplitudes
[0057]
[0058] The real-time simulator generates a power quality disturbance signal; the conditioning circuit processes the collected signal to match it with the A / D converter input; the A / D converter converts the analog signal into a digital signal; the DSP processor performs downsampling processing on the digital signal to reduce the amount of data; the host computer is used to use an algorithm to extract features from the sampled data transmitted by the DSP processor to complete the feature extraction of the disturbance signal; the display synchronously displays the detection signal waveform and gives the feature extraction result.
[0059] like Figure 3 As shown, the user interface of the present invention is designed by LabVIEW, which can realize: waveform display of disturbance signal, display of fundamental frequency amplitude curve and its characteristic parameters, display of frequency amplitude envelope curve and its characteristic parameters.
[0060] In summary, the present invention has a high recognition accuracy when extracting the features of power quality disturbance signals. The results of building a power quality disturbance signal detection platform show that the detection amplitude error of the present invention is as low as almost 0, which meets the requirements for accurate feature extraction of power quality disturbance signals in field environments; the present invention has a fast computing speed when diagnosing power quality disturbances, and the single diagnosis time is less than 25ms, which meets the requirements for fast extraction of power quality disturbance signal features.
Claims
1. A method for accurately and quickly extracting power quality disturbance characteristics, characterized by: The method comprises: transmitting a power quality disturbance signal to a host computer, the host computer performing an optimal Bohman window improved S transform on the received disturbance signal, obtaining a modulus time-frequency matrix by determining an optimal length parameter, and then performing time domain feature extraction and frequency domain feature extraction through the modulus time-frequency matrix; The optimal Bohman window improved S transform specifically comprises the following steps: (1) Calculate the discrete Fourier transform sequence of the disturbance signal x(nT): Where: k = 0, 1, ..., N-1; T is the sampling time interval; N is the total number of sampling points; (2) X(k) is the discrete Fourier transform result of the disturbance signal. The maximum sequence formed by all the maximum points in |X(k)| is denoted as |X1(i)|, where i = 0, 1, 2, ..., P-1. Suppose there are H1, H2, ..., H between adjacent maximum points of |X(k)|. i …, H P-1 frequency points, the maximum envelope of |X(k)| is: Where j = 0, 1, 2, ..., H i-1; i=1,2,…,P-1; P is the number of maximum points; (3) A fast dynamic measurement algorithm based on the maximum envelope is used to reduce the amount of calculation, and the main characteristic frequency points m, m∈h are screened out by setting a threshold; (4) Determine the window width adjustment factor r according to the frequency band where the characteristic frequency point m is located; (5) X(m) is translated to obtain X(m+l), where l is the translation factor of the discrete Fourier transform translation, l = 0, 1, ..., N-1; (6) Calculate the discrete Fourier spectrum of the Bohman window function. First, the expression of the Bohman window function is given as follows: Where: t is time; The discrete Fourier spectrum of the optimal Bohman window function is obtained from formula (3): Where: n = 0, 1, ..., N-1; (7) Calculate the optimal Bohman window function to improve the S-transform modulus time-frequency matrix FBST: Where: n=0,1,…,N-1.
2. The method for accurately and quickly extracting power quality disturbance characteristics according to claim 1 is characterized in that: Determining the optimal length parameter refers to setting the window length parameter L. In the low frequency band of the frequency range of 0 to 100 Hz, L=96 is taken; in the high frequency band of the frequency range of 100 to 1600 Hz, L=346 is taken; the time domain feature extraction refers to calculating the time domain amplitude vector |FBST(m, n)| to obtain the time domain feature, that is, the fundamental frequency amplitude curve; the frequency domain feature extraction refers to calculating the frequency domain amplitude vector Obtain frequency domain characteristics and frequency domain amplitude envelope curve.
