A method and device for detecting and alarming low-frequency oscillations in a power system

Abstract

The present application provides a method and an alarm device for detecting and alarming low-frequency oscillations in a power system. The method includes: determining the power of a measurement point in the power system; performing multi-stage low-pass filtering and decimation on the power of the measurement point to obtain power data P m (t) with M frequency bands and different sampling rates; determining the time window length T m and the data window length L m for each frequency band; performing windowed FFT calculation based on the data window length, the time window length, and the power data every preset time N, and fitting to determine the corrected frequency and corrected amplitude of the main oscillation signal in the frequency band; determining the number of oscillation cycles c m of the main oscillation signal in this period based on the corrected frequency and corrected amplitude; accumulating the number of oscillation cycles c m of frequency band m to obtain the accumulated number of oscillation cycles C m of frequency band m; if the accumulated number of oscillation cycles C m of frequency band m is greater than or equal to the low-frequency oscillation cycle number threshold C set , perform low-frequency oscillation alarm. If the accumulated number of oscillation cycles C m is less than zero, perform a clearing process.

Description

Technical Field

[0001] This application relates to the technical field of power systems and their automation, and particularly to a method and device for detecting and alarming low-frequency oscillations in a power system. Background Art

[0002] Low-frequency oscillations in a power system pose a serious threat to the safe and stable operation of the power system, and timely and reliable detection and alarming are required.

[0003] According to the technical specifications of domestic synchronous phasor measurement devices, such as the "Technical Specifications for Power System Real-time Dynamic Monitoring System (Q-GDW 10131-2017)", synchronous phasor measurement devices should have the ability to detect low-frequency oscillations in the power system in real time and online. When the power fluctuation frequency at the monitoring point is within the frequency band of 0.1 - 2.5 Hz, and the power fluctuation amplitude exceeds the power threshold value P set , and the number of consecutive power fluctuation cycles exceeds the cycle threshold value N set , the synchronous phasor measurement device needs to give a low-frequency oscillation alarm flag in a timely manner.

[0004] Currently, the methods for detecting low-frequency oscillations in a power system are mainly divided into three categories.

[0005] (1) Time-domain method: Analyze the power waveform in the time domain, calculate the average value, maximum value, and minimum value, and calculate the oscillation period based on the time interval between the maximum value and the minimum value. This method is only effective for low-frequency oscillation waveforms with relatively simple frequency components (as shown in the appendix Figure 1 ), but is easily affected by high-order harmonics in the signal and misjudge the maximum and minimum points (as shown in the appendix Figure 2 ).

[0006] (2) Digital Fourier transform algorithm (DFT) or fast Fourier transform algorithm (FFT) with a fixed time window.

[0007] (3) Parameter identification methods, such as the Prony method or the empirical mode method. This method is greatly affected by the selection of the time window, the order of the signal, and noise. The iterative calculation method determines that its calculation amount is unstable, and various factors will lead to low stability and reliability of its analysis results.

[0008] The methods (2) and (3) fail to detect the oscillation cycle number, so it is difficult to meet the requirements of the synchronous phasor measurement technical specifications.

[0009] Since the span of the low-frequency oscillation period is very large (0.4 s - 10 s), it is not easy to select a suitable calculation time window for these three types of methods.

[0010] With the further development of the new power system in China, the degree of power electronics is getting higher and higher, the proportion of wind power, photovoltaic and other distributed fluctuating energy sources is increasing, and the components of the low-frequency oscillation signal in the power system will be more complex. Therefore, it is necessary to study a reliable and fast detection and warning method for the low-frequency oscillation of the power system with large-span oscillation period, multi-frequency components and dynamic random fluctuation characteristics. Summary of the Invention

[0011] An embodiment of the present application provides a method for detecting and warning low-frequency oscillation of a power system, including: determining the power of a measurement point in the power system; performing multi-stage low-pass filtering and decimation on the power of the measurement point to obtain power data P m (t) in M frequency bands with different sampling rates, where m is a natural number from 1 to M; determining the time window length T m and data window length L m of each frequency band; fitting and determining the corrected frequency F m,max and corrected amplitude P m,max of the main oscillation signal in the frequency band based on the data window length, time window length and windowed FFT calculation of the power data at every preset time N m,max ; determining the oscillation cycle number c m,max of the main oscillation signal in the current cycle based on the corrected frequency F m and corrected amplitude P m ; accumulating the oscillation cycle number c m of frequency band m to obtain the accumulated oscillation cycle number C m of frequency band m; if the accumulated oscillation cycle number C set of frequency band m is greater than or equal to the low-frequency oscillation cycle number threshold C m , perform low-frequency oscillation warning. If the accumulated oscillation cycle number C m of frequency band m is less than zero, perform a clearing process.

[0012] According to some embodiments, the performing multi-stage low-pass filtering and decimation on the power of the measurement point to obtain power data P m,smp (t) in M frequency bands with different sampling rates includes: performing multi-stage low-pass filtering on the power of the measurement point and uniformly dividing it into M frequency bands; performing frequency decimation on the frequencies of each frequency band, and the sampling rate F m,H is 2 to 10 times the cut-off frequency F

[0013] of the frequency band. m,H According to some embodiments, the upper cut-off frequencies F 1,H of the multiple frequency bands are close to a geometric sequence, F 2,H / F 2,H ≈F 3,H / F M-1,H / F M,H≈K, where K is an integer greater than 1, and the upper cut-off frequency F of the first frequency band m,H is not less than 2.5 Hz, the cut-off frequency of the last low-pass filter is not less than 0.1 Hz, and the lower cut-off frequency F of each frequency band m,L is the upper cut-off frequency F of the next frequency band m+1,H

[0014] According to some embodiments, the time window length T m is not less than twice the signal period corresponding to the lower cut-off frequency F of each frequency band m,L The data window length L m = T m ×F m,smp where F m,smp is the sampling rate of the signal after decimation.

[0015] According to some embodiments, the preset time N is a natural number, and its value range is 1 second to 10 seconds.

[0016] According to some embodiments, at every preset time N, based on the data window length, time window length, and windowed FFT calculation of the power data, the corrected frequency F m,max and corrected amplitude P m,max of the main oscillation signal in the frequency band are determined by fitting, including: in each frequency band, intercept the power data P m (t) according to the data window length L m ; subtract the DC component from the power data P m (t) and then perform windowed calculation; perform L m -order fast Fourier transform (FFT) on the windowed calculated signal to obtain the discrete spectrum of the frequency band; determine the main oscillation signal with the largest amplitude in the discrete spectrum of the frequency band; perform double-peak spectral line fitting calculation on the main oscillation signal with the largest amplitude and its adjacent second-largest oscillation signal to obtain the corrected frequency F m,max and corrected amplitude P m,max of the main oscillation signal in the frequency band.

[0017] According to some embodiments, the window function for windowed calculation includes a cosine window function, or a mixed convolution window function obtained by convolving a rectangular window and a cosine window. The spectral resolution dF m of the discrete spectrum = 1 / T m .

[0018] According to some embodiments, except for the order D1 of the low-pass filter in the first frequency band, the order D m of the low-pass filter and the order L m of the fast Fourier transform are consistent in each frequency band.

[0019] According to some embodiments, based on the corrected frequency Fm,max and the corrected amplitude P m,max Determine the number of oscillation cycles c of the main oscillation signal in this period m , including: if the corrected amplitude P m,max is greater than or equal to the low-frequency power threshold P set , then the number of oscillation cycles c of the main oscillation signal in this period m = N×F m,max ; if the corrected amplitude P m,max is less than the low-frequency power threshold P set , then the number of oscillation cycles c of the main oscillation signal in this period m = -0.5×N×F m,L , where F m,L is the lower cut-off frequency of each frequency band.

[0020] An embodiment of the present application further provides a power system low-frequency oscillation detection and warning device, including a power detection unit, a filtering and sampling unit, a parameter determination unit, an FFT and correction unit, an oscillation cycle calculation unit, an accumulation unit and a warning unit. The power detection unit is used to determine the power of the measurement point in the power system; the filtering and sampling unit is used to perform multi-stage low-pass filtering and sampling on the power of the measurement point to obtain power data P m (t) of M frequency bands and different sampling rates, where m is a natural number from 1 to M; the parameter determination unit is used to determine the time window length T m and the data window length L m of each frequency band; the FFT and correction unit is used to perform windowed FFT calculation based on the data window length, time window length and the power data every preset time N, and fit and determine the corrected frequency F m,max and the corrected amplitude P m,max of the main oscillation signal of the frequency band; the oscillation cycle calculation unit is used to determine the number of oscillation cycles c m,max of the main oscillation signal in this period based on the corrected frequency F m,max and the corrected amplitude P m ; the accumulation unit is used to accumulate the number of oscillation cycles c m of frequency band m to obtain the accumulated oscillation cycle number C m of frequency band m; the warning unit is used to perform low-frequency oscillation warning when the accumulated oscillation cycle number C m of frequency band m is greater than or equal to the low-frequency oscillation cycle number threshold C set , and perform a clearing process when the accumulated oscillation cycle number C m of frequency band m is less than zero.

[0021] The technical solution provided by the embodiments of this application, in view of the characteristic of the large period span of the low-frequency oscillation signal, decomposes the signal of the measurement point to be analyzed into multiple frequency bands through multi-stage low-pass filtering and decimation. Data with a corresponding length time window is selected in each frequency band for windowing calculation to eliminate the boundary effect. The windowed signal is subjected to fast Fourier transform and double-peak spectral line fitting calculation to obtain the corrected frequency and corrected amplitude of the accurate main oscillation signal. The main oscillation mode is searched in each frequency band, and the main station oscillation modes with oscillation amplitude exceeding the threshold value are accumulated by oscillation cycle number according to their frequency values. When the cumulative value of the oscillation cycle number in a certain frequency band reaches the threshold, a low-frequency oscillation alarm is given.

[0022] It can adapt to the large-span period of low-frequency oscillation, can obtain the oscillation amplitude, frequency and cycle number respectively in different frequency bands, and there is no problem of difficult time window selection. It is not affected by noise and multi-frequency components, the calculation amount is basically fixed, the analysis result is accurate and reliable, the oscillation cycle number is accumulated in each frequency band according to different time scales, and an oscillation alarm can be given in time when the low-frequency oscillation reaches the set value, which can meet the low-frequency oscillation detection and alarm requirements in the new power system scenario. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to these drawings.

[0024] Figure 1 It is a low-frequency oscillation power waveform (on-site measured waveform) of a power system with simple spectrum components.

[0025] Figure 2 It is a low-frequency oscillation power waveform (on-site measured waveform) of a power system with complex spectrum components.

[0026] Figure 3 It is the overall structure diagram of a low-frequency oscillation detection and alarm method for a power system in the embodiments of this application.

[0027] Figure 4 It is the schematic flow diagram of a low-frequency oscillation detection and alarm method for a power system in the embodiments of this application.

[0028] Figure 5 It is the schematic diagram of the frequency band division of the low-frequency oscillation signal using 2 frequency bands in the embodiments of this application.

[0029] Figure 6 It is the schematic diagram of the frequency band division of the low-frequency oscillation signal using 3 frequency bands in the embodiments of this application.

[0030] Figure 7 For Figure 5 The waveform and calculation analysis results of the medium frequency band 1 in a 2 - frequency band scheme of an embodiment.

[0031] Figure 8 For Figure 5 The waveform and calculation analysis results of the medium frequency band 2 in a 2 - frequency band scheme of an embodiment.

[0032] Figure 9 A power system low - frequency oscillation detection and warning device according to an embodiment of the present application. Specific implementation manners

[0033] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present application.

[0034] It should be understood that the terms "including" and "comprising" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0035] Figure 3 A general structure diagram of a power system low - frequency oscillation detection and warning method according to an embodiment of the present application.

[0036] As Figure 3 shown, in the present application, the power of the point to be measured is subjected to multi - stage low - pass filtering and decimation to obtain power data of multiple frequency bands with uniform distribution. In each frequency band, data with an appropriate time window length is used for windowed FFT calculation and double - peak spectrum fitting, so as to obtain the corrected accurate amplitude and frequency of the main oscillation signal of the dominant oscillation mode in that frequency band.

[0037] For the dominant oscillation mode with an ultra - high threshold value, the number of oscillations is calculated by accumulating the oscillation cycle number and compared with the oscillation cycle number threshold. Each frequency band is calculated independently, and low - frequency oscillation warnings can be given separately. The high - frequency band uses a short - time window for calculation, and the low - frequency band uses a long - time window for calculation, so as to improve the detection accuracy and real - time performance in each frequency band.

[0038] Figure 4 A flow schematic diagram of a power system low - frequency oscillation detection and warning method according to an embodiment of the present application, including the following processes.

[0039] In S10, determine the power of the point to be measured in the power system.

[0040] At the low-frequency oscillation signal detection point of the power system, the power of the signal at the point to be measured is calculated based on the phasor values of the three-phase voltage and current, or the power of the signal at the point to be measured is calculated based on the instantaneous values of the three-phase voltage and current.

[0041] When calculating the power based on the instantaneous values of the three-phase voltage and current, the calculation formula is:

[0042] P m (t) = u a (t) × i a (t) + u b (t) × i b (t) + u c (t) × i c (t)

[0043] In the formula, u j (t), i j (t) are the instantaneous values of the signal, and j = a, b, c.

[0044] In the synchronous phasor measurement device, the power curve has generally been calculated based on the phasor values of the three-phase voltage and current. Its output frequency is 100 Hz, and its calculation formula is:

[0045]

[0046] In the formula, U j (t), I j (t) are the phasor amplitudes of the signal, is the phase angle difference between the voltage and the current, and j = a, b, c.

[0047] In S20, the power at the point to be measured is subjected to multi-stage low-pass filtering and decimation to obtain power data P m (t), where m is a natural number from 1 to M.

[0048] The order of the low-pass filter corresponding to each frequency band is D m , and the sampling rate of the signal after decimation is F m，smp .

[0049] In S21, the power at the point to be measured is subjected to multi-stage low-pass filtering and evenly divided into M frequency bands.

[0050] The method of frequency band division is: the upper cut-off frequency F 1，H of the first frequency band is taken as 2.5 Hz or slightly higher, and the cut-off frequency of the last-stage low-pass filter is not less than 0.1 Hz.

[0051] Each frequency band is evenly divided according to the number of frequency band levels M, so that the upper limit frequencies F m，H of each level are close to a geometric progression, that is: F 1,H / F2,H ≈F 2,H / F 3,H ≈...F M-1,H / F M,H ≈K, where K is an integer greater than 1, which facilitates the decimation operation. The lower cut-off frequency F of each frequency band m,L is the upper cut-off frequency F of the next frequency band m+1,H .

[0052] In S22, frequency decimation is performed on the frequencies of each frequency band, and the sampling rate F m,smp is 2 to 10 times the cut-off frequency F of the frequency band, to ensure signal sampling accuracy and save computational effort, but not limited thereto. m,H

[0053] As Figure 5 , Figure 6 shown, examples of dividing the signal into 2 or 3 signal frequency bands are given respectively. During specific implementation, the appropriate number of frequency band divisions can be comprehensively weighed according to the computing power of the device and the measurement accuracy requirements.

[0054] As Figure 5 shown in the embodiment, the signal is divided into 2 frequency bands, and the common ratio coefficient of each level of frequency is K = 5. The power data comes from the synchronized phasor measurement module, whose original data rate is 100Hz, and the signal frequency range is 0 - 5Hz. Since the power fluctuation frequency of the power system low-frequency oscillation is within the 0.1 - 2.5Hz frequency band, the original 100Hz signal is filtered and decimated in 2 levels, and the 0.1 - 2.5Hz is divided into 2 frequency bands. After the frequency band division, the frequency range of frequency band 1 is 0.5 - 2.5Hz, and the frequency range of frequency band 2 is 0.1 - 0.5Hz.

[0055] The curves after each order of low-pass filtering are decimated at a ratio of 5:1, and the sampling rate after decimation is 1 / 5 of the original sampling rate. The sampling rate after decimation of frequency band 1 is 20Hz, which is 8 times the upper limit frequency 2.5Hz of frequency band 1. The sampling rate after decimation of frequency 2 is 4Hz, which is 8 times the upper limit frequency 0.5Hz of frequency band 2. The sampling rate being 8 times the upper limit frequency of the signal can ensure a high analysis accuracy.

[0056] As Figure 6 shown in the embodiment, the signal is divided into 3 frequency bands, and the common ratio coefficient of each level of frequency is K = 3. The power data comes from the synchronized phasor measurement module, whose original data rate is 100Hz, and the signal frequency range is 0 - 5Hz. After filtering and decimation of this signal to achieve frequency band division, the frequency range of frequency band 1 is 0.9 - 2.5Hz, the frequency range of frequency band 2 is 0.3 - 0.9Hz, and the frequency range of frequency band 3 is 0.1 - 0.3Hz.

[0057] ​The curves after low-pass filtering of each order are decimated at a ratio of 3:1, and the sampling rate after decimation is 1 / 3 of the original sampling rate. The sampling rate after decimation in frequency band 1 is 20 Hz, which is 8 times the upper limit frequency of 2.5 Hz in frequency band 1. The sampling rate after decimation in frequency band 2 is 20 / 3 Hz, which is 7.4 times the upper limit frequency of 0.9 Hz in frequency band 2. The frequency after decimation in frequency band 3 is 20 / 9 Hz, which is 7.4 times the upper limit frequency of 0.3 Hz in frequency band 3. A sampling rate that is 7.4 times the upper limit frequency of the signal can also ensure a relatively high analysis accuracy.

[0058] In S30, determine the time window length T for each frequency band m and the data window length L m .

[0059] Specifically, the appropriate time window length T m is not less than 2 times the signal period corresponding to the lower cut-off frequency F of each frequency band m,L , and can take 3 to 10 times to achieve effective coverage of the signal frequency and improve the calculation accuracy. The data window length L m = T m × F m,smp , where F m,smp is the sampling rate of the signal after decimation.

[0060] It should be noted that the calculation accuracy and the response speed restrict each other, and a trade-off needs to be made to select an appropriate calculation time window length L m . L m Preferably, an integer power of 2 is selected, or other values are selected according to the comprehensive requirements of the calculation accuracy and the response speed.

[0061] In S40, a sliding time window is adopted, and based on the data window length, the time window length, and the power data windowed FFT calculation, the corrected frequency F m,max and the corrected amplitude P m,max of the main oscillation signal in the frequency band are determined by fitting. Among them, N is a natural number, and the value range of N is 1 second to 10 seconds. A typical N can be taken as 1 second.

[0062] In S41, in each frequency band, intercept the power data P m at the data window length L m (t).

[0063] In S42, the power data P m (t) is windowed after subtracting the DC component.

[0064] According to some embodiments, the window function for windowed calculation includes, but is not limited to, a cosine window function, or a mixed convolution window function obtained by convolving a rectangular window and a cosine window, to reduce the truncation effect and suppress spectral leakage, which will not be elaborated here.

[0065] In some embodiments, the window function selects a Hanning window, and the windowing effect of the signal is as shown in the appended Figure 7 and the appended Figure 8 figures.

[0066] In S43, an L m -order fast Fourier transform (FFT) is performed on the windowed signal to obtain the discrete spectrum of the frequency band, and the spectral resolution dF m of the discrete spectrum is 1 / T m .

[0067] Specifically, the order D1 of the low-pass filter in the first frequency band is affected by the sampling rate F0 of the original power signal. Each of the other frequency bands is evenly divided according to a geometric progression, and the order D m of the low-pass filter and the order L m of the fast Fourier transform remain consistent in each frequency band, so that the low-pass filtering calculation module and the FFT calculation module can be reused in each frequency band.

[0068] As Figure 5 shown, the data time windows T m corresponding to frequency band 1 and frequency band 2 are 12.8 seconds and 64 seconds respectively, and their corresponding frequency resolutions dF m are 0.078 Hz and 0.016 Hz respectively.

[0069] As Figure 5 shown, the number of FFT calculation points for both frequency band 1 and frequency band 2 is 256 points (2 to the 8th power). The corresponding time window lengths for frequency band 1 and frequency band 2 are 12.8 seconds and 64 seconds respectively, and the periods of the lowest frequency points in frequency band 1 and frequency band 2 are 2 seconds and 10 seconds respectively. Therefore, the data window covers more than 6 cycles of the lowest frequency signal, which is beneficial to improving the calculation accuracy.

[0070] As Figure 6 shown, the time window lengths T m corresponding to frequency band 1, frequency band 2, and frequency band 3 are 6.4 seconds, 19.2 seconds, and 57.6 seconds respectively, and their corresponding frequency resolutions dF m are 0.167 Hz, 0.053 Hz, and 0.017 Hz respectively.

[0071] As Figure 6 shown, the number of FFT calculation points for frequency band 1, frequency band 2, and frequency band 3 is 128 points (2 to the 7th power). The corresponding time window lengths for frequency band 1, frequency band 2, and frequency band 3 are 6.4 seconds, 19.2 seconds, and 57.6 seconds respectively, and the periods of the lowest frequency points in frequency band 1, frequency band 2, and frequency band 3 are 1.1 seconds, 3.3 seconds, and 10 seconds respectively. Therefore, the data window covers approximately 6 cycles of the lowest frequency signal, which is beneficial to improving the calculation accuracy.

[0072] The waveform and spectrum analysis results of dividing the frequency band into two frequency bands are as follows Figure 7 and Figure 8 shown

[0073] As an example of L m not being an integer power of 2, in Figure 7 , the number of calculation points selected for frequency band 1 is 200 points. At this time, its time window length is 10 seconds, and its corresponding frequency resolution dF m is 0.1 Hz

[0074] In Figure 8 , the number of calculation points selected for frequency band 2 is 152 points. At this time, its time window length is 38 seconds, and its corresponding frequency resolution dF m is 0.0263 Hz

[0075] In S44, the main oscillation signal with the largest amplitude in the discrete spectrum of the frequency band is determined

[0076] As Figure 7 shown, the filtered waveform, windowed waveform, and spectrum analysis results of frequency band 1 are respectively shown. a1 is the filtered waveform of frequency band 1, a2 is the windowed waveform of frequency band 1, a3 is the spectrum analysis result of frequency band 1, and f1 is the frequency range of interest for frequency band 1. The frequency range of frequency band 1 is 0.5 - 2.5 Hz, and no oscillation exceeding the threshold (taking 5 MW as an example) is found in the frequency analysis result of the FFT within this frequency band

[0077] As Figure 8 shown, the filtered waveform, windowed waveform, and spectrum analysis results of frequency band 2 are respectively shown. b1 is the filtered waveform of frequency band 2, b2 is the windowed waveform of frequency band 2, b3 is the spectrum analysis result of frequency band 2, and f2 is the frequency range of interest for frequency band 2. The frequency range of frequency band 2 is 0.1 - 0.5 Hz, and an oscillation mode with a frequency of 0.157 Hz is found in the frequency analysis result of the FFT within this frequency band, and its amplitude exceeds the threshold (5 MW)

[0078] In S45, double - peak spectral line fitting calculation is performed on the main oscillation signal with the largest amplitude and its adjacent second - largest oscillation signal to obtain the corrected frequency F m,max and corrected amplitude P m,max of the accurate main oscillation signal of the frequency band

[0079] As Figure 8 shown, double - peak spectral line fitting calculation is performed on the main oscillation signal with the largest amplitude of 0.16 Hz and its adjacent second - largest oscillation signal of 0.183 Hz, and its more accurate corrected frequency is obtained as 0.163 Hz. Double - peak spectral line fitting uses the amplitude and spectrum information of the largest main oscillation signal and the second - largest oscillation signal for frequency and amplitude correction. This method is an existing technology and will not be elaborated here

[0080] The technical solution provided by the embodiments of this application, in view of the characteristic of the large period span of the low-frequency oscillation signal, performs multi-band decomposition on the analyzed signal through multi-stage low-pass filtering and decimation. Data with a corresponding length time window is selected in each frequency band for windowing calculation to eliminate the boundary effect. The windowed signal is subjected to fast Fourier transform and double-peak spectral line fitting calculation to obtain the corrected frequency and corrected amplitude of the precise main oscillation signal.

[0081] In S50, based on the corrected frequency F m,max and the corrected amplitude P m,max determine the number of oscillation cycles c of the main oscillation signal within this period m .

[0082] If the corrected amplitude P m,max is greater than or equal to the low-frequency power threshold P set , then the number of oscillation cycles c of the main oscillation signal within this period m = N×F m,max .

[0083] If the corrected amplitude P m,max is less than the low-frequency power threshold P set , then the number of oscillation cycles c of the main oscillation signal within this period m = -0.5×N×F m,L , where F m,L is the lower cut-off frequency of each frequency band.

[0084] As Figure 8 shown, a main oscillation signal of 0.163 Hz is detected in frequency band 2, and its corrected amplitude is greater than the low-frequency power threshold P set 5 MW, then the number of oscillation cycles c within this period m = 0.163 Hz×1 s = 0.163 times.

[0085] In S60, every preset time N, the number of oscillation cycles c of frequency band m m is accumulated to obtain the accumulated number of oscillation cycles C of frequency band m m .

[0086] In S70, if the accumulated number of oscillation cycles C of frequency band m m is greater than or equal to the low-frequency oscillation cycle number threshold C set , a low-frequency oscillation alarm is given. If the accumulated number of oscillation cycles C of the frequency band m m is less than zero, a clearing process is performed.

[0087] As Figure 8 shown, if the low-frequency oscillation of 0.163 Hz persists and the low-frequency oscillation cycle number threshold is 10 times, then after 62 seconds of accumulation, the accumulated number of oscillation cycles C mReaching 0.163×62 = 10.1, at this time, low-frequency oscillation alarm can be given in frequency band 2.

[0088] The technical solution provided by the embodiments of this application searches for the dominant oscillation mode in each frequency band, and accumulates the oscillation cycle numbers of the main station oscillation mode whose oscillation amplitude exceeds the threshold value according to its frequency value. When the accumulated value of the oscillation cycle numbers in a certain frequency band reaches the threshold, a low-frequency oscillation alarm is given. It can adapt to the large-span period of low-frequency oscillation, and can obtain the oscillation amplitude, frequency and cycle number respectively in different frequency bands. There is no problem of difficult time window selection, and it is not affected by noise and multi-frequency components. The calculation amount is basically fixed, and the analysis result is accurate and reliable. Each frequency band accumulates the oscillation cycle numbers according to different time scales, and can give an oscillation alarm in time when the low-frequency oscillation reaches the set value, which can meet the low-frequency oscillation detection and alarm requirements in the new power system scenario.

[0089] Figure 9 A low-frequency oscillation detection and alarm device for the power system according to the embodiments of this application includes a power detection unit 1, a filtering and sampling unit 2, a parameter determination unit 3, an FFT and correction unit 4, an oscillation cycle calculation unit 5, an accumulation unit 6 and an alarm unit 7.

[0090] The power detection unit 1 is used to determine the power of the measurement point in the power system. The sampling unit 2 is used to perform M-level low-pass filtering and sampling on the power of the measurement point to obtain power data P m (t) at multiple frequency bands and different sampling rates, where m is a natural number from 1 to M. The parameter determination unit 3 is used to determine the time window length T m and the data window length L m of each frequency band. The FFT and correction unit 4 is used to perform windowed FFT calculation based on the data window length, time window length and power data every preset time N, and fit and determine the corrected frequency F m,max and the corrected amplitude P m,max of the main oscillation signal in the frequency band. The oscillation cycle calculation unit 5 is used to determine the oscillation cycle number c m of the signal in the current cycle in the frequency band based on the corrected amplitude. The accumulation unit 6 is used to accumulate the N oscillation cycle numbers c m in the frequency band m to obtain the accumulated oscillation cycle number C m of the frequency band m. The alarm unit 7 is used to give a low-frequency oscillation alarm when the accumulated oscillation cycle number C m in the frequency band m is greater than or equal to the low-frequency oscillation cycle number threshold C set , and perform a clearing process when the accumulated oscillation cycle number C m in the frequency band m is less than zero.

[0091] The above embodiments are only to illustrate the technical idea of the present application, and the protection scope of the present application cannot be limited thereby. Any modification made on the basis of the technical solution in accordance with the technical idea proposed in the present application shall fall within the protection scope of the present application.

Claims

1. A method for detecting and alarming low-frequency oscillations in a power system, comprising: Determining the power of a measurement point in the power system; Perform multi-level low-pass filtering and decimation on the power of the point to be measured, and obtain power data P m (t) with M frequency bands and different sampling rates, where m is a natural number from 1 to M; Determine the time window length T for each of the said frequency bands m and the data window length L m ; Perform windowed FFT calculation based on the data window length, time window length, and the power data every preset time N, and fit to determine the corrected frequency F of the main oscillation signal in the frequency band m,max and the corrected amplitude P m,max ; Based on the corrected frequency F m,max and the corrected amplitude P m,max determine the number of oscillation cycles c of the main oscillation signal in the frequency band within this period m ; Accumulate the oscillation frequency c of frequency band m m to obtain the accumulated oscillation frequency C of frequency band m m ; If the cumulative oscillation frequency C of frequency band m m is greater than or equal to the low-frequency oscillation frequency threshold C set a low-frequency oscillation alarm is given. If the cumulative oscillation frequency C of the frequency band m m is less than zero, a clearing process is performed.

2. The method according to claim 1, wherein Performing multi-stage low-pass filtering and decimation on the power of the point to be measured to obtain power data P m (t) with M frequency bands and different sampling rates, including: Perform multi-stage low-pass filtering on the power of the point to be measured and divide it into M frequency bands such that the upper cut-off frequency F of each frequency band m,H is close to a geometric progression; Frequency sampling is performed on the frequencies of each frequency band, and the sampling rate is F m,smp is 2 to 10 times the upper cut-off frequency F m,H times.

3. The method according to claim 2, wherein, The upper cut-off frequency F of the M frequency bands m,H is close to a geometric progression, and F 1,H / F 2,H ≈F 2,H / F 3,H ≈...F M-1,H / F M,H ≈K, where K is an integer greater than 1. The upper cut-off frequency F of the first frequency band m,H is not lower than 2.5 Hz, and the cut-off frequency of the last low-pass filter is not less than 0.1 Hz. The lower cut-off frequency F of each frequency band m,L is the upper cut-off frequency F of the next frequency band m+1,H .

4. The method according to claim 1, wherein The time window length T m is not less than twice the signal period corresponding to the lower cut-off frequency F m,L of each frequency band, and the data window length L m = T m × F m,smp , where F m,smp is the sampling rate of the signal after decimation.

5. The method according to claim 1, wherein, The preset time N is a natural number, and its value range is from 1 second to 10 seconds.

6. The method according to claim 1, wherein, Performing windowed FFT calculation based on the data window length, time window length, and the power data every preset time N, and fitting to determine the corrected frequency F of the main oscillation signal in the frequency band m,max and the corrected amplitude P m,max , including: At each frequency band, according to the data window length L m Intercept the power data P m (t); Subtract the DC component from the power data P m (t) and then perform windowing calculation; Perform an L m -order fast Fourier transform (FFT) on the windowed signal to obtain the discrete spectrum of the frequency band; Determining the main oscillation signal with the largest amplitude in the discrete spectrum of the frequency band; Perform a double-peak spectral line fitting calculation on the main oscillation signal with the largest amplitude and its adjacent second-largest oscillation signal to obtain the corrected frequency F of the main oscillation signal in the frequency band m,max and the corrected amplitude P m,max .

7. The method according to claim 6, wherein The window function for windowing calculation includes a cosine window function, or a mixed convolution window function obtained by convolving a rectangular window and a cosine window. The spectral resolution dF of the discrete spectrum m = 1 / T m .

8. The method according to claim 3, wherein, Except for the order D1 of the low-pass filter in the first frequency band, the order D of the low-pass filter m and the order L of the fast Fourier transform m remain consistent in each frequency band.

9. The method according to claim 1, wherein Based on the corrected frequency F m,max and the corrected amplitude P m,max determine the number of oscillation cycles c of the main oscillation signal in this frequency band within this period m , including: If the correction amplitude P m,max is greater than or equal to the low-frequency power threshold P set , then the oscillation cycle number c of the main oscillation signal in this period m = N × F m,max ; If the correction amplitude P m,max is less than the low-frequency power threshold P set , then the oscillation cycle number c of the main oscillation signal in this period m = -0.5×N×F m,L， where F m,L is the lower cut-off frequency of each frequency band.

10. A device for detecting and alarming low-frequency oscillations in a power system, comprising: A power detection unit for determining the power of a measurement point in the power system; Filter decimation unit, which is used to perform multi-stage low-pass filtering and decimation on the power of the point to be measured, so as to obtain power data P m (t) in M frequency bands with different sampling rates, where m is a natural number from 1 to M; A parameter determination unit for determining the time window length T of each of the frequency bands m and the data window length L m ; An FFT and correction unit is used to perform windowed FFT calculation based on the data window length, time window length, and the power data every preset time N, and fit and determine the corrected frequency F of the main oscillation signal in the frequency band m,max and the corrected amplitude P m,max ; An oscillation cycle calculation unit for determining the number of oscillation cycles c of the main oscillation signal in the present cycle based on the corrected frequency F m,max and the corrected amplitude P m,max ; m ; An accumulation unit for accumulating the oscillation frequency c of frequency band m m to obtain the accumulated oscillation frequency C of frequency band m m ; An alarm unit, which is used for the cumulative oscillation cycle number C in frequency band m m being greater than or equal to the low-frequency oscillation cycle number threshold C set to perform a low-frequency oscillation alarm. When the cumulative oscillation cycle number C in the frequency band m m is less than zero, a clearing process is performed.

Images