Low-frequency oscillation early warning and warning method for power system
By using a method of full-frequency sub-band downsampling, filtering, and damping ratio analysis, the problems of misjudgment and alarm lag in low-frequency oscillation detection in power systems are solved, and faster and more accurate low-frequency oscillation alarms are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HOHAI UNIV
- Filing Date
- 2026-01-27
- Publication Date
- 2026-05-19
AI Technical Summary
Existing technologies are easily affected by disturbance signals in the detection of low-frequency oscillations in power systems, leading to misjudgments and delayed alarms. They cannot effectively shield disturbance signals, and unreasonable threshold settings also cause alarm delays.
The system employs full-frequency sub-band downsampling, filtering, early warning and alarm time window setting, damping ratio analysis, and fast Fourier transform. It determines whether to issue an alarm by calculating the damping ratio and comparing it with a set threshold, thereby suppressing spectral leakage and selecting the minimum damping ratio for accurate alarming.
It improves the accuracy and speed of low-frequency oscillation detection in power systems, reduces false alarms and alarm delays, and enhances the ability to shield against disturbance signals.
Smart Images
Figure CN122068459A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power system technology, specifically relating to a method for early warning and alarm of low-frequency oscillations in power systems. Background Technology
[0002] Low-frequency oscillations in power systems pose a serious threat to the safe and stable operation of power systems, and timely and reliable detection and alarm are necessary.
[0003] Application No. 2021115896498 discloses a method and device for detecting and alarming low-frequency oscillations in power systems. In this prior art, only simple cycle accumulation is performed on the data within a window, and then an alarm is issued by comparing the accumulated value with a threshold. This judgment method cannot effectively shield against interference from disturbance signals and is prone to misjudgment. Moreover, this cycle accumulation method for alarming can lead to excessively long cycle accumulation times due to unreasonable threshold settings (such as an excessively large cycle accumulation threshold), resulting in alarm lag. Summary of the Invention
[0004] To achieve the above objectives, the present invention proposes the following technical content: A method for early warning and alarm of low-frequency oscillation in power systems includes the following steps: S1: Monitor the active power sequence at the generator output. P (t); S2: Divide the entire frequency range of the low-frequency oscillation into sub-bands, and within the corresponding sub-bands, analyze the active power sequence. P (t) is downsampled; the sub-bands are: (0.1Hz~0.3Hz), (0.3Hz~0.9Hz) and (0.9Hz~2.5Hz); S3: Filter the downsampled sequence to obtain the... i Filtering sequences for each sub-band , i ∈[1,3] and is an integer; S4: For the first i For each frequency band, set the warning time window and the alarm time window; S5: For the first i Each sub-frequency band acquires multiple early warning data windows and determines whether to issue an early warning; S6: Calculate the total number of oscillation cycles for each alarm window; start Prony analysis, calculate all damping ratios in one alarm time window after the alarm is triggered and select the smallest damping ratio, then compare the smallest damping ratio with the set threshold, and determine whether to issue an alarm based on the comparison result and the total number of oscillation cycles for the corresponding alarm window.
[0005] Further, step S6 includes the following steps: S6.1: For the firsti Filtering sequences for each sub-band , in time t As the cutoff time for the alarm data window, data is truncated forward according to the alarm time window length set in S4 to obtain... t Real-time alarm data window; S6.2: As time progresses, at a set time, the subsequent... i Filtering sequences for each sub-band , K Represents positive integers. d Indicates the set alarm time interval; in time ( t + Kd The alarm data window is cut off at the end time of the alarm data window. The data is extracted forward from the alarm time window length set in S4 to obtain the alarm data window at the corresponding time. All alarm data windows are numbered in chronological order. S6.3: For all alarm data windows, a Hanning window is added to suppress spectral leakage, followed by a Fast Fourier Transform; for the first... z For each alarm data window, extract the frequency corresponding to the maximum amplitude within that alarm data window. f peak,z Then obtain the maximum amplitude value. A t,z ; S6.4: Calculate the increment of the oscillation frequency for the z-th alarm data window. C m,z The formula is:
[0006] N alarm This indicates a long alarm time window; f L,i Indicates the first i The lower limit frequency of each sub-band; S6.5: Calculate the total number of oscillation cycles in the z-th alarm data window, using the following formula:
[0007] In the formula, C z-1 This represents the number of oscillation cycles in the (z-1)th alarm data window; due to the penalty mechanism in the formula, if no oscillation occurs within a certain period of time, then... C z-1 It is possible for it to become negative, when C z-1 When the value is less than 0, it is forced to be 0; S6.6: Calculate all damping ratios within one warning time window after the warning is triggered; for the i-th sub-band, obtain the data of one warning time window after the warning, and set this window as the judgment window; decompose the judgment window first using linear constant coefficient difference, then construct the characteristic polynomial and find the poles in the Z-plane, deduce the pole formula in the time domain based on the poles in the Z-plane, and finally calculate the damping ratio of all poles based on the pole formula in the time domain, and select the damping ratio with the smallest damping ratio; S6.7: Compare the minimum damping ratio in S6.6 with the set threshold, and issue an alarm based on the comparison result; when When the value is less than 0, an alarm signal is immediately issued based on the comparison result, and the total number of oscillation cycles is reset to 0. when When the value is greater than or equal to 0 and less than 0.1, if the total number of oscillation cycles is greater than or equal to the set threshold, an alarm signal will be issued immediately based on the judgment result; if the total number of oscillation cycles is less than the set threshold, the total number of oscillation cycles will continue to be accumulated until it is greater than or equal to the set threshold. when If the value is greater than or equal to 0.1, it is determined to be a single disturbance signal, the total number of oscillation cycles is reset to 0, and no alarm is issued, but a warning of misjudgment is output.
[0008] Furthermore, step S2 includes the following steps: S2.1: Set the target sampling rates for each sub-band (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) to 2.5Hz, 8Hz, and 20Hz, respectively; calculate the... i The downsampling factor for each sub-band is calculated using the following formula:
[0009] In the formula, A i Indicates the first i Downsampling factor for each sub-band; f raw Represents active power sequence P The sampling rate of (t); f smp,i Indicates the first i The target sampling rate is set for each sub-band; Based on the i The downsampling factor of each sub-band affects the active power sequence. P (t) is used to extract the corresponding downsampling sequence. P i (t).
[0010] Further, step S5 includes the following steps: S5.1: For the first i Filtering sequences for each sub-band , in time t As the cutoff time for the early warning data window, data is truncated forward according to the early warning time window length set in S4 to obtain... t A real-time early warning data window; S5.2: As time progresses, at a set time ( t + Km ), to obtain the first i Filtering sequences for each sub-band , K Represents positive integers. m Indicates the set warning time interval; in time ( t + Km The warning data window is used as the cutoff time. Data is extracted forward from the warning time window length set in S4 to obtain the warning data window at the corresponding time. All warning data windows are numbered in chronological order. S5.3: Add a Hanning window to all warning data windows to suppress spectral leakage, then perform a fast Fourier transform to extract the frequency domain peak amplitude of each warning data window, and determine whether to issue a warning according to the set rules; The rules are as follows:
[0011] In the formula, t k Indicates the first k The deadline for each warning data window; t last_alarm Indicates the time of the last alarm; T cool Indicates the preset cooldown time; A peak ( k ) indicates the first k The peak amplitude of each warning data window; A peak _ set This indicates the set peak amplitude. A peak ( k- 2) indicates the first k- Peak amplitude of the two warning data windows; A peak ( k- 1) indicates the first k- The peak amplitude of one warning data window.
[0012] When the above rules are met, the first... k The deadline for each warning data window is the warning time.
[0013] Furthermore, the formula for the warning window length is:
[0014] In the formula, T pre,i Indicates for the first i The warning window length is set for each sub-frequency band; K 1 represents the periodicity coefficient, which is set to 2~3; The formula for alarm window length is:
[0015] In the formula, T pre,i Indicates for the first i Alarm window length set for each sub-frequency band; K 2 represents the periodic coefficient, which is greater than or equal to 6.
[0016] Furthermore, the warning time window lengths for the three frequency bands (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) are set to 25.6s, 8s, and 3.2s, respectively. The alarm time window lengths for the three frequency bands (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) are set to 102.4s, 32s, and 12.8s, respectively.
[0017] Furthermore, d Greater than m .
[0018] The beneficial effects that can be achieved by adopting the above-mentioned technology are: In this solution, a time window of data after the warning is triggered is obtained, and the damping ratio is obtained based on the time window data. Then, the damping ratio is compared with a set threshold, and the result of the comparison determines whether to trigger an alarm or block a single disturbance signal. Compared with the alarm methods in the background technology, the speed and accuracy of the alarm are improved. Attached Figure Description
[0019] Figure 1 This is the existing technology in the background section, specifically the alarm time diagram for frequency band 1; Figure 2 This is the alarm time diagram for frequency band 1 after incorporating the damping ratio in this solution; Figure 3 This is an existing technology in the background section, specifically an alarm time diagram for frequency band 2; Figure 4 This is the alarm time diagram for frequency band 2 after incorporating the damping ratio in this solution; Figure 5 This is an existing technology in the background section, specifically an alarm time diagram for frequency band 3; Figure 6 This is the alarm time diagram for frequency band 3 after incorporating damping ratio into this solution. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] A method for early warning and alarm of low-frequency oscillation in power systems includes the following steps: S1: Use a PMU (Phasor Measurement Unit) to monitor the active power sequence at the generator outlet. P (t); S2: Based on the set rules, the entire frequency range of the low-frequency oscillation is divided into sub-bands; based on the target sampling rate of each sub-band, the active power sequence is... P (t) is used to extract the downsampled sequence for each sub-band, resulting in the following steps: S2.1: Divide into sub-frequency bands.
[0022] To cover the full frequency range (0.1Hz~2.5Hz) of low-frequency oscillations while also considering computational efficiency, the frequency bands are divided into three sub-bands using a geometric progression: Low frequency band: (0.1Hz~0.3Hz); Mid-frequency band: (0.3Hz~0.9Hz); High frequency band: (0.9Hz~2.5Hz).
[0023] S2.2: Set the target sampling rate for each sub-band.
[0024] The target sampling rate is set to 2.5Hz for the (0.1Hz~0.3Hz) band, 8Hz for the (0.3Hz~0.9Hz) band, and 20Hz for the (0.9Hz~2.5Hz) band.
[0025] S2.3: For the first i Each frequency band, i Given integers ∈ [1,3], calculate the downsampling factor for each sub-band, and apply the downsampling factor to the active power sequence. P (t) is used to extract the downsampled sequence.
[0026] No.i The formula for the downsampling factor of each sub-band is:
[0027] In the formula, A i Indicates the first i Downsampling factor for each sub-band; f raw Represents active power sequence P The sampling rate of (t); f smp,i Indicates the first i The target sampling rate is set for each sub-band.
[0028] Based on the i The downsampling factor of each sub-band affects the active power sequence. P (t) is used to extract the corresponding downsampling sequence. P i (t).
[0029] S3: Using a Butterworth bandpass filter, for the first... i downsampling sequence of each sub-band P i (t) is filtered to obtain the first (t) i Filtering sequences for each sub-band The passband range of the Butterworth bandpass filter is set as follows: [ f L,i, f H,i In the formula, f L,i Indicates the first i The lower limit frequency of each sub-band, such as the lower limit frequency of the first sub-band (0.1Hz~0.3Hz) is 0.1Hz; f H,i Indicates the first i The upper limit frequency of each sub-band, such as the upper limit frequency of the first sub-band (0.1Hz~0.3Hz) is 0.3Hz.
[0030] S4: For the first i For each sub-frequency band, set the warning time window and alarm time window. Specifically, this includes the following steps: S4.1: Set the warning window length.
[0031] The formula is:
[0032] In the formula, T pre,i Indicates for the first i The warning window length is set for each sub-frequency band; K1 represents the periodicity coefficient, which is set to 2~3; S4.2: Set alarm window length.
[0033]
[0034] In the formula, T pre,i Indicates for the first i Alarm window length set for each sub-frequency band; K 2 represents the periodic coefficient, which is greater than or equal to 6.
[0035] In order to quickly capture the signal in the early stage of oscillation, the warning time window lengths for the three frequency bands (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) are set to 25.6s, 8s, and 3.2s, respectively.
[0036] To ensure the accuracy of spectrum analysis and eliminate the picket fence effect, the alarm time window lengths for the three frequency bands (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) are set to 102.4s, 32s, and 12.8s, respectively.
[0037] The aforementioned steps S1-S4 are identical to the existing technologies mentioned in the background section and are not the core innovation of this solution. The innovation of this solution lies in steps S5-S6.
[0038] S5: For the first i The system acquires multiple warning data windows across several sub-frequency bands and determines whether to issue a warning. Specifically, it includes the following steps: S5.1: For the first i Filtering sequences for each sub-band , in time t As the cutoff time for the early warning data window, data is truncated forward according to the early warning time window length set in S4 to obtain... t A real-time early warning data window; S5.2: As time progresses, at a set time ( t + Km ), to obtain the first i Filtering sequences for each sub-band , K Represents positive integers. m Indicates the set warning time interval; in time ( t + Km The warning data window is used as the cutoff time. Data is extracted forward from the warning time window length set in S4 to obtain the warning data window at the corresponding time. All warning data windows are numbered in chronological order. S5.3: Add a Hanning window to all warning data windows to suppress spectral leakage, and then perform a fast Fourier transform; extract the frequency domain peak amplitude of each warning data window; and determine whether to issue a warning according to the set rules.
[0039] The rules are as follows:
[0040] In the formula, t k Indicates the first k The deadline for each warning data window; t last_alarm Indicates the time of the last alarm; T cool This indicates the preset cooldown time, set to 40 seconds. A peak ( k ) indicates the first k The peak amplitude of each warning data window; A peak _ set This indicates the set peak amplitude. A peak ( k- 2) indicates the first k- Peak amplitude of the two warning data windows; A peak ( k- 1) indicates the first k- The peak amplitude of one warning data window.
[0041] When the above rules are met, the first... k The deadline for each warning data window is the warning time.
[0042] Step S5 proposes an early warning method, which is a prerequisite for improving the speed of alarms in the subsequent S6.
[0043] S6: Calculate the total number of oscillation cycles for each alarm window; initiate Prony analysis, calculate the damping ratio of the warning time window after the warning is triggered in S5, then compare the damping ratio with the set threshold, and determine whether to issue an alarm based on the comparison result and the total number of oscillation cycles for the corresponding alarm window. Specifically, this includes the following steps: S6.1: For the first i Filtering sequences for each sub-band , in time t As the cutoff time for the alarm data window, data is truncated forward according to the alarm time window length set in S4 to obtain... t Real-time alarm data window; S6.2: As time progresses, at a set time, the subsequent... iFiltering sequences for each sub-band , K Represents positive integers. d This indicates the set alarm time interval. d Greater than m ; by time ( t + Kd The alarm data window is cut off at the end time of the alarm data window. The data is extracted forward from the alarm time window length set in S4 to obtain the alarm data window at the corresponding time. All alarm data windows are numbered in chronological order. S6.3: For all alarm data windows, a Hanning window is added to suppress spectral leakage, followed by a Fast Fourier Transform; for the first... z For each alarm data window, extract the frequency corresponding to the maximum amplitude within that alarm data window. f peak,z Then obtain the maximum amplitude value. A t,z .
[0044] S6.4: Calculate the increment of the oscillation frequency for the z-th alarm data window. C m,z The formula is:
[0045] In equation (1), N alarm This indicates a long alarm time window; f L,i Indicates the first i The lower limit frequency of each sub-band.
[0046] S6.5: Calculate the total number of oscillation cycles in the z-th alarm data window. The formula is:
[0047] In the formula, C z-1 This represents the number of oscillation cycles in the (z-1)th alarm data window; due to the penalty mechanism in the second row of equation (1), ... C m,z It may be negative, which further leads to C z-1 It is possible for it to become negative, when C z-1 When it is negative, it is set to 0; the aforementioned steps S6.1-S6.5 are similar to the comparative document technology mentioned in the background technology and are not the core innovation point. The innovation point lies in S6.6-S6.7.
[0048] S6.6: Calculate all damping ratios within one warning time window after the warning is triggered; specifically including the following steps: S6.6.1: For the i-th sub-frequency band, acquire the data of the warning time window after the warning is issued, and set this window as the judgment window; for example, for the sub-frequency band 0.9-2.5 Hz, acquire the data within the 3.2s window after the warning signal is issued, based on its sampling rate, there are a total of 64 sampling points; For the data in this judgment window x i [ n Perform a Prony analysis, assuming it is composed of the superposition of p complex exponential signals, and analyze this data. x i [ n The decomposition is performed using linear constant-coefficient difference, and the formula is as follows:
[0049] In the formula, x'p Indicates the decomposition of the first... p One data point; a p Indicates the first p The corresponding data point is the first... p Each coefficient.
[0050] Therefore, the characteristic polynomial can be constructed, with the following formula:
[0051] In the formula, Z Representing characteristic variables, P The degree is indicated by the number of solutions. The root obtained is the pole Z in the Z-plane. k 。
[0052] From the pole Z of the Z-plane k Inversely tracing the poles in the time domain S k From this, we can obtain the corresponding real and imaginary part values, as shown in the formula:
[0053] In the formula, f s Indicates the first i Sampling rate of each sub-band; Indicates the first k The real part of each root; Indicates the first k The imaginary part of each root; j It is the imaginary unit.
[0054] S6.6.2: Calculate the damping ratio based on the real and imaginary parts, using the following formula:
[0055] In the formula, Indicates the first k The damping ratio of each root.
[0056] S6.6.3: From k Among the damping ratios of the roots, select the smallest damping ratio.
[0057] S6.7: The minimum damping ratio in S6.6 It compares the result with a set threshold and issues an alarm based on the comparison result.
[0058] when When the value is less than 0, an alarm signal is immediately issued based on the comparison result, and the total number of oscillation cycles before the alarm time is reset to 0. For example, if the total number of oscillation cycles up to the z-th alarm cycle is calculated at the alarm time... Then Reset to 0; when When the value is greater than or equal to 0 and less than 0.1, then the total number of oscillation cycles in the z-th alarm data window is... Greater than or equal to the set threshold C 阈 Immediately upon receiving the judgment result, an alarm signal is issued, and the total number of oscillation cycles in the z-th alarm data window is recorded. Reset to 0. If the total number of oscillation cycles... Less than the set threshold C 阈 If the total number of oscillation cycles is greater than or equal to the set threshold, an alarm signal will be issued immediately based on the judgment result, and the total number of oscillation cycles in the alarm data window will be reset to 0.
[0059] when If the value is greater than or equal to 0.1, it is determined to be a single disturbance signal. After the disturbance signal is determined, the total number of oscillation cycles before the determination result time will be reset to 0. If the disturbance signal occurs exactly when the total number of oscillation cycles up to the z-th alarm cycle is calculated... Then Reset to 0; and do not issue an alarm, but output a warning message indicating a false alarm.
[0060] Example Analysis: A simulation comparison was conducted between the proposed solution and the solution mentioned in the background section.
[0061] Figure 1 and Figure 2 This refers to the processing of the disturbance signal by both methods after a single disturbance signal is added to frequency band 1 at 250.0s. Figure 1The existing technology simply accumulates the cycles and issues an alarm at 256.8s. In this solution, after the warning (252.0s) is triggered, data for a trigger window length (3.2s) is analyzed, and based on the damping ratio, the warning signal at 250.0s is identified as a false alarm signal at 255.2s.
[0062] Figure 3 and Figure 4 This is a warning chart for oscillation alarms in frequency band 2. Among them... Figure 3 It is an alarm graph obtained by accumulating cycles using existing technology. Figure 4 This solution combines early warning triggering with damping ratio analysis to obtain the alarm diagram, from... Figure 3 and Figure 4 It can be seen that both existing technologies and this solution can issue an alarm signal within 424.0 seconds. Although both this solution and existing technologies have alarm functions, this solution has an advantage because it sets a warning before the alarm is issued.
[0063] Figure 5 and Figure 6 This is a hysteresis information graph for oscillation alarms in frequency band 3. From... Figure 5 and Figure 6 As can be seen, the alarm time in the existing technology is 160.0s, while the alarm time of this solution is 115.6s. The alarm time of this solution is significantly reduced compared to the existing technology, thus improving the speed of alarm response.
[0064] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.
Claims
1. A method for early warning and alarm of low-frequency oscillations in power systems, characterized in that, Includes the following steps: S1: Monitor the active power sequence at the generator output. P (t); S2: Divide the full frequency range of the low-frequency oscillation into sub-bands, and within the corresponding sub-bands, analyze the active power sequence. P (t) is downsampled; the sub-bands are: (0.1Hz~0.3Hz), (0.3Hz~0.9Hz) and (0.9Hz~2.5Hz); S3: Filter the downsampled sequence to obtain the... i Filtering sequences for each sub-band , i ∈[1,3] and is an integer; S4: For the first i Each sub-frequency band can be configured with a warning time window and an alarm time window. S5: For the first i Each sub-frequency band acquires multiple early warning data windows and determines whether to issue an early warning; S6: Calculate the total number of oscillation cycles for each alarm window; start Prony analysis, calculate all damping ratios in one alarm time window after the alarm is triggered and select the smallest damping ratio, then compare the smallest damping ratio with the set threshold, and determine whether to issue an alarm based on the comparison result and the total number of oscillation cycles for the corresponding alarm window.
2. The method for early warning and alarm of low-frequency oscillation in power systems according to claim 1, characterized in that, Step S6 includes the following steps: S6.1: For the first i Filtering sequences for each sub-band , in time t As the cutoff time for the alarm data window, data is truncated forward according to the alarm time window length set in S4 to obtain... t Real-time alarm data window; S6.2: As time progresses, at a set time, the subsequent... i Filtering sequences for each sub-band , K Represents positive integers. d Indicates the set alarm time interval; in time ( t + Kd The alarm data window is cut off at the end time of the alarm data window. The data is extracted forward from the alarm time window length set in S4 to obtain the alarm data window at the corresponding time. All alarm data windows are numbered in chronological order. S6.3: For all alarm data windows, a Hanning window is added to suppress spectral leakage, followed by a Fast Fourier Transform; for the first... z For each alarm data window, extract the frequency corresponding to the maximum amplitude within that alarm data window. f peak,z Then obtain the maximum amplitude value. A t,z ; S6.4: Calculate the increment of the oscillation frequency for the z-th alarm data window. C m,z The formula is: ； N alarm This indicates a long alarm time window; f L,i Indicates the first i The lower limit frequency of each sub-band; S6.5: Calculate the total number of oscillation cycles in the z-th alarm data window, using the following formula: ； In the formula, C z-1 This represents the number of oscillation cycles in the (z-1)th alarm data window; due to the penalty mechanism in the formula, if no oscillation occurs within a certain period of time, then... C z-1 It is possible for it to become negative, when C z-1 When the value is less than 0, it is forced to be 0; S6.6: Calculate all damping ratios within one warning time window after the warning is triggered; for the i-th sub-band, obtain the data of one warning time window after the warning, and set this window as the judgment window; decompose the judgment window first using linear constant coefficient difference, then construct the characteristic polynomial and find the poles in the Z-plane, deduce the pole formula in the time domain based on the poles in the Z-plane, and finally calculate the damping ratio of all poles based on the pole formula in the time domain, and select the damping ratio with the smallest damping ratio; S6.7: Compare the minimum damping ratio in S6.6 with the set threshold, and issue an alarm based on the comparison result; when When the value is less than 0, an alarm signal is immediately issued based on the comparison result, and the total number of oscillation cycles in the z-th alarm data window is recorded. Reset to 0; when When the value is greater than or equal to 0 and less than 0.1, if the total number of oscillation cycles is greater than or equal to the set threshold, an alarm signal will be issued immediately based on the judgment result; if the total number of oscillation cycles is less than the set threshold, the total number of oscillation cycles will continue to be accumulated until it is greater than or equal to the set threshold. when If the value is greater than or equal to 0.1, it is determined to be a single disturbance signal, and the total number of oscillation cycles in the z-th alarm data window is set accordingly. Reset to 0 and do not issue an alarm, but output a warning message indicating a false alarm.
3. The method for early warning and alarm of low-frequency oscillation in a power system according to claim 2, characterized in that, Step S2 includes the following steps: S2.1: Set the target sampling rates for each sub-band (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) to 2.5Hz, 8Hz, and 20Hz, respectively; calculate the... i The downsampling factor for each sub-band is calculated using the following formula: ； In the formula, A i Indicates the first i Downsampling factor for each sub-band; f raw Represents active power sequence P The sampling rate of (t); f smp,i Indicates the first i The target sampling rate is set for each sub-band; Based on the i The downsampling factor of each sub-band affects the active power sequence. P (t) is used to extract the corresponding downsampling sequence. P i (t).
4. The method for early warning and alarm of low-frequency oscillation in a power system according to claim 3, characterized in that, Step S5 includes the following steps: S5.1: For the first i Filtering sequences for each sub-band , in time t As the cutoff time for the early warning data window, data is truncated forward according to the early warning time window length set in S4 to obtain... t A real-time early warning data window; S5.2: As time progresses, at a set time ( t + Km ), to obtain the first i Filtering sequences for each sub-band , K Represents positive integers. m Indicates the set warning time interval; in time ( t + Km The warning data window is used as the cutoff time. Data is extracted forward from the warning time window length set in S4 to obtain the warning data window at the corresponding time. All warning data windows are numbered in chronological order. S5.3: Add a Hanning window to all warning data windows to suppress spectral leakage, then perform a fast Fourier transform to extract the frequency domain peak amplitude of each warning data window, and determine whether to issue a warning according to the set rules; The rules are as follows: ； In the formula, t k Indicates the first k The deadline for each warning data window; t last_alarm Indicates the time of the last alarm; T cool Indicates the preset cooldown time; A peak ( k ) indicates the first k Peak amplitude of each warning data window; A peak _ set This indicates the set peak amplitude; A peak ( k- 2) indicates the first k- Peak amplitude of the two warning data windows; A peak ( k- 1) indicates the first k- Peak amplitude of one warning data window; When the above rules are met, the first... k The deadline for each warning data window is the warning time.
5. The method for early warning and alarm of low-frequency oscillation in a power system according to claim 1, characterized in that, The formula for the warning window length is: ； In the formula, T pre,i Indicates for the first i The warning window length is set for each sub-frequency band; K 1 represents the periodicity coefficient, which is set to 2~3; The formula for alarm window length is: ； In the formula, T pre,i Indicates for the first i Alarm window length set for each sub-frequency band; K 2 represents the periodic coefficient, which is greater than or equal to 6.
6. The method for early warning and alarm of low-frequency oscillation in a power system according to claim 5, characterized in that, The warning time window lengths for the three frequency bands (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) are set to 25.6s, 8s, and 3.2s, respectively; the alarm time window lengths for the three frequency bands (0.1Hz~0.3Hz), (0.3Hz~0.9Hz), and (0.9Hz~2.5Hz) are set to 102.4s, 32s, and 12.8s, respectively.
7. A method for early warning and alarm of low-frequency oscillation in a power system according to claim 6, characterized in that, d Greater than m .