Band adaptive anti-noise power system high frequency oscillation detection method

Abstract

The application provides a kind of high-frequency oscillation detection method based on band adaptive anti-noise power system, comprising: installing three-phase high-frequency voltage transformer at resonance sensitive point, selecting data acquisition window width, collecting three-phase voltage data and constructing three-phase voltage tensor mathematical model;Voltage tensor in three-phase voltage tensor mathematical model is decomposed to Tucker to realize voltage denoising;The voltage tensor after denoising is converted into three-phase voltage, and peak detection is carried out using dynamic measure;According to the detected high-frequency oscillation signal and its frequency range, wavelet packet decomposition is carried out on three-phase voltage, so that high-frequency oscillation signal and low-frequency fundamental wave and related harmonic components are separated to the wavelet coefficients of each frequency band;The wavelet coefficients containing oscillation signal are subjected to oscillation signal restoration;The restored oscillation signal is subjected to exponential decomposition using Prony method, and high-frequency oscillation characteristics are obtained.The application can improve the detection accuracy of high-frequency oscillation.

Description

Technical Field

[0001] The present invention relates to the technical field of protecting the safety of new power systems, and in particular to a method for detecting high-frequency oscillations in power systems based on frequency band self-adaptation and noise immunity. Background Art

[0002] The rapid rise of various new energy sources, including distributed photovoltaics, microgrid clusters, and energy storage, has led to a high penetration of power electronics devices, significantly increasing grid noise levels. These devices also induce a large number of high-frequency oscillations due to various impedance mismatches, posing significant challenges to the safe operation of the grid. Accurately detecting oscillation frequency and amplitude is essential for effectively suppressing oscillations, but existing methods struggle to accurately detect high-frequency oscillations in low signal-to-noise ratio environments. Summary of the Invention

[0003] The present invention provides a method for detecting high-frequency oscillations in power systems based on frequency band adaptive anti-noise. By constructing a third-order tensor of the three-phase voltage and utilizing the data's own structure to achieve data adaptive noise reduction, a high signal-to-noise ratio oscillation signal is obtained. The oscillation amplitude and frequency are then accurately obtained through frequency band adaptive analysis, thereby improving the accuracy of high-frequency oscillation detection.

[0004] A method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise includes the following steps:

[0005] Step 1: Install a three-phase high-frequency voltage transformer at a resonant sensitive point, select a data acquisition window width, collect three-phase voltage data, and construct a three-phase voltage tensor mathematical model based on the collected three-phase voltage data;

[0006] Step 2: performing Tucker decomposition on the voltage tensor in the three-phase voltage tensor mathematical model to achieve voltage noise reduction;

[0007] Step 3: Convert the denoised voltage tensor into a three-phase voltage, and perform peak detection on the three-phase voltage using dynamic measurement to detect the high-frequency oscillation signal and estimate the frequency range of the high-frequency oscillation signal;

[0008] Step 4: Perform wavelet packet decomposition on the three-phase voltage based on the high-frequency oscillation signal and its frequency range obtained by dynamic measurement in step 3, so that the high-frequency oscillation signal and the low-frequency fundamental wave and related harmonic components are separated into wavelet coefficients of each frequency band;

[0009] Step 5: restore the oscillation signal to the wavelet coefficients containing the oscillation signal;

[0010] Step 6: Use the Prony method to perform exponential decomposition on the oscillation signal restored in step 5 to obtain high-frequency oscillation characteristics.

[0011] Furthermore, in step 1, the duration T of the data acquisition window width is selected according to the calculation requirements. W , taking the integer multiple M of the period as the sampling rate, assuming N points per cycle, the three-phase sampled voltage is constructed into a three-phase voltage tensor mathematical model U(M,N,3).

[0012] Furthermore, the step 2 specifically includes:

[0013] 2.1) Perform Tucker decomposition on the three-phase voltage tensor mathematical model U(M,N,3) to obtain

[0014] U≈G×1A×2B×3C(2-1)

[0015] in is the core tensor, is the projection matrix of tensor decomposition, λ1, λ2, λ3 are the ranks of Tucker decomposition and are less than M, N, 3 respectively;

[0016] 2.2) Select an appropriate rank so that the core tensor G contains more than 99% of the energy of the third-order tensor U, and as the rank increases, the energy share of G increases by a gradient less than 1%.

[0017] Furthermore, the step 3 specifically includes:

[0018] 3.1) Determine the dynamic measurement threshold Th D , when the amplitude difference between two adjacent voltage extreme points detected by the peak is greater than Th D When , these two extreme points are marked using dynamic measurement;

[0019] 3.2) The extreme points marked by the dynamic measurement are made into a set, and the average duration of each two adjacent extreme points is calculated, which is recorded as t1, and the average duration of each f adjacent points is recorded as t f , find the inverse of the average duration to get the corresponding frequency, and record these frequencies as an array The frequency bands are divided with each element in the array as the center frequency, and the frequency bands are aligned, that is, [F1, F2, ..., F f ].

[0020] Furthermore, the step 4 specifically includes:

[0021] 4.1) Select the high-order vanishing moment function as the mother wavelet;

[0022] 4.2) Perform wavelet packet decomposition on the three-phase voltage signal based on the obtained frequency band and the mother wavelet, so that the frequency band of the last layer of detail wavelet is consistent with the frequency band in step 3.2):

[0023]

[0024] in and are the odd and even wavelet coefficients of wavelet decomposition respectively, and h and g are orthogonal low and high filter bank functions determined by the mother wavelet.

[0025] Furthermore, the step 5 specifically includes:

[0026] 5.1) Set the wavelet threshold according to the frequency band of each layer:

[0027] Assume that the wavelet coefficient d of the lth layer l , then the wavelet threshold is as follows:

[0028]

[0029] Where median is to find the middle value of the sequence;

[0030] 5.2) Perform inverse transform on each layer of wavelet coefficients layer by layer to restore the time domain form of the signal components in the frequency band. When there is a dynamic measurement threshold Th D If the signal duration is less than the window T W , it is determined to be a high-frequency oscillation, and the maximum extreme value P of the signal is taken as the oscillation amplitude.

[0031] Furthermore, the step 6 specifically includes:

[0032] 6.1) Use the high-frequency signal x obtained in 5.2) Perform the fit:

[0033]

[0034] Where p is the number of signal components and N is the number of signal sampling points;

[0035] 6.2) Use least squares to find Then the amplitude A of the i-th signal component is i , frequency f i , attenuation factor α i Obtained by the following formula:

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1) By constructing a three-phase voltage tensor and performing Tucker decomposition, the noise can be reduced and the high-frequency oscillation signal can be fully preserved in the case of low signal-to-noise ratio in a data-adaptive manner;

[0038] 2) Through dynamic measurement, a preliminary estimate of the high-frequency oscillation frequency is made to obtain the frequency range of the high-frequency oscillation, and then the oscillation characteristics can be obtained in a targeted manner without being interfered with by signals in other frequency bands;

[0039] 3) The oscillation signal is extracted through frequency-band adaptive wavelet packet decomposition, and the high-frequency oscillation characteristics are accurately obtained through the Prony algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of a three-phase voltage signal containing oscillations according to an embodiment of the present invention, with the unit being per-unit value pu and the signal-to-noise ratio being 30 dB.

[0041] Figure 2 Schematic diagram of a voltage signal after Tucker decomposition and noise reduction according to an embodiment of the present invention.

[0042] Figure 3 3. It is a schematic diagram of the components of the three-phase oscillation signal after dynamic measurement and frequency band adaptive wavelet decomposition and restoration according to an embodiment of the present invention.

[0043] Figure 4 This is a flow chart of a method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0045] See also Figure 4 The embodiment of the present invention provides a method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise, comprising the following steps:

[0046] Step 1: Install a three-phase high-frequency voltage transformer at the resonance sensitive point, select the data acquisition window width, and collect three-phase voltage data (such as Figure 1 As shown), and construct a three-phase voltage tensor mathematical model based on the collected three-phase voltage data;

[0047] The specific process of step 1 is as follows:

[0048] 1.1) Select the duration T of the data acquisition window width according to the calculation requirements W , measured in integer multiples of the period M;

[0049] 1.2) Assume that the sampling rate is N points per cycle, and construct the three-phase sampled voltage into a three-phase voltage tensor mathematical model U(M,N,3);

[0050] Step 2: Perform Tucker decomposition on the voltage tensor in the three-phase voltage tensor mathematical model to achieve voltage noise reduction. The noise reduction effect is as follows: Figure 2 ;

[0051] The specific process of step 2 is as follows:

[0052] 2.1) Perform Tucker decomposition on the three-phase voltage tensor mathematical model U(M,N,3) to obtain

[0053] U≈G×1A×2B×3C(2-1) is the core tensor, is the projection matrix of tensor decomposition, λ1, λ2, λ3 are the ranks of Tucker decomposition and are less than M, N, 3 respectively.

[0054] 2.2) Select an appropriate rank such that the core tensor G contains more than 99% of the energy of the third-order tensor U, and that the energy contribution of G increases by less than 1% as the rank increases. At this point, G contains almost no noise from U, and reconstructing it using G yields the denoised three-phase voltage.

[0055] Step 3: Convert the denoised voltage tensor into three-phase voltage and perform peak detection on the three-phase voltage using dynamic measurement. When dense peaks are measured in a short period of time, the dynamic measurement will increase sharply. Based on this, it can be determined whether there is a high-frequency oscillation signal and the frequency range of the high-frequency oscillation signal can be estimated.

[0056] The specific process of step 3 is as follows:

[0057] 3.1) Determine the dynamic measurement threshold Th D , when the amplitude difference between two adjacent voltage extreme points detected by the peak is greater than Th D , these two extreme points will be marked by dynamic measurement.

[0058] 3.2) The extreme points marked by the dynamic measurement are made into a set, and the average duration of each two adjacent extreme points is calculated, which is recorded as t1, and the average duration of each f adjacent points is recorded as t f , we can get the corresponding frequency by taking the inverse of this duration. These frequencies are recorded as an array The frequency bands are divided with each element in the array as the center frequency, and the frequency bands are aligned, that is, [F1, F2, ..., F f ].

[0059] Step 4: Perform wavelet packet decomposition on the three-phase voltage based on the high-frequency oscillation signal and its frequency range obtained by dynamic measurement in step 3, so that the high-frequency oscillation signal and the low-frequency fundamental wave and related harmonic components are separated into wavelet coefficients of each frequency band;

[0060] The specific process of step 4 is as follows:

[0061] 4.1) Select a high-order vanishing moment function as the mother wavelet, such as db40. This is done to minimize energy leakage in each wavelet layer, which helps separate oscillation signals.

[0062] 4.2) Perform wavelet packet decomposition on the three-phase voltage signal according to the obtained frequency band and the mother wavelet, so that the frequency band of the last layer of detail wavelet is consistent with the frequency band in step 3.2).

[0063]

[0064] in and are the odd and even wavelet coefficients of wavelet decomposition respectively, and h and g are orthogonal low and high filter bank functions determined by the mother wavelet.

[0065] Step 5: restore the oscillation signal to the wavelet coefficients containing the oscillation signal;

[0066] The specific process of step 5 is as follows:

[0067] 5.1) Set the wavelet threshold according to the frequency band of each layer:

[0068] Assume that the wavelet coefficient d of the lth layer l , then the wavelet threshold is as follows

[0069]

[0070] The median is the median value of the sequence. The wavelet threshold is mainly used to remove the interference of energy leakage of other wavelet layer coefficients on the wavelet coefficients of this layer.

[0071] 5.2) Perform inverse transform on each layer of wavelet coefficients layer by layer to restore the time domain form of the signal components in the frequency band. D If the signal duration is less than the window T W , then it is a high frequency oscillation. At this time, the maximum extreme value P of the signal is obtained as the oscillation amplitude, such as Figure 3 .

[0072] Step 6: Use the Prony method to perform exponential decomposition on the oscillation signal restored in step 5 to obtain high-frequency oscillation characteristics.

[0073] The specific process of step 6 is as follows:

[0074] 6.1) Use the high-frequency signal x obtained in 5.2) Perform the fit:

[0075]

[0076] Where p is the number of signal components and N is the number of signal sampling points.

[0077] 6.2) Use least squares to find Then the amplitude A of the i-th signal component is i , frequency f i , attenuation factor α i Obtained by the following formula:

[0078]

[0079] Where Δt is the sampling interval, Im is the imaginary part, and Re is the real part.

[0080] In a 30dB noise environment, the actual characteristics of the oscillation signal are compared with the oscillation characteristics obtained by the method of the present invention, as shown in Table 1:

[0081] Table 1 Comparison of the real characteristics of the oscillation signal in a 30dB noise environment and the oscillation characteristics obtained by the method of the present invention

[0082]

[0083] As shown in Table 1, the present invention constructs a three-phase voltage tensor and performs Tucker decomposition, which can reduce noise and completely preserve high-frequency oscillation signals in a data-adaptive manner under low signal-to-noise ratio conditions. Dynamic measurement is used to make a preliminary estimate of the high-frequency oscillation frequency and obtain the frequency range of the high-frequency oscillation, thereby specifically obtaining the oscillation characteristics without being interfered with by signals in other frequency bands. Band-adaptive wavelet packet decomposition is used to extract the oscillation signal, and the Prony algorithm is used to accurately obtain the amplitude, frequency, and attenuation of the oscillation.

[0084] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise, characterized in that: The steps include: Step 1: Install a three-phase high-frequency voltage transformer at a resonant sensitive point, select a data acquisition window width, collect three-phase voltage data, and construct a three-phase voltage tensor mathematical model based on the collected three-phase voltage data; Step 2: performing Tucker decomposition on the voltage tensor in the three-phase voltage tensor mathematical model to achieve voltage noise reduction; Step 3: Convert the denoised voltage tensor into a three-phase voltage, and perform peak detection on the three-phase voltage using dynamic measurement to detect the high-frequency oscillation signal and estimate the frequency range of the high-frequency oscillation signal; Step 4: Perform wavelet packet decomposition on the three-phase voltage based on the high-frequency oscillation signal and its frequency range obtained by dynamic measurement in step 3, so that the high-frequency oscillation signal and the low-frequency fundamental wave and related harmonic components are separated into wavelet coefficients of each frequency band; Step 5: restore the oscillation signal to the wavelet coefficients containing the oscillation signal; Step 6: Perform exponential decomposition on the oscillation signal restored in step 5 using the Prony method to obtain high-frequency oscillation characteristics; The step 5 specifically includes: 5.1) Set the wavelet threshold according to the frequency band of each layer: Assume that the wavelet coefficient of the first layer is , then the wavelet threshold is as follows: (5-1)； in To find the middle value of a sequence; 5.2) Perform inverse transform on each layer of wavelet coefficients layer by layer to restore the time domain form of the signal components within the frequency band. When there is a dynamic measurement threshold Th D If the signal duration is less than the window T W , it is determined to be a high-frequency oscillation, and the maximum extreme value P of the signal is taken as the oscillation amplitude; The step 6 specifically includes: 6.1) Use the high-frequency signal x obtained in 5.2) Perform the fit: (6-1)； Where p is the number of signal components and N is the number of signal sampling points; 6.2) Use least squares to find , then the amplitude of the i-th signal component ,frequency , attenuation factor Obtained by the following formula: (6-2)； in is the sampling interval, Im is the imaginary part, and Re is the real part.

2. The method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise according to claim 1, characterized in that: In step 1, the duration T of the data acquisition window width is selected according to the calculation requirements. W , taking the integer multiple M of the period as the sampling rate, assuming N points per cycle, the three-phase sampled voltage is constructed into a three-phase voltage tensor mathematical model .

3. The method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise as claimed in claim 2, wherein: The step 2 specifically includes: 2.1) Mathematical model of three-phase voltage tensor Perform Tucker decomposition to obtain (2-1)； in is the core tensor, , , is the projection matrix of tensor decomposition, , , is the rank of Tucker decomposition and is less than M, N, 3 respectively; 2.2) Select an appropriate rank so that the core tensor G contains more than 99% of the energy of the third-order tensor U, and as the rank increases, the energy share of G increases by a gradient less than 1%.

4. The method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise as claimed in claim 1, wherein: The step 3 specifically includes: 3.1) Determine the dynamic measurement threshold Th D , when the amplitude difference between two adjacent voltage extreme points detected by the peak is greater than Th D When , these two extreme points are marked using dynamic measurement; 3.2) The extreme points marked by the dynamic measurement are made into a set, and the average duration of each two adjacent extreme points is calculated, which is recorded as t1, and the average duration of each f adjacent points is recorded as t f , find the inverse of the average duration to get the corresponding frequency, and record these frequencies as an array , divide the frequency bands with each element in the array as the center frequency, and align the frequency bands, that is, .

5. The method for detecting high-frequency oscillation in a power system based on frequency band adaptive anti-noise as claimed in claim 4, characterized in that: The step 4 specifically includes: 4.1) Select the high-order vanishing moment function as the mother wavelet; 4.2) Perform wavelet packet decomposition on the three-phase voltage signal based on the obtained frequency band and the mother wavelet, so that the frequency band of the last layer of detail wavelet is consistent with the frequency band in step 3.2): (4-1)； in and are the even and odd wavelet coefficients of wavelet decomposition, and The low-high filter bank functions are orthogonal and are determined by the mother wavelet.

Images