New energy station network source coordination synchronization reference signal detection method
By employing the Hanning window and EMD algorithm to decompose signals in new energy power plants, and combining it with the HHT and windowed interpolation DFT algorithm, the problem of large detection error of reference signals in low-inertia power systems by the traditional DFT algorithm is solved, and high-accuracy detection of synchronous reference signals is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-17
- Publication Date
- 2026-04-07
AI Technical Summary
In the grid connection testing and performance evaluation of new energy power plants, the traditional DFT algorithm has a large error when detecting the synchronization reference signal in low-inertia power systems, and cannot effectively cope with frequency offset and rapid changes, affecting the safe and stable operation of the system.
The signal is captured using the Hanning window, and decomposed using the DFT and EMD algorithms. By identifying the instantaneous frequency of the IMF principal component, the reference signal is obtained using the HHT algorithm and the windowed interpolation DFT algorithm, which can adapt to the frequency shift and changes of new energy power plants.
It improves the accuracy and scientific nature of the synchronization reference signal detection for new energy power stations, reduces detection errors, is easy to standardize, and is adaptable to signal analysis under complex operating conditions.
Smart Images
Figure CN121808472A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of reference signal detection. Background Technology
[0002] In low-inertia power systems with a high proportion of renewable energy (renewable energy power generation facilities themselves do not have mechanical inertia, which makes it difficult for renewable energy to buffer frequency changes through inertial response when facing load changes or power generation fluctuations), the system frequency exhibits characteristics such as severe fundamental frequency offset and rapid frequency changes. This leads to increased detection errors of synchronization reference signals when renewable energy power stations are conducting grid connection tests and supporting performance assessments, which may threaten the safe and stable operation of the system.
[0003] When the power grid signal only experiences frequency shift, i.e., under static conditions, the reference signal... Represented as:
[0004] (1)
[0005] In the formula, The amplitude of the static reference signal; , This represents the original reference frequency and offset; This represents the initial phase corresponding to the reference frequency. It can be seen that in the static reference signal, the amplitude, frequency offset, and RoCoF (rate of change of frequency) are all constants, and RoCoF is 0; the phase is linearly time-varying. For ease of analysis, the static and dynamic power grid signals only represent the fundamental frequency component, ignoring harmonic and interharmonic components.
[0006] When a new energy power station undergoes tests such as synchronous support, rapid frequency regulation, and voltage regulation, or when the system experiences power deficit, power oscillations occur at the grid connection point, but the system does not lose synchronization. At this time, the amplitude and phase angle of the dynamic grid signal will undergo sinusoidal modulation at the same frequency, with an initial phase difference of 180°. The analytical expression is as follows:
[0007] (2)
[0008] In the formula, The amplitude of the dynamic reference signal; , This represents the original reference frequency and the corresponding initial phase. , This represents the modulation frequency and the initial modulation phase; , These represent the modulation depth of amplitude and phase, respectively.
[0009] From equation (2), the frequency shift and RoCoF of the dynamic signal can be further calculated:
[0010] (3)
[0011] Equation (3) shows that the frequency shift and RoCoF are both nonlinear time-varying in the form of trigonometric functions.
[0012] Considering the complex operating conditions such as power deficit that occur during grid connection testing or commissioning of power stations, the difficulty in evaluating the synchronous support performance of new energy power stations lies in accurately detecting the synchronous reference signal.
[0013] In traditional synchronous power systems, Direct Frequency Theory (DFT) is typically used to detect the reference signal and harmonics. This method is computationally inexpensive, easy to implement, and standardized. However, in new low-inertia power systems, frequencies exhibit diverse characteristics such as power frequency deviation and rapid frequency changes. Therefore, it is necessary to evaluate the adaptability of the DFT algorithm and quantify the detection error.
[0014] Traditional DFT-based algorithms for reference signal and harmonic detection use a rectangular window to extract discretized data, then employ the FFT algorithm to obtain the signal spectrum, and subsequently derive the fundamental frequency and harmonic parameters. When using traditional DFT algorithms to detect reference signals, the maximum error in fundamental frequency amplitude detection can reach 10.5% in power frequency offset scenarios, while standards require an error within 1%, severely reducing detection accuracy. To address this issue and suppress spectral leakage, high-performance window functions such as the Hanning window, Hamming window, and Blackman window are commonly used to extract data. To reduce the impact of the picket fence effect, spectral interpolation algorithms are often used to correct harmonic characteristics, such as single-line, double-line, and triple-line interpolation algorithms. Taking the Hanning windowed double-line interpolation DFT algorithm as an example, existing research on windowed interpolation DFT algorithms is relatively mature, often improving accuracy by increasing the single window length. However, in high RoCoF scenarios, the frequency of the signal within the window varies greatly, and DFT cannot characterize the time-varying characteristics of the frequency. It only reflects the spectral performance of the signal within that time period. Therefore, traditional DFT-based reference signals and harmonic detection algorithms are not applicable, and the detection error is large. Summary of the Invention
[0015] The purpose of this invention is to solve the problem of large errors in the detection of reference signals by traditional DFT-based reference signal and harmonic detection algorithms, and to propose a method for detecting reference signals for grid-source coordination and synchronization in new energy power stations.
[0016] A method for detecting reference signals for grid-source coordination synchronization at new energy power stations, the method comprising the following:
[0017] Step 1: Collect the signal generated during the grid connection of new energy sources as the signal to be tested, use the Hanning window to extract the signal from the signal to be tested, and use the DFT algorithm to obtain the frequency offset of the extracted signal.
[0018] Step 2: Determine whether the frequency offset of the intercepted signal is within the preset range. If yes, proceed to step 3; otherwise, proceed to step 4.
[0019] Step 3: The DFT algorithm obtains the reference signal based on the truncated signal;
[0020] Step 4: Decompose the truncated signal using the EMD algorithm to obtain each IMF component, calculate the correlation coefficient between each IMF component and the signal to be detected, select the IMF component corresponding to the maximum correlation coefficient as the reference component, use the HHT algorithm in combination with the reference component to obtain the instantaneous frequency distribution, and use the windowed interpolation DFT algorithm to obtain the reference signal based on the instantaneous frequency distribution.
[0021] Preferably, the preset range is [-0.32, 0.47].
[0022] Preferably, the correlation coefficient is expressed as:
[0023] ,
[0024] In the formula, for and covariance, for The variance.
[0025] Preferably, instantaneous frequency distribution Represented as:
[0026] ,
[0027] In the formula, , , The instantaneous phase of the reference component, The signal after performing a Hilbert transform on the reference component.
[0028] Preferably, in step 4, the EMD algorithm is used to decompose the truncated signal. The specific process is as follows:
[0029] If the truncated signal undergoes a sudden change, the truncated signal is divided into two parts with the moment of the change as the dividing point. Then, the EMD algorithm is used to decompose the two parts of the signal separately to obtain each IMF component.
[0030] If the truncated signal does not undergo a sudden change, the EMD algorithm is used to decompose the truncated signal to obtain each IMF component.
[0031] Preferably, step 3 further includes: using the DFT algorithm to select the peak spectral line near the i-th harmonic, extracting the amplitude and phase of the peak spectral line, and completing the harmonic detection.
[0032] Preferably, step 3 further includes: the DFT algorithm can obtain the peak spectral line near the i-th harmonic based on the truncated signal, extract the amplitude and phase of the peak spectral line, and complete the harmonic detection.
[0033] Preferably, step 4 further includes: the windowed interpolation DFT algorithm can also obtain the amplitude and phase of the peak spectral line near the i-th harmonic based on the instantaneous frequency distribution, thus completing the harmonic detection.
[0034] The beneficial effects of this invention are:
[0035] First, obtain the frequency offset during the grid connection of new energy sources. If the signal is indeed at the optimal level, the traditional DFT algorithm is used to obtain the reference signal; otherwise, EMD decomposition is performed, and the equivalent fundamental frequency is obtained by identifying the instantaneous frequency of the IMF principal component. This equivalent fundamental frequency is then input into the dynamic dual DFT spectral interpolation algorithm to obtain the reference signal. This invention provides high accuracy in detecting the reference signal. It is scientifically sound, requires no subjective experience, and is easy to implement and standardize. Attached Figure Description
[0036] Figure 1 A flowchart for a method of detecting reference signals for grid-source coordination and synchronization at new energy power plants;
[0037] Figure 2 The spectrum of the test signal after adding a rectangular window;
[0038] Figure 3 This is the spectrum of the test signal after the fundamental frequency shift occurs. Detailed Implementation
[0039] The technical solutions of the embodiments 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, and 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.
[0040] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.
[0041] Example:
[0042] A method for detecting reference signals for grid-source coordination synchronization at new energy power stations, the method comprising the following:
[0043] Step 1: Collect the signal generated during the grid connection of new energy sources as the signal to be tested, use the Hanning window to extract the signal from the signal to be tested, and use the DFT algorithm to obtain the frequency offset of the extracted signal.
[0044] Step 2: Determine whether the frequency offset of the intercepted signal is within the preset range. If yes, proceed to step 3; otherwise, proceed to step 4.
[0045] Step 3: The DFT algorithm obtains the reference signal based on the truncated signal;
[0046] Step 4: Decompose the truncated signal using the EMD algorithm to obtain each IMF component, calculate the correlation coefficient between each IMF component and the signal to be detected, select the IMF component corresponding to the maximum correlation coefficient as the reference component, use the HHT algorithm in combination with the reference component to obtain the instantaneous frequency distribution, and use the windowed interpolation DFT algorithm to obtain the reference signal based on the instantaneous frequency distribution.
[0047] Further, the preset range is [-0.32, 0.47].
[0048] Further specifying, the correlation coefficient is expressed as:
[0049] ,
[0050] In the formula, for and covariance, for The variance.
[0051] Further specifying, instantaneous frequency distribution Represented as:
[0052] ,
[0053] In the formula, , , The instantaneous phase of the reference component, The signal after performing a Hilbert transform on the reference component.
[0054] Further specifying, in step 4, the EMD algorithm is used to decompose the truncated signal. The specific process is as follows:
[0055] If the truncated signal undergoes a sudden change, the truncated signal is divided into two parts with the moment of the change as the dividing point. Then, the EMD algorithm is used to decompose the two parts of the signal separately to obtain each IMF component.
[0056] If the truncated signal does not undergo a sudden change, the EMD algorithm is used to decompose the truncated signal to obtain each IMF component.
[0057] Further specifying step 3, it also includes: using the DFT algorithm to select the peak spectral line near the i-th harmonic, extracting the amplitude and phase of the peak spectral line, and completing the harmonic detection.
[0058] Further specifying step 3, it also includes: the DFT algorithm obtains the peak spectral line near the i-th harmonic based on the truncated signal, extracts the amplitude and phase of the peak spectral line, and completes the harmonic detection.
[0059] Further specifying, step 4 also includes: the windowed interpolation DFT algorithm obtains the amplitude and phase of the peak spectral line near the i-th harmonic based on the instantaneous frequency distribution, thus completing the harmonic detection.
[0060] Specifically, the error analysis of the synchronization reference signal detection in traditional DFT:
[0061] Traditional DFT-based harmonic detection algorithms extract discretized data through a rectangular window, then use the FFT algorithm to obtain the signal spectrum, and subsequently determine the fundamental frequency and harmonic parameters. For ease of explanation, a single-frequency signal is used to analyze the error generation mechanism of the DFT algorithm and to quantize it.
[0062] Let the discretized measured signal S(k) be...
[0063] (4)
[0064] In the formula, , and They represent the fundamental frequency. The amplitude, frequency, and phase of the offset signal The sampling step size, It is a step-size sequence. =0, 1, 2...N-1.
[0065] rectangular window The time-frequency domain expression is
[0066] (5)
[0067] (6)
[0068] In the formula, Due to the data window length, and referring to signal measurement standards such as "GB / T 20840.103 Application of Instrument Transformers in Power Quality Measurement", this article... Take 200ms, that is, the frequency resolution. It is 5Hz.
[0069] The spectrum of the test signal after adding a rectangular window is
[0070] (7)
[0071] It can be seen that because the rectangular window covers the entire frequency domain, the spectrum of the test signal after adding the window is no longer a single spectral line, but becomes a spectrum containing a single main lobe and multiple side lobes, causing spectral energy leakage. This is also the inherent leakage caused by windowing the signal. In addition, combined with equations (3) and (4), it can be seen that the spectrum of the test signal after adding the rectangular window is 0 at f-f1=5n (n is an integer other than 0). Therefore, when the power frequency does not shift, because For frequencies that are integer multiples of the frequency resolution, the sidelobe spectrum in the negative frequency region does not affect the main lobe spectrum in the positive frequency region. However, when the power frequency shifts to... When, and does not satisfy the condition of being an integer multiple of the frequency resolution, such as Figure 2 As shown, the relative error of the main lobe spectrum amplitude detection at this time is... It can be represented as:
[0072] (8)
[0073] When the power frequency offset is denoted as ,Right now = - ,and =200ms When the frequency is 50Hz, equation (8) can be further simplified to:
[0074] (9)
[0075] Considering that the power frequency deviation in current frequency accidents is within 1.25Hz, this range is used to analyze the detection error caused by spectrum leakage. From equation (9), it can be seen that when only considering spectrum leakage error, when -1.25 ≤ For frequencies ≤1.25Hz, the relative error increases with the increase of the power frequency offset, reaching a maximum of 1.63%. For multi-frequency signals, the error may increase further when considering the influence of other spectral sidelobes.
[0076] In the k-point discrete DFT algorithm, the actual obtained spectrum is the discrete spectrum after k-point equally spaced sampling. Frequency interval That is, the frequency resolution is 5Hz, so all the frequency points in the observed discrete spectrum are... Integer multiples of the power frequency. When the power frequency shifts, this condition is not met, and the characteristics of the measured signal can only be represented by the parameters of nearby spectral lines, thus generating analysis errors. This phenomenon is called the picket fence effect. At this time, considering both spectral leakage and the picket fence effect, the relative detection error can be expressed as:
[0077] (10)
[0078] Similarly, when T=200ms and f0=50Hz, equation (4-38) can be further simplified to:
[0079] (11)
[0080] By solving for the monotonicity of equation (11), it can be seen that when -1.25 ≤ For frequencies ≤1.25Hz, the relative error of the detection will also increase with the increase of the power frequency offset, reaching up to 11.8%.
[0081] (2) Error analysis of synchronization reference signal detection based on windowed interpolation DFT:
[0082] When using traditional DFT algorithms to detect reference signals, the maximum error in detecting the fundamental frequency amplitude can reach 10.5% in power frequency offset scenarios, while current domestic and international standards require an error within 1%, severely reducing the accuracy of detection. To address this issue and suppress spectral leakage, high-performance window functions are often used to truncate the data, such as the Hanning window, Hamming window, and Blackman window. To reduce the impact of the picket fence effect, spectral interpolation algorithms are often used to correct harmonic characteristics, such as single-line, double-line, and triple-line interpolation algorithms. Taking the double-line interpolation DFT algorithm with a Hanning window as an example, the basic principle is as follows.
[0083] Define the fundamental frequency Corresponding spectral line sequence number for = / Due to power frequency shift, Generally, these are not integers, meaning they cannot be observed in actual DFT discrete spectra. Therefore, the spectral lines k2 and k3 to the left and right of the fundamental frequency are used to characterize the fundamental spectrum. First, auxiliary variables a and b are introduced, defined as follows:
[0084] (12)
[0085] In the formula, , These represent the amplitudes of the spectral lines to the left and right of the fundamental frequency.
[0086] Combining equations (7) and (8), and ignoring the sidelobe effect of the negative frequency part, the analytical relationship between a and b can be derived.
[0087] (13)
[0088] Using polynomial fitting algorithms such as Prony, a can be solved, and then the correction formulas for frequency, amplitude and phase can be obtained, as shown in equation (12). It can be seen that the amplitude correction formula in equation (13) is obtained by taking the weighted average of the amplitudes of the spectral lines to the left and right of the fundamental frequency.
[0089] Existing research on windowed interpolation DFT algorithms is relatively mature, and accuracy is often improved by increasing the single window length. However, in high RoCoF scenarios, the frequency variation of the signal within the window is large, and DFT cannot characterize the time-varying frequency features; it only reflects the spectral performance of the signal within that time period. Therefore, such algorithms are not applicable.
[0090] (17)
[0091] Because the inertia levels of my country's power grid are inconsistent, the degree of frequency offset and RoCoF caused by disturbances will vary. Therefore, indiscriminately using the proposed algorithm for synchronization reference signal detection will reduce detection efficiency. Furthermore, as shown in the formula, both power frequency offset and RoCoF are related to the system's inertia level and power fluctuations, meaning that the degree of power frequency offset can also reflect RoCoF characteristics. Therefore, this paper uses the degree of frequency offset as the basis for algorithm selection. Referring to the power quality measurement accuracy requirements in "GB / T20840.103 Application of Instrument Transformers in Power Quality Measurement": the allowable error for the 1st to 2nd harmonics is 1%, and the allowable error for the 3rd to 50th harmonics is 5%, substituting into the correlation coefficient formula, the algorithm selection threshold range based on the degree of frequency offset can be obtained: [ , = [-0.32, 0.47].
[0092] Considering that the Hanning window has a suitable main lobe width and side lobe attenuation rate, and thus a strong ability to suppress spectral leakage; and that the bispectral interpolation algorithm performs well in both accuracy and computational speed, this embodiment utilizes the Hanning window for sampling and employs the bispectral interpolation algorithm to correct harmonic characteristics. The overall detection scheme is as follows: Figure 1 As shown.
[0093] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.
Claims
1. A method for detecting reference signals for grid-source coordination synchronization at new energy power stations, characterized in that, The method includes the following: Step 1: Collect the signal generated during the grid connection of new energy sources as the signal to be tested, use the Hanning window to extract the signal from the signal to be tested, and use the DFT algorithm to obtain the frequency offset of the extracted signal. Step 2: Determine whether the frequency offset of the intercepted signal is within the preset range. If yes, proceed to step 3; otherwise, proceed to step 4. Step 3: The DFT algorithm obtains the reference signal based on the truncated signal; Step 4: Decompose the truncated signal using the EMD algorithm to obtain each IMF component, calculate the correlation coefficient between each IMF component and the signal to be detected, select the IMF component corresponding to the maximum correlation coefficient as the reference component, use the HHT algorithm in combination with the reference component to obtain the instantaneous frequency distribution, and use the windowed interpolation DFT algorithm to obtain the reference signal based on the instantaneous frequency distribution.
2. The method for detecting the reference signal for grid-source coordination synchronization at new energy power stations according to claim 1, characterized in that, The preset range is [-0.32, 0.47].
3. The method for detecting the reference signal for grid-source coordination synchronization at new energy power stations according to claim 1, characterized in that, The correlation coefficient is expressed as: ， In the formula, Cov(X,Y) is the covariance of X and Y, and Var(X) is the variance of X.
4. The method for detecting the reference signal for grid-source coordination synchronization at new energy power stations according to claim 1, characterized in that, Instantaneous frequency distribution Represented as: ， In the formula, , , The instantaneous phase of the reference component, The signal after performing a Hilbert transform on the reference component.
5. The method for detecting the reference signal for grid-source coordination synchronization at new energy power stations according to claim 1, characterized in that, In step 4, the EMD algorithm is used to decompose the truncated signal. The specific process is as follows: If the truncated signal undergoes a sudden change, the truncated signal is divided into two parts with the moment of the change as the dividing point. Then, the EMD algorithm is used to decompose the two parts of the signal to obtain each IMF component. If the truncated signal does not undergo a sudden change, the EMD algorithm is used to decompose the truncated signal to obtain each IMF component.
6. The method for detecting the reference signal for grid-source coordination synchronization at new energy power stations according to claim 1 or 5, characterized in that, Step 3 also includes: the DFT algorithm obtains the peak spectral line near the i-th harmonic based on the truncated signal, extracts the amplitude and phase of the peak spectral line, and completes the harmonic detection.
7. The method for detecting the reference signal for grid-source coordination synchronization at new energy power stations according to claim 6, characterized in that, Step 4 also includes: the windowed interpolation DFT algorithm obtains the amplitude and phase of the peak spectral line near the i-th harmonic based on the instantaneous frequency distribution, and completes the harmonic detection.