A non-stationary sub-synchronous oscillation signal detection method, system and terminal

Abstract

The application discloses a non-stationary subsynchronous oscillation signal detection method, system and terminal, belongs to the field of power system oscillation detection, and comprises the following steps: designing an adaptive reset strategy of an EKF, introducing a residual error as a self-reset index, designing an adaptive reset threshold adjustment method based on an M2M4 estimator, and adaptively adjusting the reset threshold under different noise environments. When the self-reset criterion takes effect, the EKF parameters are initialized, and the signal is continuously detected, so that real-time monitoring of a noisy non-stationary SSO signal is realized. Finally, the effectiveness of the proposed method is verified through simulation analysis.

Description

Technical Field

[0001] This invention relates to the field of power system oscillation detection, specifically to a method, system, and terminal for detecting non-stationary subsynchronous oscillation signals. Background Technology

[0002] SSO (Signal Shift) characteristics in new energy grid-connected power systems exhibit time-varying and non-stationary features in terms of frequency and amplitude. Furthermore, SSO signals in real-world systems typically contain significant noise, making online identification difficult. Currently, commonly used SSO identification methods can be broadly categorized into two types: phasor-based methods and waveform-based methods. Phasor-based methods measure SSO parameters through spectral leakage components of the phasor and require complex algorithms to eliminate the influence of the fundamental frequency signal on the SSO component measurement. Additionally, these methods require observation windows longer than 2 seconds, making it impossible to capture rapidly changing SSO modes. On the other hand, waveform-based methods primarily rely on DFT (Discrete Functional Theory) and Prony. Classical DFT suffers from picket-fence effects and spectral leakage, and depends on a long observation window to improve resolution. Prony is a high-precision SSO identification method, but it lacks time-varying tracking capabilities. Moreover, when the SSO signal is subject to noise interference, Prony's identification accuracy is significantly affected.

[0003] Kalman filtering (KF) has been widely used for online synchronous phasor estimation, accurately estimating the amplitude, phase, and frequency of a signal under noise in a recursive manner. Recently, KF has also been applied to SSO detection. However, KF faces filter divergence problems when the SSO frequency changes abruptly. Furthermore, KF-based methods require prior knowledge of the SSO frequency. Some methods combine KF with a reset criterion to address these issues, but these methods do not provide a threshold selection standard for the reset criterion. Consequently, they cannot adaptively select the reset threshold for different noise environments when the signal-to-noise ratio of the measured signal varies, affecting the algorithm's detection performance and resulting in low adaptability. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a method for detecting non-stationary subsynchronous oscillation signals, aiming to accurately detect non-stationary SSO information occurring in a system under Gaussian noise interference.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A method for detecting non-stationary subsynchronous oscillation signals includes:

[0007] The signal data is preprocessed, and the signal parameters are calculated and the signal is reconstructed using a four-state EKF to obtain the fitted SSO signal at the current time. The residual ratio e is also calculated. rati o: wherein e k is the residual at kT s time, is the estimated value of the SSO signal;

[0008] the M2M4 estimator is used to calculate the signal-to-noise ratio of the signal and establish an adaptive reset threshold r e :

[0009]

[0010] wherein SNR is the signal-to-noise ratio;

[0011] when it is satisfied that if e ratio is greater than r e , then the error covariance matrix P in the four-state EKF k is reset to nI 4×4 , and the initial frequency value is updated.

[0012] Further, the residual ratio e ratio is calculated based on the residual e k :

[0013] the residual e at kT s time k is expressed as follows:

[0014]

[0015] wherein, y k is a measured quantity; the variance of e k can be approximately calculated by the following formula:

[0016]

[0017] wherein, m is a window length, if 1≤k<m, then m is replaced by k; the variance of the estimated SSO signal value can be calculated by the following formula:

[0018]

[0019] then the residual ratio e ratio satisfies:

[0020] Further, said step of using the M2M4 estimator to calculate the signal-to-noise ratio of the signal and establishing the adaptive reset threshold r e specifically comprises the following steps:

[0021] after calculation by the four-state EKF, when formulas (9) to (12) are satisfied, e ratio can be approximated as formula (13):

[0022]

[0023]

[0024]

[0025]

[0026]

[0027] Then the measured value y k The signal-to-noise ratio (SNR) can be calculated using the following formula:

[0028]

[0029] According to equations (13) and (14), e ratio This can be further expressed as:

[0030]

[0031] When the filter diverges, the fitted SSO signal will deviate significantly from the true value and satisfy... Therefore, the following inequalities hold:

[0032]

[0033] e k >v k (17)

[0034]

[0035] According to e ratio Greater than r e Given the conditions, equations (15), (16), and (18), the following inequalities hold:

[0036]

[0037] Combining equations (15) and (19), the threshold can be calculated using the following formula:

[0038]

[0039] According to equation (20), y can be measured. k Determine the threshold r when the SNR is calculated. e .

[0040] Furthermore, the process of calculating the signal-to-noise ratio of the signal using an M2M4 estimator and establishing an adaptive reset threshold includes the following steps:

[0041] Let M2 represent y k The second moment:

[0042]

[0043] Let M4 represent y k The fourth moment:

[0044]

[0045] Signal power is defined as S, and noise power is defined as N; M2 can be given by the following formula:

[0046] M2 = S + N (23)

[0047] Furthermore, M4 can be given by the following formula:

[0048] M4 = k x S 2 +6SN+k v N 2 (twenty four)

[0049] Where k x and k v These are signal kurtosis and noise kurtosis, respectively.

[0050] Solving equations (23) and (24) for S and N, we obtain:

[0051]

[0052]

[0053] form and The estimator of the ratio is expressed as the M2M4 estimator; for a zero-mean sinusoidal signal, k x =1.5, for noise, k v =3, therefore:

[0054]

[0055] The second and fourth moments are estimated by the time average of the measured signal, satisfying equations (28) and (29):

[0056]

[0057]

[0058] The measured signal y k The estimated SNR is:

[0059]

[0060] Considering the SNR estimation bias, an additional term Δ is introduced into equation (20) as follows to make the threshold adjustment more reliable:

[0061]

[0062] As mentioned above, the threshold r e Adaptive adjustments can be made by using equations (27) to (31).

[0063] Furthermore, when e is satisfied ratio Greater than r e At that time, and the time interval between resets is greater than the time interval T. Δ Only then is a reset performed, and the error covariance matrix P in the four-state EKF is changed. k Reset to nI 4×4 Furthermore, the initial frequency value is updated, the initialized signal data is re-initialized, and the signal parameters are calculated and the signal is reconstructed using the four-state EKF to obtain the SSO signal fitted at the current time.

[0064] Otherwise, the initialized signal data is directly re-initialized and the signal parameters are calculated and the signal is reconstructed using a four-state EKF to obtain the SSO signal fitted at the current time.

[0065] Furthermore, the additional term Δ is set to 2.5.

[0066] Furthermore, the feature is that the SSO signal and its parameters can be obtained using equations (1) to (4):

[0067]

[0068]

[0069] f S (kT s )=f S0 +x3(k) (3)

[0070] α S (kT s )=-x4(k) (4)

[0071] In the formula It is the SSO signal fitted by EKF using the state vector. Here, x1(k), x2(k), x3(k), and x4(k) are the estimated values ​​of the SSO signal, and x1(k), x2(k), x3(k), and x4(k) are the four state variables of the EKF; f S0 This is the initial value for the frequency.

[0072] Secondly, the present invention also provides a non-stationary subsynchronous oscillation signal detection system, comprising the following modules:

[0073] Signal preprocessing module: preprocesses the signal data and inputs the preprocessed signal data into the fitting module;

[0074] Fitting module: Calculates signal parameters and reconstructs the signal from the initialized signal data using a four-state EKF, and outputs the fitted SSO signal at the current time.

[0075] Calculation module: used to calculate the residual ratio e ratio The signal-to-noise ratio of the signal is calculated using an M2M4 estimator, and an adaptive reset threshold r is established. e ;

[0076] Self-reset judgment module: based on e ratio and r e Determine whether to reset the error covariance matrix P k and initial value of frequency;

[0077] When e is satisfied ratio Greater than r e At that time, and the time interval between resets is greater than the time interval T. Δ If so, a reset is performed, and the error covariance matrix P in the four-state EKF is changed. k Reset to nI 4×4 The system updates the initial frequency value and then inputs the initialized signal data back into the signal processing module for processing; otherwise, it directly inputs the initialized signal data into the signal processing module for processing.

[0078] Thirdly, the present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The memory stores the computer program capable of running on the processor, and when the processor loads and executes the computer program, it employs a non-stationary subsynchronous oscillation signal detection method as described in any of the above claims.

[0079] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, wherein when the computer program is loaded and executed by a processor, it employs a non-stationary subsynchronous oscillation signal detection method as described in any of the preceding claims.

[0080] The beneficial effects of this invention are:

[0081] A threshold adjustment method based on the M2M4 estimator is proposed, which can automatically determine the threshold of the reset criterion under different signal-to-noise ratios, ensuring the effectiveness of the reset criterion in different scenarios and making the self-reset EKF method more adaptive. Attached Figure Description

[0082] The invention will now be further described with reference to the accompanying drawings.

[0083] Figure 1 This is a flowchart illustrating the specific method in a particular embodiment of the present invention;

[0084] Figure 2 The image shows the signal being tested in scenario 1.

[0085] Figure 3 The result diagram of SSO reconstruction using the self-reset EKF for Scenario 1;

[0086] Figure 4 The image shows the SSO parameter identification results for scenario 1.

[0087] Figure 5 The image shows the signal being tested in scenario 2.

[0088] Figure 6 The SSO result diagram for the self-reset EKF reconstruction in Scenario 2;

[0089] Figure 7 The image shows the SSO parameter identification results for scenario 2. Detailed Implementation

[0090] 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.

[0091] This specific embodiment discloses an improved method for detecting non-stationary subsynchronous oscillation signals in an adaptively reset EKF. The detailed flowchart of the method is shown below. Figure 1 As shown, it includes the following steps:

[0092] S1: Initialize the parameters;

[0093] S2: Obtain three-phase measurement values, perform DQ conversion, and remove the DC component;

[0094] S3: If the discrete time k is less than the ratio T of the reset time interval and the sampling time interval. Δ / T s If yes, proceed to step S4; otherwise, proceed to step S5.

[0095] S4: Use parallel EKF to obtain the most suitable initial frequency value f S0 Then increment the discrete time by 1 and return to step S2;

[0096] S5: Calculate signal parameters and reconstruct the signal using a four-state EKF;

[0097] S6: Calculate the residual ratio e ratio ;

[0098] S7: Calculate the signal-to-noise ratio of the signal using the M2M4 estimator, and then use this to calculate the adaptive reset threshold r. e ;

[0099] S8: If e ratio Greater than r e And the time interval between resets is greater than T. Δ If yes, proceed to step S9; otherwise, proceed to step S10.

[0100] S9: The error covariance matrix P k Reset to nI 4×4 And update the initial frequency value;

[0101] S10: Output the calculation result at time k obtained from step S5, and then increment the discrete time by 1;

[0102] S11: If the signal detection is complete, then the process ends; otherwise, proceed to step S2.

[0103] In step S1, the following parameters are initialized:

[0104]

[0105] In the formula, The initial value of the state vector; P1 is the initial value of P. k The initial value of represents the covariance matrix of the state vector estimation error. 4×4 It is a 4×4 identity matrix, where n is a large integer, set to 100 in this invention. Q and R are the covariance matrix of the process noise equation and the noise covariance of the measurement equation, respectively, and their values ​​remain unchanged in this invention. m is the window length, which should be at least greater than the corresponding number of oscillation periods. For example, if the sampling interval T s If the value is 0.001s, then m should be greater than 1 / (f). Smin T s ) = 200, where f Smin It is the minimum SSO frequency, f Smin =5Hz. T Δ This is the minimum time interval between two resets, and its purpose is to prevent unnecessary resets during the convergence of the extended Kalman filter iteration.

[0106] In step S2, the three-phase measurement signal, after being transformed by the DQ converter, is represented by the instantaneous signal x in the DQ axis. dq (t) consists of the fundamental frequency and time-varying SSO components, and can be modeled as:

[0107]

[0108] In the formula, and These are the fundamental frequency component and the SSO component, respectively. Note that... There is a DC bias under the DQ axis. This invention focuses on the SSO component, therefore the DC bias needs to be removed first. The time-varying SSO component in equation (2) can be expressed as:

[0109]

[0110] In the formula, A S (t), f S (t) and φ S These are the time-varying amplitude, time-varying frequency, and initial phase of the SSO component, respectively. α S It is the damping factor of SSO.

[0111] In step S4, the present invention applies a set of EKFs (without using adaptive reset) to obtain a relatively accurate initial frequency value f. S0 Since the SSO frequency range is typically 5Hz to 45Hz (synchronization frequency 50Hz) or 55Hz (synchronization frequency 60Hz), the initial frequencies of multiple EKFs are set to 5Hz, 10Hz, 15Hz, ..., 45Hz (or 55Hz), respectively. All EKFs can run in parallel without increasing the computational burden. The selection of parallel EKFs is related to the residual ratio expectation E(e... ratio The minimum corresponding frequency is taken as the most suitable frequency. Then, the adaptive reset method and the most suitable initial frequency are applied to the EKF.

[0112] In step S5, a fixed sampling interval T is used. s If the measured signal is obtained, then the continuous-time model at the kth sampling point in (3) can be expressed as:

[0113]

[0114] With 2πf S (k) is a rotating reference frame, and the above signal can be rewritten as:

[0115]

[0116] and These are the in-phase and quadrature components, represented by state variables x1 and x2, respectively. The subsynchronous frequency offset is represented by the third state variable x3. The negative value of the damping factor is –α. S The fourth state variable, x4, represents this. Therefore, equation (5) can be written as the following four-state signal model:

[0117]

[0118] The SSO signal is calculated by the following formula:

[0119]

[0120] In the formula, is the SSO signal fitted by EKF using the state vector, is the estimated value of the SSO signal. The amplitude, frequency and damping factor of the SSO signal are calculated from the state estimation using the following formula:

[0121]

[0122] f S (kT s )=f S0 +x3(k) (9)

[0123] α S (kT s )=-x4(k) (10)

[0124] In the formula, x1(k), x2(k), x3(k) and x4(k) are four state variables of EKF; f S0 is the initial value of frequency.

[0125] In step S6, for EKF with Gaussian process, the residual e at time kT s is expressed as follows: k

[0126]

[0127] Wherein, y k is the measured quantity. The variance of e k can be approximately calculated by the following formula:

[0128]

[0129] In the formula, m is the window length, and if 1≤k<m, m is replaced by k. The variance of the estimated SSO signal can be calculated by the following formula:

[0130]

[0131] The self-reset criterion is defined as follows:

[0132]

[0133] In the formula, r e is the self-reset threshold. e ratio ​The residual ratio reflects the error between the estimated value and the true value.

[0134] In step S7, after the four-state EKF calculation, the following relationship exists:

[0135]

[0136]

[0137]

[0138]

[0139] Among them, v k This represents the noise in the EKF measurement equation.

[0140] Therefore, e ratio It can be approximated as:

[0141]

[0142] Note the measured value y k The signal-to-noise ratio (SNR) can be calculated using the following formula:

[0143]

[0144] According to equations (19) and (20), e ratio It can be represented as:

[0145]

[0146] According to equation (21), when the estimated SSO signal perfectly matches the true value, e ratio Size and y k The SNR is directly related. When the filter diverges, the estimated SSO signal will deviate significantly from the true value, and often has... Therefore, the following inequalities hold:

[0147]

[0148] e k >v k (twenty three)

[0149]

[0150] Based on equations (14), (21), (22), and (24), it is clear that the following inequality holds:

[0151]

[0152] Considering equations (21) and (25), the threshold can be calculated by the following equation:

[0153]

[0154] According to equation (26), the measurement y can be estimated. k Determine the threshold r when the SNR is calculated. e Therefore, the M2M4 estimator was applied to SNR estimation.

[0155] Let M2 represent y k The second moment:

[0156]

[0157] Let M4 represent y k The fourth moment:

[0158]

[0159] Signal power is defined as S, and noise power is defined as N. Therefore, M2 can be given by the following formula:

[0160] M2 = S + N (29)

[0161] Furthermore, M4 can be given by the following formula:

[0162] M4 = k x S 2 +6SN+k v N 2 (30)

[0163] Where k x and k v These are signal kurtosis and noise kurtosis, respectively.

[0164] Solving equations (29) and (30) for S and N, we obtain:

[0165]

[0166]

[0167] form and The estimator of the ratio is denoted as the M²M⁴ estimator. For a zero-mean sinusoidal signal, k... x =1.5, for noise, k v =3, therefore:

[0168]

[0169] In practical applications, the second and fourth moments are estimated from the time average of the measured signal:

[0170]

[0171]

[0172] The measured signal y k The estimated SNR is:

[0173]

[0174] Considering the SNR estimation bias, an additional term Δ is introduced into equation (26) as follows to make the threshold adjustment more reliable:

[0175]

[0176] As mentioned above, the threshold r e Adaptive adjustment can be made using equations (33)-(37). In equation (37), based on experimental testing, the present invention sets the additional term Δ to 2.5.

[0177] In step S8, since EKF is an iterative algorithm, the deviation between the estimated value and the true value may be relatively large during the convergence process, which may satisfy equation (8) and lead to continuous resets. To avoid unnecessary resets, each reset requires a time interval T. Δ The next reset can only be allowed after the algorithm converges.

[0178] In step S9, it is assumed that the SSO component undergoes nonlinear changes such as frequency variations, causing the EKF to lose its tracking capability. At this time, the error covariance matrix P k It needs to be reset to nI 4×4 (n=100), and the initial frequency value f in EKF S0 It needs to be updated to time (k–1)T s The estimated frequency.

[0179] Figures 2 to 4 and Figures 5 to 7 Two case studies are presented to demonstrate the effectiveness and accuracy of the proposed process. The artificially constructed non-stationary SSO signal is expressed as equation (32):

[0180]

[0181] In the formula, n(t) represents white noise. The entire simulation lasts 20 seconds and is divided into four time periods. In time period 1 (0-5s), the signal frequency is 10Hz, the amplitude is 10, and the attenuation coefficient is -0.05. In time period 2 (5-10s), the frequency changes to 15Hz, while the amplitude remains constant at 12.8. In time periods 3 (10-15s) and 4 (15-20s), the frequencies change to 20Hz and 15Hz respectively, while the amplitude remains constant at 12.8.

[0182] In scenario 1, the signal-to-noise ratio is set to 30dB, the sampling frequency is 1000Hz, and T... Δ The time was set to 0.3s, the window length m was 300, and a simulation was performed using the Monte Carlo method 100 times. The results are as follows. Figures 2 to 4 As shown.

[0183] In scenario 2, the signal-to-noise ratio is set to 10dB, the sampling frequency is 1000Hz, and T... Δ The time was set to 0.3s, the window length m was 300, and a simulation was performed using the Monte Carlo method 100 times. The results are as follows. Figures 5 to 7 As shown.

[0184] It can be seen that the method used in this invention can effectively detect noisy, non-stationary SSO signals.

[0185] This application discloses a non-stationary subsynchronous oscillation signal detection system, which includes the following modules:

[0186] Signal processing module: initializes the signal data and inputs the initialized signal data into the fitting module;

[0187] Fitting module: Calculates signal parameters and reconstructs the signal from the initialized signal data using a four-state EKF, and outputs the fitted SSO signal at the current time.

[0188] Calculation module: used to calculate the residual ratio e ratio The signal-to-noise ratio of the signal is calculated using an M2M4 estimator, and an adaptive reset threshold r is established. e ;

[0189] Self-reset judgment module: based on e ratio and r e Determine whether to reset the error covariance matrix P k and initial value of frequency;

[0190] When e is satisfied ratio Greater than r e At that time, and the time interval between resets is greater than the time interval T. Δ If so, a reset is performed, and the error covariance matrix P in the four-state EKF is changed. k Reset to nI 4×4The system updates the initial frequency value and then inputs the initialized signal data back into the signal processing module for processing; otherwise, it directly inputs the initialized signal data into the signal processing module for processing.

[0191] This application also discloses a terminal device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it employs any of the non-stationary subsynchronous oscillation signal detection methods described in the above embodiments.

[0192] The terminal device can be a computer device such as a desktop computer, a laptop computer, or a cloud server. The terminal device includes, but is not limited to, a processor and a memory. For example, the terminal device may also include input / output devices, network access devices, and buses.

[0193] The processor can be a central processing unit (CPU). Of course, depending on the actual use, it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it.

[0194] The memory can be an internal storage unit of the terminal device, such as a hard disk or RAM of the terminal device, or an external storage device of the terminal device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD), or flash memory card (FC) equipped on the terminal device. Furthermore, the memory can be a combination of internal storage units and external storage devices of the terminal device. The memory is used to store computer programs and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or will be output. This application does not limit this.

[0195] In this terminal device, any one of X in the above embodiments is stored in the memory of the terminal device and loaded and executed on the processor of the terminal device for convenient use.

[0196] This application also discloses a computer-readable storage medium, which stores a computer program, wherein when the computer program is executed by a processor, it employs any of the non-stationary subsynchronous oscillation signal detection methods described in the above embodiments.

[0197] The computer program can be stored in a computer-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The computer-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the computer-readable medium includes, but is not limited to, the above-mentioned components.

[0198] In this computer-readable storage medium, any one of the non-stationary subsynchronous oscillation signal detection methods in the above embodiments is stored in the computer-readable storage medium and loaded and executed on the processor to facilitate the storage and application of the above methods.

[0199] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for detecting non-stationary subsynchronous oscillation signals, characterized in that, include: The signal data is preprocessed, and the signal parameters are calculated and the signal is reconstructed using a four-state EKF to obtain the fitted SSO signal at the current time. The residual ratio is then calculated. e ratio : In the formula e k for kT s Time residuals This is an estimate of the SSO signal; use M 2 M 4. The estimator calculates the signal-to-noise ratio of the signal and establishes an adaptive reset threshold. r e : (20) In the formula, SNR is the signal-to-noise ratio; When the following conditions are met e ratio Greater than r e When, the error covariance matrix in the four-state EKF is... P k Reset to nI 4×4 And update the initial frequency value; use M 2 M 4. The estimator calculates the signal-to-noise ratio of the signal and establishes an adaptive reset threshold, including the following steps: make M 2 indicates y k The second moment: (21) make M 4 represents y k The fourth moment: (22) Signal power is defined as S Noise power is defined as N ; M 2 can be given by the following formula: (23) and, M 4 can be given by the following formula: (24) in k x and k v These are signal kurtosis and noise kurtosis, respectively. Solving equations (23) and (24) for S and N, we obtain: (25) (26) form and The estimator of the ratio is expressed as M 2 M 4. Estimators; for zero-mean sinusoidal signals, k x =1.5, for noise, k v =3, therefore: (27) The second and fourth moments are estimated by the time average of the measured signal, satisfying equations (28) and (29): (28) (29) Measured signal y k The estimated SNR is: (30) Considering the SNR estimation bias, an additional term Δ is introduced into equation (20) as follows to make the threshold adjustment more reliable: (31) As mentioned above, threshold r e Adaptive adjustments can be made by using equations (27) to (31).

2. The method for detecting non-stationary subsynchronous oscillation signals according to claim 1, characterized in that, Residual ratio e ratio Based on residuals e k Perform the calculation: kT s Time residuals e k It is expressed as follows: (5) in, y k It is a measurement; e k The variance can be approximated by the following formula: (6) wherein, m is the window length, if 1≤ k <m, then m is replaced by k ; the variance of the estimated SSO signal can be calculated by the following formula: (7) Then the residual ratio e ratio satisfy: .

3. The method for detecting non-stationary subsynchronous oscillation signals according to claim 2, characterized in that, The use M 2 M 4. The estimator calculates the signal-to-noise ratio of the signal and establishes an adaptive reset threshold. r e Specifically, the following steps are included: After the four-state EKF calculation, if equations (9) to (12) are satisfied, then... e ratio It can be approximated as equation (13): (9) (10) (11) (12) (13) Then the measured value y k The signal-to-noise ratio (SNR) can be calculated using the following formula: (14) According to equations (13) and (14), e ratio This can be further expressed as: (15) When the filter diverges, the fitted SSO signal will deviate significantly from the true value and satisfy... Therefore, the following inequalities hold: (16) (17) (18) according to e ratio Greater than r e Given the conditions, equations (15), (16), and (18), the following inequalities hold: (19) Combining equations (15) and (19), the threshold can be calculated using the following formula: (20) According to equation (20), it is possible to measure y k Determining the threshold when calculating SNR r e .

4. The method for detecting non-stationary subsynchronous oscillation signals according to claim 1, characterized in that, When the following conditions are met e ratio Greater than r e At that time, and the time interval between resets is greater than the time interval. T Δ Only then is a reset performed, and the error covariance matrix in the four-state EKF is changed. P k Reset to nI 4×4 Furthermore, the initial frequency value is updated, the initialized signal data is re-initialized, and the signal parameters are calculated and the signal is reconstructed using the four-state EKF to obtain the SSO signal fitted at the current time. Otherwise, the initialized signal data is directly re-initialized and the signal parameters are calculated and the signal is reconstructed using a four-state EKF to obtain the SSO signal fitted at the current time.

5. The method for detecting non-stationary subsynchronous oscillation signals according to claim 1, characterized in that, The additional Δ is set to 2.

5.

6. The method for detecting non-stationary subsynchronous oscillation signals according to claim 1, characterized in that... The SSO signal and its parameters can be obtained using equations (1) to (4): (1) (2) (3) (4) In the formula It is the SSO signal fitted by EKF using the state vector. This is an estimate of the SSO signal. x 1( k ), x 2( k ), x 3( k )and x 4( k ) are the four state variables of EKF; f S0 This is the initial value for the frequency.

7. A detection system for implementing the non-stationary subsynchronous oscillation signal detection method according to any one of claims 1-6, characterized in that, Includes the following modules: Signal preprocessing module: preprocesses the signal data and inputs the preprocessed signal data into the fitting module; Fitting module: Calculates signal parameters and reconstructs the signal from the initialized signal data using a four-state EKF, and outputs the fitted SSO signal at the current time. Calculation module: used to calculate the residual ratio e ratio and through M 2 M 4. The estimator calculates the signal-to-noise ratio of the signal and establishes an adaptive reset threshold. r e ; Self-reset judgment module: based on e ratio and r e Determine whether to reset the error covariance matrix. P k and initial value of frequency; When the following conditions are met e ratio Greater than r e At that time, and the time interval between resets is greater than the time interval. T Δ If so, a reset is performed, and the error covariance matrix in the four-state EKF is changed. P k Reset to nI 4×4 It also updates the initial frequency value and inputs the initialized signal data back into the signal processing module for processing. Otherwise, the initialized signal data is directly input into the signal processing module for processing.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, The memory stores a computer program that can run on the processor. When the processor loads and executes the computer program, it employs a non-stationary subsynchronous oscillation signal detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is loaded and executed by the processor, it employs a non-stationary subsynchronous oscillation signal detection method according to any one of claims 1 to 6.

Images