A method, system, medium and device for identifying offshore wind power synchronous oscillation

Abstract

The present application relates to the field of power system stability control, and discloses a kind of offshore wind power synchronous oscillation identification method, system, medium and equipment, comprising: the measured data of wide area measurement system in offshore wind farm grid-connected point and flexible direct current station AC side is collected, and the signal to be analyzed in sub or super synchronous frequency band is obtained after preprocessing;The signal to be analyzed is decoupled in multiple time scales, and is adaptively decomposed into multiple corresponding independent sub or super synchronous oscillation modes of intrinsic mode function, and the instantaneous characteristics of each mode function are extracted;The instantaneous characteristics and priori knowledge of physical model are input into the data-physical double-drive fusion feature extraction framework constructed, and high-precision oscillation characteristic quantity is output;Based on high-precision oscillation characteristic quantity, the initial parameters of subspace identification are optimized, the key parameters of each oscillation mode are identified online, and the identification result is obtained;Based on the identification result, calculate the parameter identification residual error, and dynamically adjust the identification strategy through the parameter adaptive correction mechanism.

Description

Technical Field

[0001] This invention relates to the field of power system stability control technology, and in particular to a method, system, medium, and equipment for identifying synchronous oscillations in offshore wind power based on multi-timescale decoupling and dual-drive fusion. Background Technology

[0002] Offshore wind power, with its abundant resources, high power generation efficiency, and lack of reliance on land resources, has become an important direction for renewable energy development. However, due to the long distance between offshore wind farms and the onshore power grid, traditional AC transmission methods suffer from limitations in transmission distance and poor voltage stability. Flexible DC transmission based on voltage source converters (VSC-HVDC), with its advantages of independent active and reactive power control, adaptability to weak grids, and large transmission capacity, has become the mainstream technology for offshore wind power grid connection. However, the offshore wind power VSC-HVDC grid connection system is a complex dynamic system with strong nonlinearity, multiple inputs and outputs, and multiple time scale coupling, composed of a large number of power electronic devices. The interaction between the wind turbine converter control, the phase-locked loop control of the VSC-HVDC converter station, the DC voltage control, and the AC grid impedance can easily induce continuous power oscillations, i.e., subsynchronous or supersynchronous oscillations, in the subsynchronous frequency band (2–45Hz) and the supersynchronous frequency band (55–95Hz). Such oscillations are characterized by wide propagation range, slow decay rate, and high degree of harm: at best, they can cause overcurrent tripping of wind turbine converters and power fluctuations in flexible DC converter stations, affecting power quality; at worst, they can cause torsional vibration fatigue damage to wind turbine shafts, lockout of flexible DC converter stations, and even large-scale power outages. Numerous subsynchronous or supersynchronous oscillation accidents have occurred both domestically and internationally in offshore wind power systems connected to flexible DC grids, causing serious economic losses and social impacts, and severely threatening the safe and stable operation of power systems.

[0003] Rapid and accurate identification of key parameters (frequency, damping ratio, phase) of subsynchronous or supersynchronous oscillations is the prerequisite and foundation for achieving oscillation early warning, suppression, and control. Existing subsynchronous or supersynchronous oscillation identification technologies for offshore wind power cannot simultaneously meet the requirements of high precision, strong robustness, high real-time performance, and adaptability to changes in operating conditions, and cannot effectively cope with the complex oscillation characteristics of offshore wind power clusters connected to the grid via flexible DC transmission. Summary of the Invention

[0004] To address the aforementioned problems, the purpose of this invention is to provide a method, system, medium, and equipment for identifying synchronous oscillations in offshore wind power, which can achieve novel subsynchronous or supersynchronous oscillation identification by dynamically decoupling multiple time scales and deeply integrating data and physics, thus providing technical support for the safe and stable operation of power systems.

[0005] To achieve the above objectives, in a first aspect, the technical solution adopted by the present invention is as follows: a method for identifying synchronous oscillations in offshore wind power, comprising: collecting measured data from the grid connection point of an offshore wind farm and the AC side of a flexible DC converter station in a wide-area measurement system, and preprocessing the data to obtain a signal to be analyzed in the subsynchronous or supersynchronous frequency band; using an improved variational mode decomposition algorithm to dynamically decouple the signal to be analyzed across multiple time scales, adaptively decomposing it into several intrinsic mode functions corresponding to independent subsynchronous or supersynchronous oscillation modes, and extracting the instantaneous features of each mode function; inputting the instantaneous features and prior knowledge of the physical model into a constructed data-physical dual-drive fusion feature extraction framework, and outputting high-precision oscillation feature quantities; optimizing the initial parameters for subspace identification based on the high-precision oscillation feature quantities, identifying the key parameters of each oscillation mode online through a recursive subspace identification algorithm, and obtaining identification results; calculating the parameter identification residual based on the identification results, and dynamically adjusting the identification strategy through a parameter adaptive correction mechanism.

[0006] Furthermore, the extraction of instantaneous features of each modal function includes: The initial number of modes K is adaptively determined based on the power spectral density of the signal to be analyzed, and the initialization penalty factor and convergence accuracy are set to construct a variational problem constraint model. Solving the variational problem constraint model yields the initial intrinsic mode functions, which are then used to introduce the mode overlap index. Valid intrinsic mode functions with instantaneous amplitudes greater than preset values ​​are selected, and noise and spurious modes are eliminated. For each effective intrinsic mode function, a Hilbert transform is performed to calculate the instantaneous characteristics, which include instantaneous amplitude, instantaneous frequency, and instantaneous phase.

[0007] Furthermore, the variational problem constraint model is as follows: ; The constraints are: ; in, Let be the component of the k-th order intrinsic mode function; The center frequency corresponding to the k-th intrinsic mode function; It is the first-order partial derivative with respect to time t, used to characterize the modal time-domain change rate; is the unit impulse function; j is the imaginary unit; t is the continuous-time variable; Pi is a constant. This is the convolution operator; Let be the instantaneous signal value of the k-th natural mode function at time t; It is the rotation factor in the complex frequency domain; The signal to be analyzed is denoted as ; K represents the initial number of modes.

[0008] Furthermore, the modal overlap index for: ; in, and These are the bandwidths of the i-th and j-th modes, respectively; These are the center frequencies corresponding to the i-th and j-th intrinsic mode functions, respectively.

[0009] Furthermore, the data-physical dual-drive fusion feature extraction framework includes a physical drive module, a data drive module, and a feature fusion module; The physical drive module is used to construct the multi-input multi-output impedance matrix on the offshore wind farm side and the flexible DC converter station side. The multi-input multi-output impedance matrix is ​​reduced to the single-input single-output equivalent positive sequence impedance through the Schur complement transformation. For the subsynchronous or supersynchronous frequency band, the phase-locked loop control link coefficient matrix and bridge arm dynamic model affected by the power operation level are introduced. The prior feature library containing the frequency range, damping ratio range and phase relationship of each typical mode is generated by traversing multiple power operation levels. The data-driven module is used to construct an adaptive sliding window long short-term memory network, which automatically learns the nonlinear dynamic characteristics of the oscillation signal using the effective intrinsic mode function as input; an attention mechanism is introduced to weight and fuse feature vectors at different time scales to output a preliminary feature vector. The feature fusion module calculates the cosine similarity between the initial feature vector and the feature vectors in the prior feature library, and uses the parameter range corresponding to the feature vector with the highest matching degree as the initial constraint; it then uses a Bayesian fusion algorithm to combine the prior probability of the physical model with the likelihood probability of the measured data to correct and obtain a high-precision oscillation feature quantity.

[0010] Furthermore, an improved particle swarm optimization algorithm is used to optimize the initial parameters used for subspace identification, including: Initialize particle swarm parameters by using chaotic mapping to initialize particle position and velocity; Introducing adaptive inertia weights ,in, Let T be the current iteration number, and T be the maximum number of iterations for the particle swarm optimization algorithm. For the pre-set maximum weight, The minimum weight is set in advance; The formulas for updating particle velocity and position are as follows: ; ; in, As a learning factor, A random number in the interval [0,1]. For the optimal position of an individual, To be the globally optimal position For the updated particle velocity, The particle velocity before the update. The particle positions before the update. The updated particle positions; Add Gaussian mutation operation: ;in, These are random numbers distributed according to a standard normal distribution. The fitness function is defined as the mean square error between the identification output and the measured signal, and the formula is as follows: ; in, For the measured signal at the k-th sampling point, To simulate the oscillating response signal, N is the total number of particles.

[0011] Furthermore, based on the identification results, parameter identification residuals are calculated, and the identification strategy is dynamically adjusted through a parameter adaptive correction mechanism, including: The parameters identify residuals , For the measured signal at the k-th sampling point, To simulate the oscillation response signal; When the parameters identify residuals When the value exceeds the first set threshold and continues to exceed 5 sampling points, it is determined that the system operating condition has changed abruptly, and the improved particle swarm optimization algorithm is automatically triggered to re-optimize and update the initial parameters of subspace identification. When the damping ratio of the identified key parameters is less than the second set threshold, a first-level instability warning is issued; when the damping ratio is less than 0, a second-level emergency warning is issued, and the identified parameters are uploaded to the wideband oscillation control system.

[0012] Secondly, the technical solution adopted by this invention is as follows: a system for identifying synchronous oscillations in offshore wind power, comprising: a data acquisition and preprocessing module, which acquires measured data from the grid connection point of the offshore wind farm and the AC side of the flexible DC converter station in a wide-area measurement system, and performs preprocessing to obtain the signal to be analyzed in the subsynchronous or supersynchronous frequency band; a multi-timescale decoupling module, which uses an improved variational mode decomposition algorithm to dynamically decouple the signal to be analyzed in multiple timescales, adaptively decomposes it into several intrinsic mode functions corresponding to independent subsynchronous or supersynchronous oscillation modes, and extracts the instantaneous features of each mode function; a data-physics dual-drive feature extraction module, which inputs the instantaneous features and prior knowledge of the physical model into the constructed data-physics dual-drive fusion feature extraction framework, and outputs high-precision oscillation feature quantities; a dynamic parameter identification module, which optimizes the initial parameters used for subspace identification based on the high-precision oscillation feature quantities, identifies the key parameters of each oscillation mode online through a recursive subspace identification algorithm, and obtains the identification results; and an adaptive correction module, which calculates the parameter identification residual based on the identification results, and dynamically adjusts the identification strategy through a parameter adaptive correction mechanism.

[0013] Thirdly, the technical solution adopted by the present invention is: a computer-readable storage medium for storing one or more programs, wherein the one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform any of the methods described above.

[0014] Fourthly, the technical solution adopted by the present invention is: a computing device comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described above.

[0015] The present invention has the following advantages due to the adoption of the above technical solutions: 1. This invention employs an improved variational mode decomposition algorithm based on adaptive power spectral density order determination and mode overlap screening, which achieves accurate decoupling of sub- or supersynchronous oscillations of offshore wind power across multiple time scales. This effectively solves the problems of mode aliasing and spurious modes in traditional signal decomposition methods, and significantly improves the accuracy of oscillation feature extraction.

[0016] 2. This invention constructs a data-physics dual-driven deep fusion feature extraction framework, which organically combines the prior knowledge of the physical model with the nonlinear features of the measured data through the Bayesian fusion algorithm. This not only overcomes the shortcomings of the pure physical model, such as strong parameter dependence and high computational complexity, but also solves the problems of poor robustness and lack of physical meaning in the results of the pure data-driven method.

[0017] 3. This invention employs an improved particle swarm optimization algorithm to optimize the initial parameters for identifying the recursive subspace, significantly improving the convergence speed and accuracy of the identification algorithm. Simultaneously, it introduces a parameter adaptive correction mechanism, which can dynamically adapt to changes in system operating conditions, achieving online high-precision identification of oscillation parameters under complex operating conditions. This provides reliable technical support for oscillation early warning and suppression in offshore wind power systems connected to the flexible DC grid. Attached Figure Description

[0018] Figure 1 This is a flowchart of the offshore wind power synchronous oscillation identification method in an embodiment of the present invention. Detailed Implementation

[0019] To address the limitations of existing offshore wind power subsynchronous or supersynchronous oscillation identification technologies in simultaneously achieving high precision, robustness, real-time performance, and adaptability, and thus failing to effectively handle the complex oscillation characteristics of offshore wind power flexible DC-DC grid-connected systems, this invention provides a method, system, medium, and equipment for identifying synchronous oscillations in offshore wind power. Considering the high-dimensional nonlinear coupling characteristics of offshore wind farms and flexible DC-DC converter stations, this invention constructs an adaptive algorithm capable of extracting oscillation characteristics from measured data from a Wide Area Measurement System (WAMS). By introducing improved particle swarm optimization and subspace identification theory, it achieves millisecond-level online accurate tracking and adaptive correction of key parameters such as subsynchronous or supersynchronous oscillation mode frequency, damping ratio, and phase, solving the identification lag and error accumulation problems caused by model mismatch in weak grid environments using traditional methods. This provides core decision-making basis for broadband oscillation emergency control, adaptive instability early warning, and active damping suppression of offshore wind power clusters, improving the grid connection stability and equipment safety level of deep-sea wind farms under extreme operating conditions.

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0021] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0022] In one embodiment of the present invention, a method for identifying synchronous oscillations in offshore wind power is provided. In this embodiment, as... Figure 1As shown, the method includes the following steps: 1) Collect measured data from the grid connection point of the offshore wind farm and the AC side of the flexible DC converter station in the Wide Area Measurement System (WAMS), and preprocess the data to obtain the signals to be analyzed in the subsynchronous or supersynchronous frequency bands; among which, the measured data include measured data of three-phase voltage, three-phase current and power.

[0023] 2) An improved variational mode decomposition algorithm is used to dynamically decouple the signal to be analyzed across multiple time scales, adaptively decompose it into several intrinsic mode functions corresponding to independent sub- or supersynchronous oscillation modes, and preliminarily extract the instantaneous features of each mode function.

[0024] 3) Input instantaneous features and prior knowledge of the physical model into the constructed data-physical dual-drive fusion feature extraction framework, and output high-precision oscillation feature quantities.

[0025] 4) The initial parameters for subspace identification are optimized based on high-precision oscillation characteristic quantities. The key parameters of each oscillation mode are identified online by recursive subspace identification algorithm to obtain the identification results. The key parameters include the frequency, damping ratio and phase of each oscillation mode.

[0026] 5) Calculate the parameter identification residuals based on the identification results, and dynamically adjust the identification strategy through a parameter adaptive correction mechanism.

[0027] In step 1) above, the preprocessing of the measured data includes denoising and synchronous sampling to obtain the sub- or super-synchronous frequency band signal to be analyzed with a sampling frequency of 100Hz to 1000Hz. The denoising process adopts the wavelet threshold denoising method, the wavelet basis function is selected as db8 wavelet, and the decomposition layer is 4 layers.

[0028] In this embodiment, based on an actual offshore wind power base connected to the grid via a ±400kV / 1000MW flexible DC transmission system, the base includes 200 5MW permanent magnet direct-drive wind turbines with a total installed capacity of 1000MW. The flexible DC converter station adopts a modular multilevel converter (MMC) topology with a rated AC voltage of 220kV.

[0029] The measured data of three-phase voltage and three-phase current of the grid-connected bus and the AC side bus of the flexible DC converter station of the offshore wind farm are collected through the synchronous phasor measurement unit (PMU) of the wide area measurement system (WAMS). The sampling frequency is set to 200Hz (to meet the Nyquist sampling requirements of the sub- or super-synchronous frequency band).

[0030] The db8 wavelet thresholding denoising method was employed, with a decomposition layer of 4, to remove high-frequency noise and DC components. Time synchronization calibration was performed to ensure that the timestamp error of data from different measurement points was less than 1ms. The subsynchronous or supersynchronous frequency band signal of 2~95Hz was extracted through bandpass filtering to obtain the signal to be analyzed. .

[0031] In step 2) above, extracting the instantaneous features of each modal function includes the following steps: 2.1) The initial number of modes K is adaptively determined based on the power spectral density of the signal to be analyzed, and an initialization penalty factor is set. and convergence accuracy Construct a variational problem constraint model.

[0032] Specifically, the variational problem constraint model is as follows: ; The constraints are: ; in, These are the components of the k-th order intrinsic mode function, i.e., the single-frequency oscillation mode signal obtained by decomposition; The center frequency corresponding to the k-th intrinsic mode function; It is the first-order partial derivative with respect to time t, used to characterize the modal time-domain change rate; is the unit impulse function; j is the imaginary unit; t is the continuous-time variable; Pi is a constant. This is the convolution operator; Let be the instantaneous signal value of the k-th natural mode function at time t; is the complex frequency domain rotation factor.

[0033] 2.2) Solve the variational problem constraint model to obtain the initial intrinsic mode functions, and then introduce the mode overlap index. Valid intrinsic mode functions with instantaneous amplitudes greater than preset values ​​are selected, and noise and spurious modes are eliminated.

[0034] Specifically, modal overlap index for: ; in, and These are the bandwidths of the i-th and j-th modes, respectively; These are the center frequencies corresponding to the i-th and j-th intrinsic mode functions, respectively.

[0035] 2.3) Perform Hilbert transform on each effective intrinsic mode function to calculate the instantaneous characteristics, including the instantaneous amplitude. Instantaneous frequency and instantaneous phase .

[0036] ; ; ; ; in, Represents the Hilbert transform operator; For integration variables; It is the instantaneous phase.

[0037] In a specific embodiment of the present invention, the power spectral density of the signal to be analyzed is calculated, and the initial number of modes K=3 is adaptively determined based on the number of power spectral peaks. Initialize the penalty factor α = 2000, and the convergence accuracy ε = 1 × 10⁻⁵; A variational problem constraint model was constructed and solved to obtain three initial intrinsic mode functions (IMFs). Introducing the modal overlap index Set the overlap threshold to 0.2 to filter out Modes with a value >0.2 are merged to obtain two effective intrinsic mode functions, IMF1 and IMF2. Noise modes with instantaneous amplitude less than a preset amplitude threshold (0.05pu, where pu is a per-unit value) are removed; Perform Hilbert transforms on IMF1 and IMF2 respectively to calculate their instantaneous amplitudes. Instantaneous frequency and instantaneous phase .

[0038] In step 3) above, the data-physical dual-drive fusion feature extraction framework includes a physical-driven module, a data-driven module, and a feature fusion module. Wherein: The physics-driven module constructs the multi-input multi-output impedance matrix for both the offshore wind farm and the flexible DC converter station. It reduces the multi-input multi-output impedance matrix to a single-input single-output (SISO) equivalent positive-sequence impedance using Schur complement transformation. For sub- or super-synchronous frequency bands, it introduces the phase-locked loop (PLL) control element coefficient matrix and arm dynamic model, which are influenced by power operating levels. It then traverses multiple power operating levels to generate a priori feature library containing the frequency range, damping ratio range, and phase relationships of each typical mode. The single-input single-output equivalent positive-sequence impedance, PLL control element coefficient matrix, and arm dynamic model provide the modeling basis for the physics-driven module. Since MMC (Modular Multilevel Converter) is a high-order MIMO system, direct analysis is extremely difficult. Reducing it to a SISO equivalent positive-sequence impedance through Schur complement transformation significantly simplifies stability analysis. The PLL control element coefficient matrix and arm dynamic model are introduced because they are the main culprits causing sub- or super-synchronous oscillations and must be included in the prior knowledge base to ensure the physical model remains accurate.

[0039] The data-driven module is used to construct an adaptive sliding window long short-term memory network, which takes the effective intrinsic mode function as input and automatically learns the nonlinear dynamic characteristics of the oscillating signal; an attention mechanism is introduced to weight and fuse feature vectors at different time scales and output a preliminary feature vector.

[0040] The feature fusion module calculates the cosine similarity between the initial feature vector and the feature vectors in the prior feature library, and uses the parameter range corresponding to the feature vector with the highest matching degree as the initial constraint; the Bayesian fusion algorithm is used to combine the prior probability of the physical model and the likelihood probability of the measured data to correct and obtain the high-precision oscillatory feature quantity.

[0041] Among them, the prior feature library is used to store a set of parameter vectors for "frequency range, damping ratio range and phase relationship of typical modes"; the preliminary feature vector is also an estimate of frequency, damping, etc. Calculating the cosine similarity between the preliminary feature vector and the feature vector in the prior feature library is to find the set of operating condition parameters that are most similar to the historical data from the prior feature library.

[0042] In this embodiment, the formula for reducing the order of the Schur complement transform is as follows: ; in, This is the equivalent positive sequence impedance for a single-input single-output circuit. It is a positive-sequence self-impedance; Positive-sequence to negative-sequence coupling impedance; It is a negative sequence self-impedance; This is the negative-sequence to positive-sequence coupling impedance.

[0043] The adaptive sliding window length L of the short-term memory network is adjusted using the following formula: ; in, and These are the minimum and maximum window lengths, respectively. The standard deviation of the instantaneous frequency; This is the preset maximum frequency standard deviation.

[0044] The formula for the Bayesian fusion algorithm is as follows: ; in, The prior probabilities of the physical model follow a normal distribution. ; The likelihood probability of the measured data follows a normal distribution. ; This is the corrected high-precision oscillation characteristic.

[0045] In a specific embodiment of the present invention, a multiple-input multiple-output (MIMO) impedance matrix is ​​established on both the offshore wind farm side and the flexible DC converter station side. The MIMO impedance matrix is ​​then reduced to an equivalent positive-sequence single-input single-output (SISO) impedance using a Schur complement transformation. The coefficient matrix of the phase-locked loop control loop and the dynamic model of the MMC bridge arm, which are affected by the power operation level, are introduced. The system traverses the 0~100% rated power range to generate a priori feature library containing typical modal frequency ranges (5~30Hz, 60~90Hz), damping ratio ranges (-0.05~0.05), and phase relationships.

[0046] Construct an adaptive sliding window long short-term memory network (LSTM) and set a minimum window length L. min =100, maximum window length L max =500; based on the standard deviation of the current effective IMF instantaneous frequency. The sliding window length L is adaptively adjusted; the nonlinear dynamic characteristics of the oscillation signal are automatically learned using IMF1 and IMF2 as inputs; an attention mechanism is introduced to weight and fuse feature vectors at different time scales to output a preliminary feature vector.

[0047] The cosine similarity between the initial feature vector and each feature vector in the prior feature library is calculated. The parameter range corresponding to the feature vector with the highest matching degree is used as the initial constraint. Bayesian fusion is used to obtain high-precision oscillatory feature quantities.

[0048] In step 4) above, the improved particle swarm optimization algorithm is used to optimize the initial parameters for subspace identification, including the following steps: 4.1) Initialize the particle swarm parameters: number of particles N=30, maximum number of iterations T=50, and use chaotic mapping to initialize particle position and velocity.

[0049] 4.2) Introducing adaptive inertia weights ,in, Let T be the current iteration number, and T be the maximum number of iterations for the particle swarm optimization algorithm. For the pre-set maximum weight, The minimum weight is preset; in this embodiment, , .

[0050] The formulas for updating particle velocity and position are as follows: ; ; in, As the learning factor, in this embodiment, we take... ; A random number in the interval [0,1]. The optimal position for the individual; To be the globally optimal position For the updated particle velocity, The particle velocity before the update. The particle positions before the update. This represents the updated particle position.

[0051] 4.3) Incorporating Gaussian Mutation: When the fitness does not improve significantly after three consecutive generations, the globally optimal particle is mutated using the following formula: ; in, These are random numbers that follow a standard normal distribution.

[0052] In this embodiment, when the absolute value of the relative change rate of the fitness function value for three consecutive generations is less than 1%, it is determined that there is no significant improvement.

[0053] 4.4) The fitness function is defined as the mean square error between the identification output and the measured signal, and the formula is as follows: ; in, For the measured signal at the k-th sampling point, To simulate the oscillating response signal, N is the total number of particles.

[0054] In this embodiment, the identification step size of the recursive subspace identification algorithm is set to 10ms, and the system state matrix is ​​updated by recursive singular value decomposition to achieve millisecond-level online parameter tracking.

[0055] In step 5) above, the parameter identification residual is calculated based on the identification results, and the identification strategy is dynamically adjusted through a parameter adaptive correction mechanism, including the following steps: 5.1) Parameter Identification Residual , For the measured signal at the k-th sampling point, This is to simulate the oscillation response signal.

[0056] 5.2) When the parameters identify the residuals When the value exceeds the first set threshold and continues to exceed 5 sampling points, it is determined that the system operating condition has changed abruptly, and the improved particle swarm optimization algorithm is automatically triggered to re-optimize and update the initial parameters of subspace identification.

[0057] In this embodiment, the first set threshold is set to 0.02 pu. When the residual exceeds the first set threshold, the decomposition, feature extraction and identification are re-executed, and the window length is increased by 20% and the iteration step size is decreased by 10% to ensure stable online identification.

[0058] 5.3) When the damping ratio of the identified key parameters is less than the second set threshold, a first-level instability warning is issued; when the damping ratio is less than 0, a second-level emergency warning is issued, and the identified parameters are uploaded to the wideband oscillation control system.

[0059] In this embodiment, the second set threshold is 0.01.

[0060] In one embodiment of the present invention, a system for identifying synchronous oscillations in offshore wind power is provided, comprising: The data acquisition and preprocessing module acquires measured data from the grid connection point of the offshore wind farm and the AC side of the flexible DC converter station in the wide-area measurement system, and performs preprocessing to obtain the signal to be analyzed in the sub- or supersynchronous frequency band. The multi-timescale decoupling module uses an improved variational mode decomposition algorithm to dynamically decouple the signal to be analyzed across multiple timescales, adaptively decomposing it into several intrinsic mode functions corresponding to independent sub- or supersynchronous oscillation modes, and extracting the instantaneous features of each mode function; The data-physics dual-drive feature extraction module takes instantaneous features and prior knowledge of the physical model as inputs to the constructed data-physics dual-drive fusion feature extraction framework, and outputs high-precision oscillation feature quantities. The dynamic parameter identification module optimizes the initial parameters for subspace identification based on high-precision oscillation feature quantities, and identifies the key parameters of each oscillation mode online through a recursive subspace identification algorithm to obtain the identification results; The adaptive correction module calculates the parameter identification residual based on the identification results and dynamically adjusts the identification strategy through a parameter adaptive correction mechanism.

[0061] In the above embodiments, the instantaneous features of each modal function are extracted, including: The initial number of modes K is adaptively determined based on the power spectral density of the signal to be analyzed, and the initialization penalty factor and convergence accuracy are set to construct a variational problem constraint model. Solving the variational problem constraint model yields the initial intrinsic mode functions, which are then used to introduce the mode overlap index. Valid intrinsic mode functions with instantaneous amplitudes greater than preset values ​​are selected, and noise and spurious modes are eliminated. For each effective intrinsic mode function, a Hilbert transform is performed to calculate the instantaneous characteristics, which include instantaneous amplitude, instantaneous frequency, and instantaneous phase.

[0062] In this embodiment, the variational problem constraint model is as follows:

[0063] The constraints are: ; in, Let be the component of the k-th order intrinsic mode function; The center frequency corresponding to the k-th intrinsic mode function; It is the first-order partial derivative with respect to time t, used to characterize the modal time-domain change rate; is the unit impulse function; j is the imaginary unit; t is the continuous-time variable; Pi is a constant. This is the convolution operator; Let be the instantaneous signal value of the k-th natural mode function at time t; is the complex frequency domain rotation factor.

[0064] In this embodiment, the modal overlap index for: ; in, and These are the bandwidths of the i-th and j-th modes, respectively; These are the center frequencies corresponding to the i-th and j-th intrinsic mode functions, respectively.

[0065] In the above embodiments, the data-physical dual-drive fusion feature extraction framework includes a physical drive module, a data drive module, and a feature fusion module; The physics-driven module is used to construct the multi-input multi-output impedance matrix on the offshore wind farm side and the flexible DC converter station side. The multi-input multi-output impedance matrix is ​​reduced to the single-input single-output equivalent positive sequence impedance through the Schur complement transformation. For the subsynchronous or supersynchronous frequency band, the phase-locked loop control link coefficient matrix and bridge arm dynamic model affected by the power operation level are introduced. The prior feature library containing the frequency range, damping ratio range and phase relationship of each typical mode is generated by traversing multiple power operation levels. The data-driven module is used to construct an adaptive sliding window long short-term memory network, which automatically learns the nonlinear dynamic characteristics of oscillating signals by taking the effective intrinsic mode function as input; an attention mechanism is introduced to weight and fuse feature vectors at different time scales to output a preliminary feature vector. The feature fusion module calculates the cosine similarity between the initial feature vector and the feature vectors in the prior feature library, and uses the parameter range corresponding to the feature vector with the highest matching degree as the initial constraint; the Bayesian fusion algorithm is used to combine the prior probability of the physical model and the likelihood probability of the measured data to correct and obtain the high-precision oscillatory feature quantity.

[0066] In the above embodiments, an improved particle swarm optimization algorithm is used to optimize the initial parameters for subspace identification, including: Initialize particle swarm parameters by using chaotic mapping to initialize particle position and velocity; Introducing adaptive inertia weights: ,in, The current iteration number is T, where T is the maximum number of iterations for the particle swarm optimization algorithm. The formulas for updating particle velocity and position are as follows: ; ; in, As a learning factor, A random number in the interval [0,1]. For the optimal position of an individual, The globally optimal position; Add Gaussian mutation operation: ;in, These are random numbers distributed according to a standard normal distribution. The fitness function is defined as the mean square error between the identification output and the measured signal, and the formula is as follows: ; in, For the measured signal at the k-th sampling point, This is to simulate the oscillation response signal.

[0067] In the above embodiments, the parameter identification residual is calculated based on the identification result, and the identification strategy is dynamically adjusted through a parameter adaptive correction mechanism, including: Parameter identification residuals , For the measured signal at the k-th sampling point, To simulate the oscillation response signal; When parameters identify residuals When the value exceeds the first set threshold and continues to exceed 5 sampling points, it is determined that the system operating condition has changed abruptly, and the improved particle swarm optimization algorithm is automatically triggered to re-optimize and update the initial parameters of subspace identification. When the damping ratio of the identified key parameters is less than the second set threshold, a first-level instability warning is issued; when the damping ratio is less than 0, a second-level emergency warning is issued, and the identified parameters are uploaded to the wideband oscillation control system.

[0068] The system provided in this embodiment is used to execute the above-described method embodiments. For specific processes and details, please refer to the above embodiments, which will not be repeated here.

[0069] In one embodiment of the present invention, a computing device is provided, which may be a terminal and may include: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors, and the one or more programs include instructions for performing any of the methods in the above embodiments.

[0070] Furthermore, when the aforementioned instructions can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] In one embodiment of the present invention, a computer program product is provided, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to perform the methods provided in the above-described method embodiments.

[0072] In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided, which stores server instructions that cause a computer to perform the methods provided in the above embodiments.

[0073] The computer-readable storage medium provided in the above embodiments has a similar implementation principle and technical effect to the above method embodiments, and will not be described again here.

[0074] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for identifying synchronous oscillations in offshore wind power, characterized in that, include: Collect measured data from the grid connection point of offshore wind farms and the AC side of flexible DC converter stations in the wide-area measurement system, and perform preprocessing to obtain the signals to be analyzed in the subsynchronous or supersynchronous frequency bands; An improved variational mode decomposition algorithm is used to dynamically decouple the signal to be analyzed across multiple time scales, adaptively decompose it into several intrinsic mode functions corresponding to independent sub- or supersynchronous oscillation modes, and extract the instantaneous features of each mode function. The instantaneous features and prior knowledge of the physical model are input into the constructed data-physics dual-drive fusion feature extraction framework, which outputs high-precision oscillation feature quantities. Based on the high-precision oscillation feature quantity, the initial parameters for subspace identification are optimized, and the key parameters of each oscillation mode are identified online by the recursive subspace identification algorithm to obtain the identification results; Based on the identification results, the parameter identification residual is calculated, and the identification strategy is dynamically adjusted through a parameter adaptive correction mechanism.

2. The offshore wind power synchronous oscillation identification method as described in claim 1, characterized in that, The extraction of instantaneous features of each modal function includes: The initial number of modes K is adaptively determined based on the power spectral density of the signal to be analyzed, and the initialization penalty factor and convergence accuracy are set to construct a variational problem constraint model. Solving the variational problem constraint model yields the initial intrinsic mode functions, which are then used to introduce the mode overlap index. Valid intrinsic mode functions with instantaneous amplitudes greater than preset values ​​are selected, and noise and spurious modes are eliminated. For each effective intrinsic mode function, a Hilbert transform is performed to calculate the instantaneous characteristics, which include instantaneous amplitude, instantaneous frequency, and instantaneous phase.

3. The offshore wind power synchronous oscillation identification method as described in claim 2, characterized in that, The variational problem constraint model is as follows: ； The constraints are: ; in, Let be the component of the k-th order intrinsic mode function; The center frequency corresponding to the k-th intrinsic mode function; It is the first-order partial derivative with respect to time t, used to characterize the modal time-domain change rate; is the unit impulse function; j is the imaginary unit; t is the continuous-time variable; Pi is a constant. This is the convolution operator; Let be the instantaneous signal value of the k-th natural mode function at time t; It is the rotation factor in the complex frequency domain; The signal to be analyzed is denoted as ; K represents the initial number of modes.

4. The offshore wind power synchronous oscillation identification method as described in claim 2, characterized in that, The modal overlap index for: ; in, and These are the bandwidths of the i-th and j-th modes, respectively; These are the center frequencies corresponding to the i-th and j-th intrinsic mode functions, respectively.

5. The offshore wind power synchronous oscillation identification method as described in claim 1, characterized in that, The data-physical dual-drive fusion feature extraction framework includes a physical drive module, a data drive module, and a feature fusion module; The physical drive module is used to construct the multi-input multi-output impedance matrix on the offshore wind farm side and the flexible DC converter station side. The multi-input multi-output impedance matrix is ​​reduced to the single-input single-output equivalent positive sequence impedance through the Schur complement transformation. For the subsynchronous or supersynchronous frequency band, the phase-locked loop control link coefficient matrix and bridge arm dynamic model affected by the power operation level are introduced. The prior feature library containing the frequency range, damping ratio range and phase relationship of each typical mode is generated by traversing multiple power operation levels. The data-driven module is used to construct an adaptive sliding window long short-term memory network, which automatically learns the nonlinear dynamic characteristics of oscillation signals by taking the effective intrinsic mode function as input. An attention mechanism is introduced to weight and fuse feature vectors at different time scales, and a preliminary feature vector is output. The feature fusion module calculates the cosine similarity between the initial feature vector and the feature vectors in the prior feature library, and uses the parameter range corresponding to the feature vector with the highest matching degree as the initial constraint; it then uses a Bayesian fusion algorithm to combine the prior probability of the physical model with the likelihood probability of the measured data to correct and obtain a high-precision oscillation feature quantity.

6. The offshore wind power synchronous oscillation identification method as described in claim 1, characterized in that, An improved particle swarm optimization algorithm is used to optimize the initial parameters used for subspace identification, including: Initialize particle swarm parameters by using chaotic mapping to initialize particle position and velocity; Introducing adaptive inertia weights ,in, Let T be the current iteration number, and T be the maximum number of iterations for the particle swarm optimization algorithm. For the pre-set maximum weight, The minimum weight is set in advance; The formulas for updating particle velocity and position are as follows: ； ； in, As a learning factor, A random number in the interval [0,1]. For the optimal position of an individual, To be the globally optimal position For the updated particle velocity, The particle velocity before the update. The particle positions before the update. The updated particle positions; Add Gaussian mutation operation: ;in, These are random numbers distributed according to a standard normal distribution. The fitness function is defined as the mean square error between the identification output and the measured signal, and the formula is as follows: ； in, For the measured signal at the k-th sampling point, To simulate the oscillating response signal, N is the total number of particles.

7. The offshore wind power synchronous oscillation identification method as described in claim 1, characterized in that, Based on the identification results, the parameter identification residual is calculated, and the identification strategy is dynamically adjusted through a parameter adaptive correction mechanism, including: The parameters identify residuals , For the measured signal at the k-th sampling point, To simulate the oscillation response signal; When the parameters identify residuals When the value exceeds the first set threshold and continues to exceed 5 sampling points, it is determined that the system operating condition has changed abruptly, and the improved particle swarm optimization algorithm is automatically triggered to re-optimize and update the initial parameters of subspace identification. When the damping ratio of the identified key parameters is less than the second set threshold, a first-level instability warning is issued; when the damping ratio is less than 0, a second-level emergency warning is issued, and the identified parameters are uploaded to the wideband oscillation control system.

8. A system for identifying synchronous oscillations in offshore wind power, characterized in that, include: The data acquisition and preprocessing module acquires measured data from the grid connection point of the offshore wind farm and the AC side of the flexible DC converter station in the wide-area measurement system, and performs preprocessing to obtain the signal to be analyzed in the sub- or supersynchronous frequency band. The multi-timescale decoupling module uses an improved variational mode decomposition algorithm to dynamically decouple the signal to be analyzed across multiple timescales, adaptively decomposing it into several intrinsic mode functions corresponding to independent sub- or supersynchronous oscillation modes, and extracting the instantaneous features of each mode function; The data-physics dual-drive feature extraction module inputs the instantaneous features and prior knowledge of the physical model into the constructed data-physics dual-drive fusion feature extraction framework, and outputs high-precision oscillation feature quantities. The dynamic parameter identification module optimizes the initial parameters for subspace identification based on the high-precision oscillation feature quantity, and identifies the key parameters of each oscillation mode online through the recursive subspace identification algorithm to obtain the identification results; The adaptive correction module calculates the parameter identification residual based on the identification results and dynamically adjusts the identification strategy through a parameter adaptive correction mechanism.

9. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods described in claims 1 to 7.

10. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods described in claims 1 to 7.

Images