A data-driven wind farm subsynchronous oscillation dynamic monitoring method and system

Abstract

The application relates to the field of new power system stability, and discloses a data-driven wind farm subsynchronous oscillation dynamic monitoring method and system, and relates to the field of new power system stability, which comprises the following steps: based on energy angle analysis of system stability characteristics, defining characteristic quantities capable of quantifying the oscillation stability level of a wind farm grid-connected system; adopting a singular spectrum analysis algorithm to realize improvement of an adaptive variational modal decomposition method, accurately extracting time-frequency signals of an oscillation mode, realizing rapid identification of oscillation stability; applying a subjective and objective comprehensive evaluation method, establishing an evaluation model, and quantitatively evaluating the influence of relevant influence factors on the energy characteristics of a dominant subsynchronous oscillation mode. The subsynchronous oscillation stability quantification evaluation method provided by the application has high practicability, can evaluate system stability only by using measurement data, and can analyze the dominant oscillation mode characteristics of the system, thereby having certain guiding significance for subsynchronous oscillation stability quantification evaluation, disturbance positioning and oscillation suppression in actual engineering.

Description

Technical Field

[0001] This invention relates to the field of novel power system stability, specifically to a data-driven method and system for dynamic monitoring of subsynchronous oscillations in wind farms. Background Technology

[0002] To accelerate the achievement of "dual carbon" targets, the grid-connected capacity of renewable energy is continuously increasing, and power electronic equipment is being widely used. This has led to a decrease in power system inertia and immunity, making power oscillations increasingly prominent. Among these, subsynchronous oscillation (SSO) incidents caused by wind power grid connection occur frequently, seriously threatening the safe and stable operation of the power grid. Accurately identifying oscillation parameters and studying their SSO influencing factors helps to quickly assess system stability, optimize system parameters, and thus effectively suppress oscillation propagation.

[0003] Existing analyses of SSO characteristics are mostly based on eigenvalue analysis or impedance analysis, whose accuracy depends on accurate system modeling. This approach is increasingly ill-suited to the current complexities of large power grids, which feature a wide variety of components, difficulties in parameter acquisition, and complex and variable operating states. Furthermore, it fails to capture information on the damping characteristics of local components. The rapid development of measurement and data mining technologies has provided new avenues for system stability research. Some researchers have constructed energy functions using only port measurement data to extract characteristic quantities that characterize SSO stability, achieving online damping assessment and oscillation source tracing from an energy perspective, demonstrating strong practicality. Therefore, there is an urgent need to extract SSO stability characteristic quantities from response data to quickly analyze oscillation characteristics under various scenarios. Oscillation detection methods based on measurement data mainly include Fourier transform, Prony analysis, mode decomposition, and state estimation algorithms. To overcome the shortcomings of these single analysis methods and improve the accuracy of oscillation detection, current approaches often involve improving and integrating multiple algorithms or combining them with intelligent algorithms to achieve rapid parameter identification.

[0004] As the oscillation signals of new power systems exhibit increasingly prominent wide-frequency domain, multi-mode, nonlinear, and time-varying characteristics, existing oscillation detection technologies are less practical and require further improvement and refinement to accurately achieve rapid dynamic monitoring of multiple oscillation modes. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a data-driven method and system for dynamic monitoring of subsynchronous oscillations in wind farms, thereby solving the aforementioned problems.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions.

[0007] A data-driven method for dynamic monitoring of subsynchronous oscillations in wind farms includes the following steps: S1: Collect voltage and current data at the grid connection point of the wind farm, and perform preprocessing and initial status assessment of the data; S2: Identify oscillation mode parameters and determine the dominant mode; S3: The dissipation intensity corresponding to the oscillation energy under this working condition is calculated, which is the stable characteristic quantity under the dominant mode; S4: Combining partial least squares regression and improved analytic hierarchy process, establish a comprehensive evaluation model for the influencing factors of system oscillation energy.

[0008] Preferably, step S1 specifically involves: updating and calculating the root mean square value of the current in the sliding time window; if the value is detected to be continuously increasing or maintaining a certain increment within 2-3 consecutive time windows, an oscillation alarm signal is issued to locate the starting point of the region's growth and execute subsequent operations. Singular spectrum analysis (SSA) is used to analyze the directly acquired measurement data, obtain the principal component characteristics in the time-series signal, and filter out high-frequency noise. Specifically, this includes: Step S11: Construct a trajectory matrix from the one-dimensional oscillation data by embedding the dimension, and then decompose the matrix into multiple components sorted by the magnitude of the singular values ​​using singular value decomposition. The key is to set a reasonable window length. This invention takes into account the characteristic that the effective rank of the dominant signal matrix is ​​usually low, and selects... To balance the accuracy and computational efficiency of the analysis ( (where is the sequence length of the time-series signal to be analyzed).

[0009] Step S12: High-frequency noise corresponds to smaller singular values, reflecting irregular, high-frequency oscillation characteristics in the components; while the main oscillation signal corresponds to larger singular values, appearing in pairs in the singular value spectrum. By analyzing the singular value spectrum and the corresponding eigenvector spectrum, component groups representing high-frequency noise are identified and eliminated. (Previous selection...) By reconstructing a larger component, signal-to-noise separation can be achieved. Considering the complex characteristics of power grid signals, a ranking index is set. To help determine Value, of which, To select the number of recombinations, It reflects the change in the contribution of the selected order to the system. Step S13: The retained effective component group is restored to the time domain sequence by diagonal averaging, so as to effectively filter out high-frequency noise and extract a clean signal.

[0010] Preferably, step S2 specifically involves: using an improved adaptive variational mode decomposition (VMD) method to obtain the dominant oscillation mode components of the system. Specifically, this includes: Step S21: Select a window length of 1 / 2 and perform mirror extension on the reconstructed denoised data. Although SSA can eliminate the influence of noise and high-frequency components on algorithm analysis in practical engineering, given that the VMD calculation process involves Hilbert transform, there is still an endpoint effect, that is, the identification results of each modal component obtained by decomposition have large errors at the boundaries, which easily produces spurious frequency components. The data mirror extension used in this invention is simple to implement, has a small computational load, and can suppress the port signal divergence caused by energy loss to a certain extent.

[0011] Step S22: Adaptive parameter optimization to determine the optimal number of decomposition modes for variational mode decomposition. With penalty factor Taking into account both decomposition effectiveness and computational speed, this invention selects... =2000 has strong adaptability; if aliasing occurs in subsequent studies, it can be appropriately increased. , The value is generally in the range of 2000-4000. Principal component information is determined based on SSA calculations. Considering that oscillation modes always appear in pairs, one residual component is retained, and the selected component is... Modal number can be quickly achieved Preliminary selection.

[0012] It is worth noting that when the noise content is high or the modal frequencies are dense, the selection of the order determination index obtained from the solution is not clear. To avoid filtering out important components with low content but fast decay rate, the algorithm stores the number of order determination indices that can be satisfied and selects the maximum value as the final order determination index. Value. At this time, If the value is too large, further filtering is needed to determine the number of valid modes. A correlation coefficient criterion is introduced, by calculating the correlation coefficients between each modal component and between each modal component and the original data. ,in This represents the number of signal sampling points. The threshold is set to 0.1, meaning that if the cross-correlation coefficient between two modal components is greater than 0.1, or greater than the cross-correlation coefficient between the two components and the original data, then the cross-correlation coefficient is set to 0.1. If the similarity is too high, it is considered that there is over-decomposition and it needs to be reduced. The value can ultimately yield accurate and effective modal components, facilitating subsequent identification of oscillation parameters and stability assessment.

[0013] Step S23: Solve for detailed oscillation parameters using the Prony method. Since the preceding steps have essentially determined the dominant mode order of the denoised reconstructed signal, applying the Prony method directly at this point allows for the rapid and accurate extraction of mode parameters from the oscillation signal, including frequency, damping, amplitude, and phase, further determining the dominant oscillation mode components of the system.

[0014] Preferably, step S3 specifically involves: extracting the dissipated energy that characterizes the dominant oscillation mode of the wind farm grid-connected system through dynamic energy analysis, and evaluating the stability level of the system based on this. Combining the extracted modal parameters, the dissipation intensity corresponding to the oscillation energy under this operating condition can be calculated.

[0015] Given that oscillations are essentially the accumulation and dissipation of disturbance energy among various devices in a system, when constructing the system's energy function, the potential energy term can simultaneously characterize the magnitude of both the disturbance energy and the oscillation amplitude, while the dissipation energy term reflects both the magnitude of disturbance energy consumption and oscillation damping. Wind farm modeling is generally based on... In a coordinate system, the port energy can be defined as: in For port node via side road To the node The energy flow. Existing research shows that, since subsynchronous oscillations are dynamically controlled by non-power frequency quantities, the energy function calculation method cannot fully express the complex electromagnetic dynamic characteristics of oscillations. To avoid leading to unreasonable stability assessment conclusions, this invention uses the following formula, based on the Hamiltonian model, to calculate the oscillation energy generated by a system. To accurately obtain the subsynchronous oscillation mode energy of the system under study, it is necessary to collect the instantaneous values ​​of the equipment port voltage and current over a period of time, then perform coordinate transformation, and substitute them into the calculation to obtain the energy. Considering that when power oscillations occur in the grid-connected system of a direct-drive wind farm, frequency coupling often occurs on the AC side, and the amplitude of the supersynchronous mode component is relatively large, therefore, the subsynchronous / supersynchronous components must be considered simultaneously when analyzing a direct-drive wind farm. a Taking a phase as an example, the voltage and current components when a direct-drive wind farm grid-connected system experiences a short-term stable electrical (SSO) event are as follows: in, and These are the effective value and initial phase of the fundamental frequency voltage. i 0 and γ 0 represents the effective value and initial phase of the fundamental frequency current, while the subscripts "sub" and "sup" indicate the corresponding subsynchronous and supersynchronous frequency components. Combining this with the Park transform formula, we can obtain... dq The voltage and current in the coordinate system are as follows: in, ,definition Substitute the obtained voltage and current components into the oscillation energy calculation formula, and then compare the obtained oscillation energy with time. Differentiation yields the corresponding oscillation power, an equation composed of oscillation and non-oscillation terms, which is relatively complex and will not be elaborated upon here. The non-oscillation term represents the energy flow power corresponding to energy dissipation, i.e., the fitting slope of the oscillation energy. This is called dissipation intensity. According to Lyapunov's second principle, the trend of energy flow power variation can represent the damping of the power system; therefore, dissipation intensity can serve as a primary basis for oscillation stability judgment and disturbance source location. Retaining the aperiodic components in the expression, the dissipation intensity expression for the wind farm subsystem is obtained as follows: when When this value is high, it indicates that the energy emitted by the subsystem is continuously increasing, exhibiting negative damping characteristics for oscillations. Furthermore, the larger this value, the faster the emitted energy increases, and the more violent the system oscillation divergence. When the subsystem exhibits undamped characteristics to oscillations, it is in a critically stable state; when When the value is less, it indicates that the dynamic energy generated by the subsystem gradually decreases, exhibits positive damping characteristics for oscillations, and the system will gradually converge and tend to stabilize. The smaller the value, the higher the stability level of the system.

[0016] Preferably, step S4 specifically involves: First, selecting factors that may affect oscillation characteristics as independent variables and the dissipation intensity corresponding to the dominant oscillation mode under the corresponding operating condition as the dependent variable, and performing partial least squares regression to obtain the objective weights of the relevant influencing factors; then, using an improved analytic hierarchy process to calculate the corresponding subjective weights, and using a scaling method to construct a comparison judgment matrix to ensure that the subjective evaluation results are reasonable and scientific; finally, based on the principle of minimum information identification, and comprehensively considering both subjective and objective information, establishing a quantitative evaluation model for the analysis of factors affecting oscillation characteristics.

[0017] This invention also proposes a data-driven dynamic monitoring system for subsynchronous oscillations in wind farms, comprising: The data acquisition module is used to extract voltage, current or power signals from the corresponding monitoring points of the wind farm grid-connected system and to make a preliminary assessment of the system status. The data preprocessing module is used to filter out high-frequency noise in the measurement data to be analyzed and retain the principal component characteristics in the time series signal; The parameter identification module is used to extract relatively stable subsequences at each frequency component scale, obtain accurate effective modal components, and facilitate subsequent oscillation parameter identification and stability assessment. The feature extraction module is used to analyze the dominant oscillation mode energy function of the wind farm grid-connected system derived from the energy perspective, and obtain feature quantities that can be quantified to assess the stability level of the system oscillation.

[0018] The variable selection module is used to determine the relevant factors that may affect the stability of the wind farm grid-connected system based on the operating principle of the wind farm under study and in combination with existing relevant research results, for the oscillation modes caused by the action of different equipment. These factors are then used as independent variables to establish a comprehensive evaluation model of the influencing factors of the system's oscillation energy.

[0019] The model generation module is used to determine the influencing factors of the dominant oscillation mode of the wind farm grid-connected system under study, based on the selected independent variables and the corresponding calculated stable characteristic quantities as dependent variables, and by comprehensively considering subjective and objective information.

[0020] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the response data-driven dynamic monitoring and influencing factor assessment method for oscillation parameters of wind farm grid-connected systems as described above.

[0021] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any of the above-described data-driven method for dynamic monitoring and stability assessment of oscillations in a wind farm grid-connected system.

[0022] Compared to existing technologies, this invention discloses a data-driven method and system for dynamic monitoring of subsynchronous oscillations in wind farms. ①The proposed SSA-improved VMD method can accurately extract signals of each mode and is suitable for analyzing various complex oscillation signals with multimodality, time-varying and nonlinearity; ② Determined stability characteristic quantities of the wind farm grid-connected system It can quantitatively assess the stability level of SSO in a wind farm. This value integrates measured voltage and current information, and its difference is significant compared to that reflected by the system damping ratio. It can accurately reflect the energy change of SSO in the system and has higher sensitivity and stronger interpretability. ③ The comprehensive evaluation model for the influencing factors of the dominant oscillation mode, which combines partial least squares regression and improved analytic hierarchy process, is reliable and reasonable, and has strong flexibility while accurately evaluating the influencing factors. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating the analysis of a data-driven dynamic monitoring method for subsynchronous oscillations in wind farms proposed in this invention. Figure 2 This is a schematic diagram of the direct-drive wind farm grid-connected system via LCC-HVDC, studied in the feasibility analysis of Example 1 of this invention. Figure 3 This is a flowchart illustrating the specific solution process of the improved adaptive variational mode decomposition method used in this invention. Figure 4 The waveform diagrams of each mode are obtained by decomposing the voltage collected at the grid connection point under the oscillation condition of a certain wind farm grid-connected system generated by the present invention through the oscillation monitoring method proposed in the present invention. Figure 5aThe image shows the time-frequency diagrams of various modes obtained by decomposing the current collected at the grid connection point under the oscillation condition of a wind farm grid-connected system, generated by the present invention through a time-domain model, using the oscillation monitoring method proposed in this invention. Figure 5b The image shows the results obtained by performing Fast Fourier Decomposition on the corresponding current data. Figure 6 As the objective evaluation part of the oscillation stability assessment model of the wind farm grid-connected system in this invention, partial least squares regression analysis is performed on the selected influencing factors and the dissipation intensity under the corresponding operating conditions to obtain the quantitative relationship diagram between the relevant influencing factors and the dissipation intensity. Figure 7 This is a schematic diagram illustrating the specific solution process of the improved analytic hierarchy process used in this invention. It combines the arithmetic mean method, geometric mean method, and characteristic method to calculate weights, ensuring the robustness of subjective analysis results. Figure 8 This is a topology diagram of a data-driven dynamic monitoring method for subsynchronous oscillations in wind farms proposed in this invention. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0025] A data-driven method for dynamic monitoring of subsynchronous oscillations in wind farms includes the following steps: S1: Collect voltage and current data at the grid connection point of the wind farm, and perform preprocessing and initial status assessment of the data; The data collected in step S1 can come from the grid connection point detection data of the actual power system, or from the established electromagnetic transient simulation model of the wind farm grid-connected system. By adjusting relevant system operating parameters such as wind speed, equivalent number of wind turbines, controller parameters, and grid-connected transmission line impedance, multiple sets of operating data under different operating conditions are obtained to form a research dataset. After obtaining the dataset for analysis, the root mean square value of the short-time window current is calculated. If this value continuously increases or shows a certain increment within 2-3 consecutive sliding analysis time windows, an oscillation alarm signal is issued. This triggers singular spectrum analysis of the data within the current time window to preliminarily determine the number of effective modes and achieve data reconstruction and noise reduction.

[0026] As the "high-voltage and high-efficiency" characteristics of power systems become increasingly pronounced, the interaction between wind farms, power grids, and related equipment becomes more complex. Therefore, this invention combines... Figure 2The study focuses on the transmission system of a direct-drive wind farm via a grid-commutated-converter based high-voltage direct current (LCC-HVDC) system. A simulation model is established in the PSCAD / EMTDC platform. Figure 2 The time-domain model of the wind farm grid-connected system shown adopts a single-unit equivalent model, with a total of 80 equivalent wind turbines, each with a rated capacity of 5MW, and a rated wind speed of 8m / s. The analysis dataset is obtained by modifying parameters such as wind speed, number of wind turbines, stabilizing DC capacitor, grid-side converter control parameters, phase-locked loop control parameters, and transmission line impedance.

[0027] S2: Identify oscillation mode parameters and determine the dominant mode; In step S2, the following is adopted: Figure 3 The improved variational mode decomposition method shown introduces a correlation coefficient criterion to evaluate the decomposition results. After determining the optimal number of decomposed modes and the penalty factor, the effective decomposition of modal components can be achieved quickly and accurately. To quickly obtain stable characteristic quantities, the Prony method is used to solve for detailed oscillation parameters.

[0028] Taking the scenario of changing the outer loop integral coefficient of the d-axis voltage of the grid-side converter under initial steady-state conditions as an example, the process of oscillation parameter identification is demonstrated. Based on... Figure 3 The process involves analyzing the a-phase component of the grid connection voltage at the direct-drive wind farm to obtain the number of SSA recombinations. =8, at this point the waveform similarity coefficient (NCC) between the initial data and the denoised data is 0.9993, and the mean square error (MSE) is 0.6376, indicating that the denoising effect is quite good and the number of effective modes is [missing information]. =4, the time-domain waveforms of each mode are obtained as follows: Figure 4 As shown. Among them, For power frequency components, and The component is the dominant secondary / supersynchronous component. The supersynchronous component is 78Hz, and the coupled subsynchronous component, due to its relatively small content, is filtered out in the SSA stage. The energy of each component of the current phase a is decomposed as shown in Figure 5a). To verify the accuracy of the decomposition algorithm, a fast Fourier analysis was performed on the denoised signal, and the results are shown in Figure 5b). As can be seen from Figure 5, the important frequency components identified by the fast Fourier analysis are basically consistent with the components identified by the improved adaptive variational mode decomposition method proposed in this invention, indicating that the decomposition effect of this identification method is quite good, and it has a very good processing accuracy for multi-mode and nonlinear signals. It can intuitively, continuously and accurately reflect the state changes of each component of the system.

[0029] To improve the accuracy and speed of energy dissipation calculation, this invention does not use fitting to solve the four extracted oscillation components. Instead, it chooses to perform improved Prony analysis on each component to directly calculate the relevant modal parameters, thereby obtaining the energy dissipation coefficient of the dominant oscillation mode under the corresponding operating condition. The calculation results of the modal parameters of the dominant voltage and current components under this operating condition are shown in Table 1.

[0030] Table 1 Dominant Mode Parameters of Output Voltage and Current Components

[0031] S3: The dissipation intensity corresponding to the oscillation energy under this working condition is calculated, which is the stable characteristic quantity under the dominant mode; As shown in Table 1, the dominant secondary / supersynchronous components separated from the wind farm's outlet current and voltage components have very similar frequencies and minimal differences in attenuation factors, exhibiting essentially equal amplitude oscillations. This indicates that the system is experiencing power oscillations around 14Hz. The dissipation intensity corresponding to the dominant oscillation mode is calculated as follows:

[0032] A dissipation intensity greater than 0 indicates that the direct-drive wind farm is continuously emitting oscillating energy, and the system is experiencing subsynchronous oscillation at this time.

[0033] S4: Combining partial least squares regression and improved analytic hierarchy process, establish a comprehensive evaluation model for the influencing factors of system oscillation energy.

[0034] In selecting factors influencing the oscillation characteristics of wind farm grid-connected systems, considering that the output active power of a wind farm significantly depends on the number of wind turbines in operation and the wind speed, existing research has shown that the design of the stabilizing DC capacitor in the internal converter of the wind farm, the controller parameter settings, and the strength of the external AC system all have a significant impact on system stability. Furthermore, real-time monitoring and analysis of the changes in voltage, current, and power components at the wind farm grid connection point can obtain key indicators reflecting system oscillation fluctuations. Based on the above analysis, this invention focuses on analyzing the voltage and current components at the grid connection point of a direct-drive wind farm, employing an improved adaptive variational mode decomposition method to perform oscillation mode decomposition to assess system stability, and calculating the system dissipation intensity based on the extracted oscillation signal as the dependent variable of the evaluation model; nine influencing factors are selected as the independent variables of the evaluation model: number of wind turbines, wind speed, DC capacitor, grid-side converter control parameters, phase-locked loop control parameters, and AC system strength. The data analyzed in this example were all obtained through time-domain simulation. The control parameters of the grid-side converter were changed at 2 seconds to induce subsynchronous oscillations, while external parameters such as the number of wind turbines, wind speed, and grid-connected transmission line impedance were randomly varied. Analysis was conducted on 600 sets of constant-amplitude oscillation data within 3-4 seconds to evaluate the impact of each factor on oscillation stability.

[0035] Partial least squares regression analysis was performed on the selected influencing factors and the dissipation intensity under the corresponding working conditions, resulting in 9 pairs of principal components. and The corresponding factor variance explanations and the cumulative projective importance index of the independent variables (VIP) were calculated, and the optimal number of principal components was determined to be 4. The corresponding component matrix is ​​shown in Table 2.

[0036] Table 2 Principal Component Matrix Table

[0037] The standardized regression equation and the original variable regression equation can be obtained as follows: ;in, and These represent the standardized variables and dissipation intensity, respectively. The obtained regression coefficients are then validated. The values ​​are all less than 0.05, indicating that the model is significant; goodness of fit This indicates that the established model can reveal the degree of change in the dependent variable quite well. The coefficients of the standard regression equation are normalized to obtain the objective weights. = Furthermore, standardized coefficients can intuitively reflect the trend and importance of the independent variable's effect on the dependent variable. The absolute values ​​of the standardized coefficients obtained from the regression are arranged in descending order, such as... Figure 6 As shown, the dominant factors negatively affecting dissipation intensity are wind speed and AC system strength, indicating that as these parameters increase, dissipation intensity decreases, meaning the oscillating energy emitted by the wind farm decreases and system stability improves. The outer voltage control coefficient of the grid-side converter, the number of wind turbines, the inner current control coefficient, and the DC capacitance value are the main factors positively affecting dissipation intensity, indicating that increasing these parameters leads to increased system oscillation energy and worsens system stability. These analytical results are largely consistent with existing studies using small-signal models to analyze oscillation characteristics, verifying the effectiveness of the dissipation intensity regression model established based on measurement data, and further clarifying that dissipation intensity is an effective characteristic quantity for oscillating stability response.

[0038] When calculating the subjective weights, the analysis results of relevant literature were referenced. Considering that the number of wind turbines remains constant during normal operation and that the phase-locked loop has little impact on the DC capacitor-dominated oscillation, the nine influencing factors are ranked in descending order of importance as follows: When improving the hierarchical method for subjective weight calculation, the scale value is set as follows: Thus, a judgment matrix of nine influencing factors was obtained. ,in, satisfy ,show It has consistency, so there is no need to adjust the judgment matrix, which improves the reliability and practicality of objective weight analysis. according to Figure 7 The process is solved to obtain the subjective weights. = .

[0039] This invention is based on the principle of minimum information discrimination, fully considers the differences between subjective and objective information, and constructs a weight optimization model as shown below, and uses the least squares method to obtain the comprehensive weight. . The comprehensive weighting coefficient can be obtained from the above calculation results. = In summary, through analysis of the measurement data, the main factors dominating the oscillation characteristics of direct-drive wind farms are wind speed and grid-side converters. The integral coefficient of the outer loop control for shaft voltage, the proportional coefficient of the inner loop control for current, the number of fans, the AC system strength, and the DC capacitor are all factors to consider. Subsequent analysis should focus on identifying the strongly correlated influencing factors, and combining this with the results of dissipation strength calculations to guide system parameter optimization control, thereby enabling faster and better resolution of oscillation suppression and disturbance source localization issues.

[0040] This invention discloses a data-driven dynamic monitoring system for subsynchronous oscillations in wind farms, comprising: The data acquisition module is configured to: acquire real-time data from monitoring points of the wind farm grid-connected system and perform preliminary analysis of the system status; upon receiving an oscillation alarm signal, the following modules will be activated. The data preprocessing module is configured to: apply singular spectrum analysis to filter out high-frequency noise in the measurement data to be analyzed, retain the principal component characteristics in the time series signal, and provide a preliminary number of modes; The parameter identification module is configured to: extract accurate effective modal components by applying an improved adaptive variational mode decomposition method, and perform improved Prony analysis on each component to directly calculate the relevant modal parameters, which facilitates subsequent oscillation parameter identification and stability assessment. The feature extraction module is configured to: analyze the dominant oscillation mode energy function of the wind farm grid-connected system derived from the energy perspective, and obtain feature quantities that can quantify the stability level of the system oscillation.

[0041] The variable selection module is configured to: based on the operating principle of the wind farm grid-connected system under study and combined with existing relevant research results, identify relevant factors that may affect the stability of the system for the oscillation modes caused by the action of different equipment, and use them as independent variables to establish a comprehensive evaluation model of the influencing factors of the system on oscillation energy.

[0042] The model generation module is configured to: determine the influencing factors of the dominant oscillation mode of the wind farm grid-connected system under study by comprehensively considering subjective and objective information, based on the selected independent variables and the corresponding calculated stable characteristic quantities as dependent variables.

[0043] This invention provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the response data-driven dynamic monitoring and influencing factor assessment method for oscillation parameters of a wind farm grid-connected system described in Example 1. These steps include: S1: Collect voltage and current data at the grid connection point of the wind farm, and perform preprocessing and initial status assessment of the data; S2: Identify oscillation mode parameters and determine the dominant mode; S3: The dissipation intensity corresponding to the oscillation energy under this working condition is calculated, which is the stable characteristic quantity under the dominant mode; S4: Combining partial least squares regression and improved analytic hierarchy process, establish a comprehensive evaluation model for the influencing factors of system oscillation energy.

[0044] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any of the above-described data-driven method for dynamic monitoring and stability assessment of oscillations in a wind farm grid-connected system.

[0045] 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 exemplary embodiments according to this application. 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.

[0046] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data-driven method for dynamic monitoring of subsynchronous oscillations in wind farms, characterized in that: Includes the following steps: S1: Collect voltage and current data at the grid connection point of the wind farm, and perform preprocessing and initial status assessment of the data; S2: Identify oscillation mode parameters and determine the dominant mode; S3: The dissipation intensity corresponding to the oscillation energy under this operating condition is calculated, which is the stability characteristic quantity under the dominant mode. Specifically, step S3 involves: extracting the dissipation energy that can characterize the dominant oscillation mode of the system through dynamic energy analysis of the wind farm grid-connected system, evaluating the stability level of the system based on this, and calculating the dissipation intensity corresponding to the oscillation energy under this operating condition by combining the extracted mode parameters. Given that the essence of oscillation is the accumulation and dissipation of disturbance energy among various devices in a system, when constructing the system's energy function, the potential energy term can simultaneously characterize the magnitude of the disturbance energy and the oscillation amplitude, while the dissipation energy term simultaneously reflects the magnitude of disturbance energy consumption and oscillation damping. Wind farm modeling is generally based on... dq In a coordinate system, the port energy can be defined as: in W ij For port node i via side road L ij To the node j Based on the Hamiltonian model, the energy flow of a given system is calculated using the following formula, which outlines the oscillating energy generated by that system. To accurately obtain the subsynchronous oscillation mode energy of the system under study, it is necessary to collect the instantaneous values ​​of the equipment port voltage and current over a period of time, then perform coordinate transformation, and substitute them into the calculation. Considering that when power oscillation occurs in the grid-connected system of a direct-drive wind farm, the AC side often exhibits frequency coupling, and the amplitude of the supersynchronous mode component is relatively large, therefore, the subsynchronous / supersynchronous components must be considered simultaneously when analyzing a direct-drive wind farm. a The voltage and current components of a direct-drive wind farm grid-connected system during a short-term stable electrical (SSO) event are as follows: in, u 0 and φ 0 represents the effective value and initial phase of the fundamental frequency voltage. i 0 and γ 0 represents the effective value and initial phase of the fundamental frequency current. The subscripts "sub" and "sup" represent the subsynchronous and supersynchronous frequency components, respectively. Using the Park transform formula, we can obtain... dq The voltage and current in the coordinate system are as follows: in, ,definition The obtained voltage and current components are substituted into the oscillation energy calculation formula, and the oscillation energy obtained is calculated relative to time. t Differentiation yields the corresponding oscillation power, where the non-oscillatory term represents the energy flow power corresponding to energy dissipation, i.e., the fitting slope of the oscillation energy. a As for dissipation intensity, according to Lyapunov's second principle, the trend of energy flow power variation can represent the damping of the power system. Therefore, dissipation intensity is the main basis for oscillation stability judgment and disturbance source location. By retaining the aperiodic component in the expression, the dissipation intensity expression of the wind farm subsystem can be obtained as follows: when E EF When the value is greater than 0, it indicates that the energy emitted by the subsystem is continuously increasing, exhibiting negative damping characteristics for oscillations. Furthermore, the larger the value, the faster the emitted energy increases, and the more violent the system oscillation divergence. E EF When = 0, it indicates that the subsystem exhibits undamped characteristics to oscillations, and the system is in a critically stable state; when E EF When the value is less than 0, it indicates that the dynamic energy generated by the subsystem gradually decreases, exhibits positive damping characteristics for oscillation, and the system will gradually converge and tend to stabilize. The smaller the value, the higher the stability level of the system. S4: Combining partial least squares regression and improved analytic hierarchy process, establish a comprehensive evaluation model for the influencing factors of system oscillation energy.

2. The data-driven dynamic monitoring method for subsynchronous oscillations in wind farms according to claim 1, characterized in that, Step S1 specifically involves: updating and calculating the root mean square value of the current in the sliding time window; when it is detected that the value continues to increase or shows a certain increment within 2-3 consecutive time windows, an oscillation alarm signal is issued to locate the starting point of the region's growth and execute subsequent operations; using singular spectrum analysis to analyze the directly acquired measurement data, obtain the principal component characteristics in the time series signal, and filter out high-frequency noise, specifically including: Step S11: Construct a trajectory matrix from the one-dimensional oscillation data by embedding the dimension, and decompose the matrix into multiple components sorted by the size of the singular values ​​using singular value decomposition. Set a reasonable window length L. Considering that the effective rank of the dominant signal matrix is ​​usually low, select L=N / 3 to balance the accuracy of the analysis and the computational efficiency. N is the sequence length of the time series signal to be analyzed. Step S12: High-frequency noise corresponds to smaller singular values, reflecting irregular, high-frequency oscillation characteristics in the components; while the main oscillation signal corresponds to larger singular values, appearing in pairs in the singular value spectrum. By analyzing the singular value spectrum and the corresponding eigenvector spectrum, the component groups representing high-frequency noise are identified and eliminated, and the top components are selected. K A large component reconstruction is performed to achieve signal-to-noise separation. Considering the complex characteristics of the power grid signal, a ranking index R is set. i To help determine K Value, of which, P To select the number of recombinations, G p This reflects the change in the contribution of the selected order to the system. Step S13: The retained effective component group is restored to the time domain sequence by diagonal averaging, so as to effectively filter out high-frequency noise and extract a clean signal.

3. The data-driven dynamic monitoring method for subsynchronous oscillations in wind farms according to claim 1, characterized in that, Step S2 specifically involves: using an improved adaptive variational mode decomposition method to obtain the dominant oscillation mode components of the system, specifically including: Step S21: Select a window length of 1 / 2 and perform mirror extension on the reconstructed denoised data; Step S22: Adaptive parameter optimization. Determine the optimal number of decomposed modes M and penalty factor α for variational mode decomposition. Introduce the correlation coefficient criterion by calculating the correlation coefficient ρ between each mode component and between each mode component and the original data, where N is the number of signal sampling points. The threshold is set to 0.1, meaning that if the cross-correlation coefficient between two modal components is greater than 0.1, or greater than the cross-correlation coefficient between the two components and the original data, then the cross-correlation coefficient is set to 0.

1. x ( n If the similarity between the two modal components is considered to be excessive, the value of M needs to be reduced to obtain accurate effective modal components, which facilitates subsequent identification of oscillation parameters and stability assessment. Step S23: Use the Prony method to solve for detailed oscillation parameters.

4. The data-driven dynamic monitoring method for sub-synchronous oscillations in wind farms according to claim 1, characterized in that, Step S4 specifically involves: First, selecting factors that may affect oscillation characteristics as independent variables and the dissipation intensity corresponding to the dominant oscillation mode under the corresponding operating condition as the dependent variable, and performing partial least squares regression to obtain the objective weights of the relevant influencing factors; then, using the improved analytic hierarchy process to calculate the corresponding subjective weights, and using the scaling method to construct a comparison judgment matrix to ensure that the subjective evaluation results are reasonable and scientific; finally, based on the principle of minimum information identification, and comprehensively considering both subjective and objective information, establishing a quantitative evaluation model for the analysis of factors affecting oscillation characteristics.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images