A method and system for quantitatively evaluating wind power fluctuation and intermittency

Abstract

This invention relates to the field of wind power new energy technology, and in particular to a method and system for quantitatively evaluating the volatility and intermittency of wind power. It introduces a sparrow search algorithm to iteratively optimize key parameters of variational mode decomposition, overcoming the shortcomings of traditional methods that rely on human experience for parameter selection, are highly subjective, and are prone to getting trapped in local optima, thus significantly reducing mode aliasing. By constructing a dual-threshold mode division rule based on center frequency and sample entropy, it achieves accurate identification of noise-dominant components. Through targeted denoising and superposition reconstruction, it effectively filters out noise interference while preserving the true fluctuation details of wind power to the greatest extent. Based on the reconstructed signal, it calculates the volatility index and wind power ramp duty cycle separately, achieving independent quantification of wind power volatility and intermittency characteristics, solving the problems of ambiguous definitions and mixed indicators in existing technologies, and providing accurate quantitative basis for power system dispatch.

Description

Technical Field

[0001] This invention relates to the field of wind power new energy technology, and in particular to a method and system for quantitatively evaluating the volatility and intermittency of wind power. Background Technology

[0002] With the global shortage of fossil fuels and increasingly severe environmental problems, wind power, as a clean and low-carbon renewable energy source, has been developed on a large scale. However, the volatility and intermittency of wind power output restrict its grid integration efficiency. Accurately quantifying these two characteristics is a key prerequisite for optimized power system dispatch and efficient wind power integration.

[0003] Currently, quantitative research on wind power output characteristics can be mainly summarized into the following three technical paths: The first is statistical analysis methods based on power series, such as using mathematical tools like power spectral density analysis and maximum power change rate calculation to process historical data. However, in practical applications, this type of method often confuses the intermittency and volatility of wind power, lacking a clear definition and distinction between the two physical concepts. The second is a judgment method based on state switching, which artificially divides the operating states of wind turbines into those with and without power by setting wind power density thresholds, thereby assessing intermittent characteristics. The set thresholds often depend on historical data from wind farms in specific regions, resulting in poor universality, regional limitations, and an inability to reflect the fluctuating characteristics of wind power during periods of power availability. The third is a reverse evaluation method based on grid connection impact. This method indirectly measures the fluctuating characteristics of wind power by monitoring the impact of wind power grid connection on system parameters such as frequency and voltage. It is easily affected by multiple factors such as load changes, network structure, and regulation by other power sources, making it difficult to isolate external interference and directly characterize the inherent characteristics of wind power.

[0004] In recent years, although some studies have proposed using wind power ramp-up duty cycle to quantify wind power intermittency, these methods do not consider spurious ramp-up events caused by noise interference when processing raw data, leading to misjudgments and statistical biases in the intermittent events. Meanwhile, in the field of signal decomposition, the parameters of traditional variational mode decomposition rely on manual experience or simple grid search methods, resulting in high subjectivity, low optimization efficiency, and a tendency to cause mode aliasing, affecting the subsequent quantification accuracy of wind power output component characteristics.

[0005] In summary, existing technologies not only have vague definitions of wind power volatility and intermittency and lack systematic quantitative indicators, but also fail to adequately explore the spatiotemporal distribution characteristics of wind power output and its deep correlation with meteorological factors, resulting in limited adaptability of quantitative results to the actual operational needs of the power system. Summary of the Invention

[0006] This invention aims to provide a method and system for quantitatively assessing the volatility and intermittency of wind power, constructing a scientific quantitative index system to improve the accuracy and reliability of wind power characteristic analysis, and providing support for power system dispatch, reserve capacity configuration, and efficient wind power consumption. It addresses the technical problems of existing technologies, such as vague definitions of wind power volatility and intermittency, lack of systematic quantitative indicators, insufficient exploration of the spatiotemporal distribution characteristics of wind power output and its deep correlation with meteorological factors, resulting in limited adaptability of quantitative results to the actual operational needs of the power system.

[0007] To achieve the above objectives, the first aspect of the present invention provides a method for quantitatively evaluating the volatility and intermittency of wind power, comprising the following steps: Obtain historical active power data of the target wind farm, and generate a signal to be decomposed based on the historical active power data; An initial parameter combination is constructed for variational mode decomposition of the signal to be decomposed, and then the initial parameter combination is iteratively optimized based on the sparrow search algorithm to obtain the target parameter combination. Based on the target parameter combination, variational mode decomposition is performed on the signal to be decomposed to obtain several intrinsic mode function components and the center frequency of each intrinsic mode function component; Obtain the sample entropy of each intrinsic mode function component; determine the mode type of each intrinsic mode function component based on the center frequency and sample entropy of each intrinsic mode function component according to a preset threshold division rule; Based on the mode type of each intrinsic mode function component, the intrinsic mode function components are denoised and superimposed to reconstruct the signal, thereby obtaining the reconstructed wind power signal. The power change rate of the reconstructed wind power signal within adjacent time intervals is obtained, and then the volatility index of the reconstructed wind power signal is analyzed based on the power change rate and the coefficient of variation formula. Obtain the rated power of the wind turbines in the target wind farm, and then, based on the power change rate and the rated power of the wind turbines, analyze the wind power ramp-up duty cycle of the reconstructed wind power signal according to the preset ramp-up event judgment rules. The volatility index is used to obtain the quantitative assessment results of wind power volatility, and the wind power ramp duty cycle is used to obtain the quantitative assessment results of wind power intermittency.

[0008] The aforementioned method for quantitatively assessing the volatility and intermittency of wind power firstly ensures data validity by preprocessing and constructing historical active power data. Secondly, it introduces a sparrow search algorithm to iteratively optimize key parameters of variational mode decomposition, effectively overcoming the shortcomings of traditional methods that rely on human experience for parameter selection, are highly subjective, and are prone to getting trapped in local optima. This significantly reduces mode aliasing and ensures the adaptability and accuracy of signal decomposition. Thirdly, by constructing a dual-threshold mode partitioning rule based on center frequency and sample entropy, it achieves the classification of noise-dominant components, high-frequency fluctuation components, and low-frequency trend components. The system achieves accurate identification by employing targeted denoising and superposition reconstruction. This effectively filters out noise interference while preserving the true fluctuation details of wind power to the greatest extent possible, solving the problem of traditional denoising methods easily losing high-frequency feature information. Finally, based on the reconstructed signal, the volatility index and wind power ramp duty cycle are calculated separately. From the two dimensions of time-domain statistical characteristics and event occurrence probability, the system achieves independent and scientific quantification of the volatility and intermittency characteristics of wind power, clarifies the physical connotation of volatility and intermittency, and solves the problems of ambiguous definitions and mixed indicators in existing technologies. This provides an accurate quantitative basis for power system dispatching.

[0009] Further, the step of acquiring historical active power data of the target wind farm and generating a signal to be decomposed based on the historical active power data includes: Outliers in the historical active power data are removed to obtain the first power time series data; The first power time series data is filled in using linear interpolation to obtain the second power time series data; Gaussian white noise is injected into the second power time series data to obtain the signal to be decomposed.

[0010] In this implementation, outlier removal and linear interpolation imputation of the original data correct for missing or erroneous data caused by downtime, power outages, measurement errors, etc., ensuring the integrity and accuracy of the input data and laying the foundation for subsequent precise decomposition. Injecting Gaussian white noise into the preprocessed data is a common technique based on signal decomposition theory. This aims to use noise to help variational mode decomposition better avoid mode aliasing when processing complex nonlinear signals, thereby improving the stability and robustness of the decomposition.

[0011] In this implementation, outlier removal and linear interpolation imputation of the original data can correct missing or erroneous data caused by wind farm shutdowns, power curtailment, measurement errors, etc., ensuring the integrity and accuracy of the input power time series data. Furthermore, injecting Gaussian white noise into the processed data can simulate actual sensor interference, making the data more closely resemble real-world operating conditions.

[0012] Furthermore, the construction of an initial parameter combination for variational mode decomposition of the signal to be decomposed, followed by iterative optimization of the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination, includes: Construct an initial total number of components and an initial penalty factor for variational mode decomposition of the signal to be decomposed, and then use the initial total number of components and the initial penalty factor as the initial parameter combination; A fitness function is constructed based on the signal to be decomposed, the initial total number of components, and the initial penalty factor. A sparrow search population is generated, wherein each individual in the sparrow search population corresponds to a set of random component totals and random penalty factors; Based on the sparrow search algorithm, the initial total number of components and the initial penalty factor are iteratively optimized in several rounds according to the fitness function and the sparrow search population until the final total number of components and the final penalty factor generated after a certain round of iterative optimization meet the preset optimization conditions. Then, the final total number of components and the final penalty factor are used as the target parameter combination. In any round of the aforementioned iterative optimization process: Based on the fitness function, obtain the fitness function value of the total number of random components and the random penalty factor corresponding to each individual in the sparrow search population; Based on the sparrow search algorithm, the two-dimensional position vector of each individual in the population is updated based on the fitness function value corresponding to each individual in the population, thereby obtaining the iterative search population; The sparrow search population is updated based on the iterative search population, and the final total number of components and the final penalty factor are obtained based on the iterative search population, thus ending this round of iterative optimization.

[0013] In this implementation, to address the issue that traditional variational mode decomposition parameters rely on manual experience for setting, a sparrow search algorithm is introduced for adaptive optimization. By constructing a fitness function that reflects the decomposition effect, the sparrow search algorithm simulates the discoverer-follower-watcher position update mechanism of sparrow population foraging behavior for iterative optimization. This efficient global search within the solution space can find the optimal total number of components and penalty factor for variational mode decomposition.

[0014] This process not only avoids the problem of excessive computation time in traditional grid search methods, but also eliminates the subjectivity and blindness of manually selecting parameters, overcoming the shortcomings of traditional methods such as strong subjectivity and low optimization efficiency. It ensures that variational mode decomposition can adapt the signal with the optimal parameter combination, thereby obtaining eigenmode function components with clear center frequencies and clear physical meanings, and greatly improving the efficiency and accuracy of signal decomposition.

[0015] Further, the variational mode decomposition of the signal to be decomposed based on the target parameter combination to obtain several intrinsic mode function components and the center frequency of each intrinsic mode function component includes: The signal to be decomposed is subjected to variational mode decomposition according to the variational mode decomposition algorithm to obtain the initial mode function components and the center frequency of each initial mode function component; The frequency domain update formula, center frequency update formula, and Lagrange multiplier update formula for modal components are constructed based on the alternating direction multiplier method. Based on the frequency domain update formula, the center frequency update formula, and the Lagrange multiplier update formula, the initial mode function components and the center frequency of each initial mode function component are iteratively updated in several rounds until the difference between the output results of two rounds of iterative updates is less than a preset threshold. Then, based on the output results of the two rounds of iterative updates, several intrinsic mode function components and the center frequency of each intrinsic mode function component are obtained.

[0016] In this implementation, the alternating direction multiplier method is used to solve the constrained variational problem of variational mode decomposition. This transforms the complex variational constraint problem into a series of easily solvable subproblems. Through alternating iterations of frequency domain update, center frequency update, and Lagrange multiplier update, the optimal solution is found through convergence. This process ensures that, given the optimal parameter combination, the original signal can be accurately decomposed into a series of eigenmode functions with specific sparsity characteristics and finite bandwidth, and the center frequency of each component can be accurately calculated. This achieves the decomposition of complex non-stationary wind power signals into several sub-band signals with different center frequencies. Compared to traditional recursive decomposition algorithms, this method effectively suppresses endpoint effects and mode aliasing, accurately captures the dynamic characteristics of wind power at different time scales, and improves the convergence speed and accuracy of the decomposition results.

[0017] Further, obtaining the sample entropy of each of the intrinsic mode function components includes: The standard deviation of the original wind power signal is obtained based on historical active power data, and then the tolerance threshold is obtained based on the standard deviation of the original wind power signal. For any of the aforementioned intrinsic mode function components: A phase space is constructed based on the intrinsic mode function component, such that each phase space vector in the phase space corresponds to a sequence value of the intrinsic mode function component; Obtain several spatial distances between any phase space vector and several other phase space vectors, and then analyze the proportion of several spatial distances that are less than the tolerance threshold to obtain the proportion parameter of the phase space vector; The sample entropy of the intrinsic mode function component is calculated based on the scaling parameter of each phase space vector.

[0018] In this implementation, sample entropy, a dimensionless metric for measuring the complexity and irregularity of a time series, is introduced to quantify the randomness of each intrinsic mode function component. By constructing a phase space, calculating vector distances, and comparing them with a tolerance threshold, sample entropy can effectively measure the probability of a signal generating new modes. The larger the sample entropy of a component, the stronger its randomness and irregularity. Furthermore, the tolerance threshold is determined based on the standard deviation of the original signal, ensuring consistency in the measurement scale for each component and objectively reflecting the complexity of each component under different wind power scenarios.

[0019] Further, the step of determining the mode type of each intrinsic mode function component based on the center frequency and sample entropy of each intrinsic mode function component according to a preset threshold division rule includes: Set the sample entropy threshold and frequency threshold; For any of the aforementioned intrinsic mode function components: If the sample entropy of the intrinsic mode function component is greater than the sample entropy threshold, and the center frequency of the intrinsic mode function component is greater than or equal to the frequency threshold, then the intrinsic mode function component is a noise-dominant mode type. If the sample entropy of the intrinsic mode function component is less than or equal to the sample entropy threshold, and the center frequency of the intrinsic mode function component is greater than or equal to the frequency threshold, then the intrinsic mode function component is a high-frequency fluctuation mode type. If the center frequency of the intrinsic mode function component is less than the frequency threshold, then the intrinsic mode function component is a low-frequency trend mode type.

[0020] This implementation considers the physical characteristics of wind power signals and proposes a dual-threshold classification rule based on center frequency and sample entropy to classify the decomposed modal components in a physical sense. Noise signals typically exhibit high frequency and high entropy, displaying high randomness; while the high-frequency fluctuations in wind power, though high in frequency, are influenced by meteorological factors and exhibit certain regularities; low-frequency components reflect the overall trend of the system. Therefore, in this implementation, high-frequency components with high sample entropy are identified as noise-dominant; high-frequency components with low sample entropy are identified as high-frequency fluctuation components, representing the main fluctuation components of wind power; and low-frequency components represent the long-term trend of wind power output. This rule effectively identifies noise components mixed in with the useful signal, avoiding the risk of misjudging useful high-frequency fluctuation features as noise and discarding them when filtering solely based on frequency, as is common in traditional methods. This ensures that the reconstructed signal retains complete wind power fluctuation characteristics, improving the accuracy and reliability of subsequent quantitative assessments.

[0021] Further, the step of denoising and superimposing several intrinsic mode function components based on the mode type of each intrinsic mode function component to obtain a reconstructed wind power signal includes: The intrinsic mode function components of the mode type that are noise-dominant mode types are taken as noise-dominant function components; the intrinsic mode function components of the mode type that are high-frequency fluctuation mode types are taken as high-frequency fluctuation function components; and the intrinsic mode function components of the mode type that are low-frequency trend mode types are taken as low-frequency trend function components. High-frequency noise in each of the noise-dominant function components is filtered out using a wavelet threshold denoising algorithm, thereby obtaining several denoised function components. The reconstructed wind power signal is obtained by superimposing and reconstructing the signal based on all the denoising function components, all the high-frequency fluctuation function components, and all the low-frequency trend function components.

[0022] In this implementation, wavelet threshold denoising technology is used to filter out high-frequency clutter from the identified noise-dominant component while retaining any potentially weak but effective information. This allows the removal of high-frequency clutter while preserving the potentially weak but effective signal components within the noise component. These components are then superimposed with the high-frequency fluctuation component and the low-frequency trend component to reconstruct the signal. This results in a reconstructed wind power signal that is free from noise interference and fully preserves the true fluctuation characteristics, significantly improving the signal-to-noise ratio and fidelity of the signal, and enhancing the accuracy of subsequent quantitative calculations and assessments of volatility and intermittency.

[0023] A second aspect of the present invention provides a quantitative assessment system for wind power fluctuation and intermittency, comprising: The data acquisition module is used to acquire historical active power data of the target wind farm and generate a signal to be decomposed based on the historical active power data. The variational mode decomposition parameter optimization module is used to construct an initial parameter combination for variational mode decomposition of the signal to be decomposed, and then iteratively optimize the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination. The variational mode decomposition module is used to perform variational mode decomposition on the signal to be decomposed based on the target parameter combination, thereby obtaining a number of intrinsic mode function components and the center frequency of each intrinsic mode function component; A dual-threshold mode segmentation module is used to obtain the sample entropy of each intrinsic mode function component; and to determine the mode type of each intrinsic mode function component based on the center frequency and sample entropy of each intrinsic mode function component according to a preset threshold segmentation rule. The signal reconstruction module is used to perform denoising and superposition reconstruction on several intrinsic mode function components based on the mode type of each intrinsic mode function component, thereby obtaining the reconstructed wind power signal; A volatility quantization module is used to obtain the power change rate of the reconstructed wind power signal in adjacent time intervals, and then analyze the volatility index of the reconstructed wind power signal based on the power change rate and the coefficient of variation formula. An intermittent quantization module is used to obtain the rated power of the wind turbines in the target wind farm, and then, based on the power change rate and the rated power of the wind turbines, analyzes the wind power ramp-up duty cycle of the reconstructed wind power signal according to a preset ramp-up event judgment rule. The results output analysis module is used to obtain the quantitative assessment results of wind power volatility based on the volatility index, and to obtain the quantitative assessment results of wind power intermittency based on the wind power ramp duty cycle.

[0024] Further, the step of acquiring historical active power data of the target wind farm and generating a signal to be decomposed based on the historical active power data includes: Outliers in the historical active power data are removed to obtain the first power time series data; The first power time series data is filled in using linear interpolation to obtain the second power time series data; Gaussian white noise is injected into the second power time series data to obtain the signal to be decomposed.

[0025] Furthermore, the construction of an initial parameter combination for variational mode decomposition of the signal to be decomposed, followed by iterative optimization of the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination, includes: Construct an initial total number of components and an initial penalty factor for variational mode decomposition of the signal to be decomposed, and then use the initial total number of components and the initial penalty factor as the initial parameter combination; A fitness function is constructed based on the signal to be decomposed, the initial total number of components, and the initial penalty factor. A sparrow search population is generated, wherein each individual in the sparrow search population corresponds to a set of random component totals and random penalty factors; Based on the sparrow search algorithm, the initial total number of components and the initial penalty factor are iteratively optimized in several rounds according to the fitness function and the sparrow search population until the final total number of components and the final penalty factor generated after a certain round of iterative optimization meet the preset optimization conditions. Then, the final total number of components and the final penalty factor are used as the target parameter combination. In any round of the aforementioned iterative optimization process: Based on the fitness function, obtain the fitness function value of the total number of random components and the random penalty factor corresponding to each individual in the sparrow search population; Based on the sparrow search algorithm, the two-dimensional position vector of each individual in the population is updated based on the fitness function value corresponding to each individual in the population, thereby obtaining the iterative search population; The sparrow search population is updated based on the iterative search population, and the final total number of components and the final penalty factor are obtained based on the iterative search population, thus ending this round of iterative optimization.

[0026] The method and system for quantitatively evaluating the volatility and intermittency of wind power provided by the present invention have at least the following advantages compared with the prior art: First, this invention achieves effective separation and identification of noise, fluctuation and trend components in wind power signals by using variational mode decomposition optimized by the sparrow search algorithm and combining it with the dual threshold criteria of sample entropy and center frequency. This solves the problems of strong subjectivity in parameter selection and easy mode aliasing in traditional methods, and significantly improves the accuracy of wind power signal decomposition.

[0027] Secondly, this invention employs a sparrow search algorithm to automatically find the optimal parameter combination for variational mode decomposition, replacing the traditional parameter setting method that relies on manual experience or simple grid search. This not only improves the accuracy and adaptability of signal decomposition and avoids mode aliasing caused by improper parameters, but also enhances the method's versatility and adaptability to data from different wind farms and time periods.

[0028] Furthermore, this invention constructs a dual-threshold mode division rule based on center frequency and sample entropy, which can accurately distinguish noise from effective high-frequency fluctuations, avoid the loss of feature information during the denoising process, and improve the fidelity of signal reconstruction.

[0029] Finally, by calculating the volatility index of the reconstructed signal and the wind power ramp-up duty cycle, this invention clearly defines and quantifies the volatility and intermittency characteristics of wind power from a physical perspective, overcoming the shortcomings of existing technologies where the concepts are confused and the indicators are singular. By implementing this invention, the inherent spatiotemporal distribution patterns of wind power output can be deeply explored, providing a more scientific and reliable quantitative basis and technical support for the optimized scheduling of power systems, the scientific allocation of reserve capacity, and the efficient absorption of wind power. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating a method for quantitatively evaluating the volatility and intermittency of wind power provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating another method for quantitatively evaluating the volatility and intermittency of wind power provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of a wind power fluctuation and intermittency quantitative evaluation system provided in an embodiment of the present invention. Detailed Implementation

[0031] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that the following detailed descriptions are exemplary and intended to provide further detailed explanation of the invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein in the specification is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings are used to distinguish different objects, not to describe a particular order.

[0032] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0033] Wind energy, as a clean and low-carbon energy source with abundant resources and mature technology, has become a core force in the global energy transition. Its large-scale grid connection is crucial for the safe dispatch and efficient absorption of power systems. However, wind power output is naturally characterized by significant fluctuations and intermittency due to complex factors such as spatiotemporal variations in wind speed and turbulence interference. Traditional quantitative methods struggle to accurately capture its nonlinear and non-stationary characteristics. Existing technologies often rely on empirically setting signal decomposition parameters such as VMD, which can easily lead to mode aliasing. Furthermore, they frequently confuse fluctuations with intermittency, lacking clear distinctions and targeted quantitative indicators. At the same time, some methods do not fully consider spurious feature interference from noise, and their fixed parameters and model structures lack dynamic adaptability, resulting in technical bottlenecks of static rigidity and limited accuracy.

[0034] Therefore, this embodiment constructs a fully adaptive quantification system for wind power volatility and intermittency. After preprocessing the collected active power data from the Supervisory Control and Data Acquisition (SCADA) system, including cleaning, missing value imputation, and noise adaptation, a sparrow optimization algorithm is introduced to adaptively optimize the VMD parameters. A dual-threshold rule is constructed based on sample entropy to achieve accurate separation of noise, volatility, and intermittency modes. Signal quality is further optimized by improving wavelet threshold denoising. Simultaneously, a complementary quantification index system based on the volatility index and wind power ramp duty cycle is constructed to characterize short-term random fluctuations and medium- to long-term ramp events, respectively. Furthermore, through adaptive parameter optimization and dynamic mode partitioning, the bottlenecks of traditional quantification methods—parameter empiricism and static structure—are overcome, significantly improving the accuracy and reliability of wind power characteristic quantification.

[0035] Please see Figure 1 and Figure 2 To address the technical problems of existing technologies, such as vague definitions of wind power volatility and intermittency, lack of systematic quantitative indicators, and insufficient exploration of the spatiotemporal distribution characteristics of wind power output and its deep correlation with meteorological factors, resulting in limited adaptability of quantitative results to the actual operational needs of the power system, the first embodiment of this invention provides a method for quantitatively evaluating wind power volatility and intermittency. This method integrates the sparrow search algorithm (SSA), variational mode decomposition (VMD), and sample entropy (SE) to quantify wind power volatility and intermittency. The aim is to achieve adaptive optimization of VMD parameters through SSA, and to accurately segment and denoise modes using SE, thereby constructing a scientific quantitative indicator system. This improves the accuracy and reliability of wind power characteristic analysis, providing support for power system dispatch, reserve capacity configuration, and efficient wind power consumption. The method specifically includes the following steps: Data Acquisition and Initialization: S1. Obtain historical active power data of the target wind farm, and generate a signal to be decomposed based on the historical active power data.

[0036] SSA adaptive optimization of VDM parameters: S2. Construct an initial parameter combination for variational mode decomposition of the signal to be decomposed, and then iteratively optimize the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination.

[0037] VMD signal decomposition: S3. Based on the target parameter combination, perform variational mode decomposition on the signal to be decomposed to obtain several intrinsic mode function components and the center frequency of each intrinsic mode function component.

[0038] Sample entropy calculation and dual-threshold mode segmentation: S4. Obtain the sample entropy of each intrinsic mode function component; determine the mode type of each intrinsic mode function component based on the center frequency and sample entropy of each intrinsic mode function component according to the preset threshold division rule.

[0039] Signal denoising and mode reconstruction: S5. Based on the mode type of each intrinsic mode function component, the intrinsic mode function components are denoised and superimposed to reconstruct the signal, thereby obtaining the reconstructed wind power signal.

[0040] Volatility and Intermittency Quantification: S6. Obtain the power change rate of the reconstructed wind power signal in adjacent time intervals, and then analyze the volatility index of the reconstructed wind power signal based on the power change rate and the coefficient of variation formula. Obtain the rated power of the wind turbines in the target wind farm, and then, based on the power change rate and the rated power of the wind turbines, analyze the wind power ramp-up duty cycle of the reconstructed wind power signal according to the preset ramp-up event judgment rules.

[0041] Results Output and Analysis: S7. Obtain the quantitative assessment result of wind power volatility based on the volatility index, and obtain the quantitative assessment result of wind power intermittency based on the wind power ramp duty cycle.

[0042] The aforementioned method for quantitatively assessing the volatility and intermittency of wind power firstly ensures data validity by preprocessing and constructing historical active power data. Secondly, it introduces a sparrow search algorithm to iteratively optimize key parameters of variational mode decomposition, effectively overcoming the shortcomings of traditional methods that rely on human experience for parameter selection, are highly subjective, and are prone to getting trapped in local optima. This significantly reduces mode aliasing and ensures the adaptability and accuracy of signal decomposition. Thirdly, by constructing a dual-threshold mode partitioning rule based on center frequency and sample entropy, it achieves the classification of noise-dominant components, high-frequency fluctuation components, and low-frequency trend components. The system achieves accurate identification by employing targeted denoising and superposition reconstruction. This effectively filters out noise interference while preserving the true fluctuation details of wind power to the greatest extent possible, solving the problem of traditional denoising methods easily losing high-frequency feature information. Finally, based on the reconstructed signal, the volatility index and wind power ramp duty cycle are calculated separately. From the two dimensions of time-domain statistical characteristics and event occurrence probability, the system achieves independent and scientific quantification of the volatility and intermittency characteristics of wind power, clarifies the physical connotation of volatility and intermittency, and solves the problems of ambiguous definitions and mixed indicators in existing technologies. This provides an accurate quantitative basis for power system dispatching.

[0043] Further, step S1, which involves acquiring historical active power data of the target wind farm and generating a signal to be decomposed based on the historical active power data, includes: Outliers in the historical active power data are removed to obtain the first power time series data; The first power time series data is filled in using linear interpolation to obtain the second power time series data; Gaussian white noise is injected into the second power time series data to obtain the signal to be decomposed.

[0044] In a preferred embodiment, high-temporal-resolution historical active power data from the wind farm's SCADA system is collected. Multi-dimensional preprocessing operations are performed on the collected data: first, outliers exceeding the rated power or with negative active power are removed to ensure data validity; then, linear interpolation is used to fill in missing data. The interpolation formula is as follows: In the formula, P ( i） Power at the missing moment t ( i） To correspond to the time interval, Gaussian white noise is injected at the end to simulate actual sensor interference, making the data more consistent with real working conditions.

[0045] In this embodiment, by removing outliers and performing linear interpolation to fill in the gaps in the original data, missing or erroneous data caused by reasons such as downtime, power outages, and measurement errors can be corrected, ensuring the integrity and accuracy of the input data and laying the foundation for subsequent precise decomposition. Injecting Gaussian white noise into the preprocessed data is a common technique based on signal decomposition theory. This aims to use the assistance of noise to help variational mode decomposition better avoid mode aliasing when processing complex nonlinear signals, thereby improving the stability and robustness of the decomposition.

[0046] In this implementation, outlier removal and linear interpolation imputation of the original data can correct missing or erroneous data caused by wind farm shutdowns, power curtailment, measurement errors, etc., ensuring the integrity and accuracy of the input power time series data. Furthermore, injecting Gaussian white noise into the processed data can simulate actual sensor interference, making the data more closely resemble real-world operating conditions.

[0047] Further, step S2, which involves constructing an initial parameter combination for variational mode decomposition of the signal to be decomposed, and then iteratively optimizing the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination, includes: Construct an initial total number of components and an initial penalty factor for variational mode decomposition of the signal to be decomposed, and then use the initial total number of components and the initial penalty factor as the initial parameter combination; A fitness function is constructed based on the signal to be decomposed, the initial total number of components, and the initial penalty factor. A sparrow search population is generated, wherein each individual in the sparrow search population corresponds to a set of random component totals and random penalty factors; Based on the sparrow search algorithm, the initial total number of components and the initial penalty factor are iteratively optimized in several rounds according to the fitness function and the sparrow search population until the final total number of components and the final penalty factor generated after a certain round of iterative optimization meet the preset optimization conditions. Then, the final total number of components and the final penalty factor are used as the target parameter combination. In any round of the aforementioned iterative optimization process: Based on the fitness function, obtain the fitness function value of the total number of random components and the random penalty factor corresponding to each individual in the sparrow search population; Based on the sparrow search algorithm, the two-dimensional position vector of each individual in the population is updated based on the fitness function value corresponding to each individual in the population, thereby obtaining the iterative search population; The sparrow search population is updated based on the iterative search population, and the final total number of components and the final penalty factor are obtained based on the iterative search population, thus ending this round of iterative optimization.

[0048] In a preferred embodiment, step S2 aims to find the optimal combination of parameters for VMD using a sparrow search algorithm. .in, K Represents the mode decomposition number, which is the total number of intrinsic mode functions (IMF) components obtained after VMD decomposition.

[0049] The modal decomposition number K determines the summation sign in the subsequent reconstruction formula. The upper limit.

[0050] and This represents the penalty factor, which physically controls the bandwidth of each IMF component. The larger the value, the narrower the modal bandwidth of the decomposed mode, and vice versa. These two parameters together determine the decomposition result. u k ( t） The shape and quality of the VMD decomposition are therefore set as the variables to be optimized in the SSA algorithm. The algorithm aims to optimize VMD decomposition performance and constructs a fitness function that integrates reconstruction error and modal independence constraints. The goal is to find the minimum value of this function. The formulas for calculating the fitness function and its components are as follows: In the formula, E recon The reconstruction error is the L2 norm of the difference between the original signal energy and the energy of the superposition of all components. x ( t） The signal to be decomposed specifically refers to the historical active power time series data after preprocessing in step 1 (anomaly removal, interpolation, and noise addition); u k ( t） for x ( t） The first result obtained after VMD decomposition under the current parameters k One IMF component; w 1 and w 2 represents the weighting coefficient. I total This is the total mutual information, used to quantify the degree of modal aliasing.

[0051] This represents all non-repeating modal component pairs ( u i , u j） Calculate mutual information and sum it.

[0052] Among them, two modal components u i and u j Mutual information between The calculation formula is defined as follows: In the above formula, H ( u i） and H ( u j） Components u i and u j marginal entropy, H ( u i , u j） Let be their joint entropy. For discrete-time series, entropy is... H ( X） The calculation formula is ,in p ( x） The probability distribution of the variable's values ​​is obtained from the statistical analysis of the data histogram. The larger the mutual information value, the stronger the correlation between the two components and the more severe the modal mixing; conversely, the lower the value, the better the independence.

[0053] The specific optimization process is as follows: Initialize the SSA population and set the population size to be... n s = 30, set the parameter constraint range as , In this embodiment, the two-dimensional position vector of each sparrow... X i Directly corresponds to a set of VMD parameter combinations .

[0054] First, calculate and sort the fitness values ​​of all individual sparrows, then select the top-fittest sparrows. n s Two individuals are designated as producers, and the rest as followers. The formula for updating the producer's position is: In the formula, t This represents the current iteration number. T The maximum number of iterations, A random number between (0, 1). n s The population size is set as described above.

[0055] Followers adjust their own state based on the discoverer's location to compete for food; their position update formula is: The current optimal producer position is represented by `rand()`, which generates a random number uniformly distributed in the interval (0, 1) to simulate the randomness of individual movement. Furthermore, a subset of individuals is randomly selected from the population as scouts for early warning updates, as shown in the formula: Through the iterative updates of the positions of producers, followers, and scouts described above, the VMD parameters are effectively updated. K and Dynamic adjustments. After each position update, the algorithm will adjust the new position accordingly. Substitute into the VMD model to generate a new u k ( t） Then calculate the fitness function. The optimal parameter combination with better fitness is retained. The iteration termination condition is set as reaching the maximum number of iterations or the difference between two consecutive optimal fitness values ​​being less than a set threshold. Once the termination condition is met, the optimal VMD parameters are output. .

[0056] In this embodiment, to address the problem that traditional variational mode decomposition parameters rely on manual experience for setting, a sparrow search algorithm is introduced for adaptive optimization. By constructing a fitness function that reflects the decomposition effect, the sparrow search algorithm simulates the discoverer-follower-watcher position update mechanism of sparrow population foraging behavior for iterative optimization. This efficient global search within the solution space can find the optimal total number of components and penalty factor for variational mode decomposition.

[0057] This process not only avoids the problem of excessive computation time in traditional grid search methods, but also eliminates the subjectivity and blindness of manually selecting parameters, overcoming the shortcomings of traditional methods such as strong subjectivity and low optimization efficiency. It ensures that variational mode decomposition can adapt the signal with the optimal parameter combination, thereby obtaining eigenmode function components with clear center frequencies and clear physical meanings, and greatly improving the efficiency and accuracy of signal decomposition.

[0058] It should also be noted that the aforementioned SSA-VMD parameter adaptive optimization mechanism, with minimizing reconstruction error as its core, integrates modal independence constraints to construct the objective function, and uses SSA to optimize the number of VMD modes. K With penalty factor Adaptive optimization solves the problems of subjectivity and modal aliasing in traditional parameter settings.

[0059] Further, step S3, which involves performing variational mode decomposition on the signal to be decomposed based on the target parameter combination to obtain several intrinsic mode function components and the center frequency of each intrinsic mode function component, includes: The signal to be decomposed is subjected to variational mode decomposition according to the variational mode decomposition algorithm to obtain the initial mode function components and the center frequency of each initial mode function component; The frequency domain update formula, center frequency update formula, and Lagrange multiplier update formula for modal components are constructed based on the alternating direction multiplier method. Based on the frequency domain update formula, the center frequency update formula, and the Lagrange multiplier update formula, the initial mode function components and the center frequency of each initial mode function component are iteratively updated in several rounds until the difference between the output results of two rounds of iterative updates is less than a preset threshold. Then, based on the output results of the two rounds of iterative updates, several intrinsic mode function components and the center frequency of each intrinsic mode function component are obtained.

[0060] In a preferred embodiment, the optimal parameters obtained in step S2 are... and Substituting the VMD model, the preprocessed wind power signal x ( t） Perform adaptive decomposition. The core idea of ​​the VMD algorithm is to construct a constrained variational problem, aiming to find... K A frequency around the center Finite bandwidth modal components u k ( t） This minimizes the sum of the estimated bandwidths of all components, while simultaneously ensuring that the sum of all components equals the original signal. x ( t） The expression for this constrained variational problem is: The constraints are: In the formula, Indicates time t The partial derivative is used to measure the smoothness of a signal, i.e., its bandwidth; where... For the Dirac function, This is a convolution operation; the expression is related to the fitness function in step S2. The relationship is that step S2 evaluates the parameters by continuously solving this variational problem. The quality of the problem is determined by the optimal parameters, and this step is to solve the problem using the determined optimal parameters to obtain the final decomposition result.

[0061] To solve the above constrained variational problem, a quadratic penalty factor is introduced. and Lagrange multipliers This is transformed into an unconstrained augmented Lagrangian function. : The components and parameters are updated iteratively using the alternating direction multiplier method. The frequency domain update formula for the modal components is shown below, which intuitively illustrates the parameters. right Impact: As can be seen from the above formula, Located in the denominator, it serves to adjust the filter bandwidth: The larger the value, the further the denominator is from the center frequency. The stronger the component suppression, the narrower the modal bandwidth. Center frequency The update formula is: The Lagrange multiplier update formula is: In the formula, Update the step size for the Lagrange multipliers, iterate until the difference between two adjacent decomposition results is less than a preset threshold, and then output. Each IMF component and its corresponding center frequency.

[0062] In this embodiment, the alternating direction multiplier method is employed to solve the constrained variational problem of variational mode decomposition. This transforms the complex variational constraint problem into a series of easily solvable subproblems. Through alternating iterations of frequency domain update, center frequency update, and Lagrange multiplier update, the optimal solution is found through convergence. This process ensures that, given the optimal parameter combination, the original signal can be accurately decomposed into a series of eigenmode functions with specific sparsity characteristics and finite bandwidth, and the center frequency of each component can be accurately calculated. This achieves the decomposition of complex non-stationary wind power signals into several sub-band signals with different center frequencies. Compared to traditional recursive decomposition algorithms, this method effectively suppresses endpoint effects and mode aliasing, accurately captures the dynamic characteristics of wind power at different time scales, and improves the convergence speed and accuracy of the decomposition results.

[0063] Further, step S4, obtaining the sample entropy of each of the intrinsic mode function components, includes: The standard deviation of the original wind power signal is obtained based on historical active power data, and then the tolerance threshold is obtained based on the standard deviation of the original wind power signal. For any of the aforementioned intrinsic mode function components: A phase space is constructed based on the intrinsic mode function component, such that each phase space vector in the phase space corresponds to a sequence value of the intrinsic mode function component; Obtain several spatial distances between any phase space vector and several other phase space vectors, and then analyze the proportion of several spatial distances that are less than the tolerance threshold to obtain the proportion parameter of the phase space vector; The sample entropy of the intrinsic mode function component is calculated based on the scaling parameter of each phase space vector.

[0064] In a preferred embodiment, for each IMF component output in step S3 u k (t) (wherein) Each of these is used to calculate its sample entropy (SE) to quantify the signal complexity.

[0065] First, set the parameters for calculating sample entropy: embedding dimension. m = 2, tolerance threshold r = 0.2 ;in This represents the standard deviation of the original wind power signal. The original signal standard deviation is used as the benchmark to ensure consistency in the measurement scale of each component. Data length. N This represents the number of sample points for the IMF component. The calculation process is as follows: Firstly, regarding the current situation... k IMF component sequence u k , build m A dimensional phase space vector is expressed as: Next, calculate the distance between any two vectors: Calculate the distance between each vector and other vectors. r Ratio: In the above formula I As an indicator function, it is ultimately expressed by the formula: The sample entropy values ​​of each IMF component are obtained.

[0066] In this embodiment, sample entropy, a dimensionless metric for measuring the complexity and irregularity of a time series, is introduced to quantify the randomness of each intrinsic mode function component. By constructing a phase space, calculating the vector distance, and comparing it with a tolerance threshold, sample entropy can effectively measure the probability of a signal generating new modes. The larger the sample entropy of a component, the stronger its randomness and irregularity. Furthermore, the tolerance threshold is determined based on the standard deviation of the original signal, ensuring consistency in the measurement scale for each component and objectively reflecting the complexity of each component under different wind power scenarios.

[0067] Further, step S4, which involves determining the mode type of each intrinsic mode function component based on the center frequency and sample entropy of each component according to a preset threshold division rule, includes: Set the sample entropy threshold and frequency threshold; For any of the aforementioned intrinsic mode function components: If the sample entropy of the intrinsic mode function component is greater than the sample entropy threshold, and the center frequency of the intrinsic mode function component is greater than or equal to the frequency threshold, then the intrinsic mode function component is a noise-dominant mode type. If the sample entropy of the intrinsic mode function component is less than or equal to the sample entropy threshold, and the center frequency of the intrinsic mode function component is greater than or equal to the frequency threshold, then the intrinsic mode function component is a high-frequency fluctuation mode type. If the center frequency of the intrinsic mode function component is less than the frequency threshold, then the intrinsic mode function component is a low-frequency trend mode type.

[0068] In one specific embodiment, the sample entropy value of the IMF component is combined with the corresponding center frequency. f k Construct a dual-threshold partitioning rule: Set a sample entropy threshold SE th and frequency threshold f th , will satisfy SE( u k SE th and The components are identified as noise-dominant modes; these components are highly complex and have high frequencies; [the following will satisfy...] and The components are determined to be high-frequency fluctuation modes; those that satisfy... The component is identified as a low-frequency trend mode. The specific mathematical expression is: Noise-dominant mode: .

[0069] High-frequency fluctuation modes: .

[0070] Low-frequency trend mode: .

[0071] In this embodiment, considering the physical characteristics of wind power signals, a dual-threshold classification rule based on center frequency and sample entropy is proposed to classify the decomposed modal components in a physical sense. Noise signals typically exhibit high frequency and high entropy, displaying high randomness; while the high-frequency fluctuations of wind power, though high in frequency, are influenced by meteorological factors and exhibit certain regularities; low-frequency components reflect the overall trend of the system. Therefore, in this implementation, high-frequency components with high sample entropy are identified as noise-dominant; high-frequency components with low sample entropy are identified as high-frequency fluctuation components, representing the main fluctuation components of wind power; and low-frequency components represent the long-term trend of wind power output. This rule effectively identifies noise components mixed in with the useful signal, avoiding the risk of misjudging useful high-frequency fluctuation features as noise and discarding them when filtering solely based on frequency, thus ensuring that the reconstructed signal retains complete wind power fluctuation characteristics and improving the accuracy and reliability of subsequent quantitative assessments.

[0072] It should also be noted that the above dual-threshold mode segmentation and denoising strategy combines sample entropy and center frequency to construct dual-threshold rules, achieving accurate separation of noise, fluctuation, and intermittent modes. Combined with improved wavelet threshold denoising, it effectively filters out sensor interference and short-term turbulence noise, while preserving the true characteristics of the signal.

[0073] Further, step S5, which involves denoising and superimposing several intrinsic mode function components based on the mode type of each intrinsic mode function component to obtain a reconstructed wind power signal, includes: The intrinsic mode function components of the mode type that are noise-dominant mode types are taken as noise-dominant function components; the intrinsic mode function components of the mode type that are high-frequency fluctuation mode types are taken as high-frequency fluctuation function components; and the intrinsic mode function components of the mode type that are low-frequency trend mode types are taken as low-frequency trend function components. High-frequency noise in each of the noise-dominant function components is filtered out using a wavelet threshold denoising algorithm, thereby obtaining several denoised function components. The reconstructed wind power signal is obtained by superimposing and reconstructing the signal based on all the denoising function components, all the high-frequency fluctuation function components, and all the low-frequency trend function components.

[0074] In a preferred embodiment, based on the dual-threshold mode segmentation result of step S4, differentiated processing strategies are adopted for different types of IMF components. Specifically, only IMF components identified as noise-dominant mode types in step 4 are specifically processed using an improved wavelet threshold denoising algorithm to filter out high-frequency clutter and retain potentially useful information, while IMF components belonging to other mode types remain unchanged. The improved wavelet threshold function is constructed as follows: In the above formula, ω These are wavelet coefficients. λ For the threshold, k = 0.5 is an adjustment parameter to balance the noise reduction effect and signal fidelity. Represents a sign function, when When the value is 1, The value is -1 when The value is 0 when the noise level is low, which is used to keep the sign of the denoised coefficients consistent with the original coefficients.

[0075] The denoised components are then superimposed with the unprocessed high-frequency fluctuation components and low-frequency trend components to reconstruct the purified wind power signal. .

[0076] In this embodiment, wavelet threshold denoising technology is used to filter out high-frequency clutter from the identified noise-dominant component while retaining any potentially weak but valid information. This allows the removal of high-frequency clutter while preserving the potentially weak but valid signal components within the noise component. These components are then superimposed with the high-frequency fluctuation component and the low-frequency trend component to reconstruct the signal. This results in a reconstructed wind power signal that is free from noise interference and fully preserves the true fluctuation characteristics, significantly improving the signal-to-noise ratio and fidelity of the signal, and enhancing the accuracy of subsequent quantitative calculations and evaluations of volatility and intermittency.

[0077] Further, in a preferred embodiment, step S6, which involves obtaining the power change rate of the reconstructed wind power signal within adjacent time intervals and then analyzing the volatility index of the reconstructed wind power signal based on the power change rate and the coefficient of variation formula, includes: Volatility quantification employs a volatility index (VI) based on the coefficient of variation, first calculating the rate of change of power between adjacent time intervals. Then through the formula Calculate the volatility index, where M is the standard deviation of the power change rate (M is the number of samples in the sliding window). This represents the average power within the window.

[0078] Further, in a preferred embodiment, step S6, which involves obtaining the rated power of the wind turbines in the target wind farm and then analyzing the wind power ramp-up duty cycle of the reconstructed wind power signal based on the power change rate and the rated power of the wind turbines according to a preset ramp-up event determination rule, includes: Intermittent quantification uses the wind power ramp (DRWPR) duty ratio, and first sets the ramp event determination rules: in, P R This refers to the rated power of the fan. This represents the power change rate between adjacent time intervals. The wind power ramp duty cycle DRWPR is then calculated using the following formula: in For time intervals, T obs Let be the length of the observation window, and numerator be the total duration of the climbing events within the window.

[0079] Further, in a preferred embodiment, step S7, which involves obtaining a quantitative assessment result of wind power volatility based on the volatility index and obtaining a quantitative assessment result of wind power intermittency based on the wind power ramp duty cycle, includes: A 4-hour non-overlapping sliding window is used to traverse the entire reconstructed signal sequence, outputting the volatility index sequence and wind power ramp duty cycle (DRWPR) sequence corresponding to each window. At the same time, based on the wind power ramp duty cycle (DRWPR) value, the distribution results of strong (DRWPR≥40%), medium (20%≤DRWPR<40%), and weak (DRWPR<20%) intermittent periods are divided, providing direct quantitative decision-making basis for power system optimization scheduling, reserve capacity configuration, and efficient wind power consumption.

[0080] The aforementioned volatility index based on the coefficient of variation complements the wind power ramp duty cycle DRWPR, quantifying short-term random fluctuations and medium- to long-term ramp events respectively, clearly distinguishing the two major characteristics, and solving the problem of existing technologies conflating wind power volatility with intermittency.

[0081] The beneficial effects of the wind power volatility and intermittency quantitative assessment method provided in the first embodiment of this invention are as follows: It constructs a complete analysis system of "parameter optimization – signal purification – precise quantification". By optimizing VMD parameters through SSA, the subjectivity and inefficiency of traditional methods are avoided; the combination of sample entropy dual-threshold mode division and improved wavelet denoising effectively eliminates spurious features and preserves the physical essence of the signal; the volatility index and wind power ramp duty cycle DRWPR form a complementary index system, achieving precise differentiation and quantification of the two major characteristics; the method parameters can be flexibly adjusted according to the actual operating conditions of the wind farm, exhibiting strong adaptability and providing reliable decision support for the power system.

[0082] It should be understood that in the quantitative evaluation method for wind power fluctuation and intermittency disclosed in the first embodiment of the present invention, the parameter constraint range and threshold can be adjusted according to the actual operating conditions such as wind farm data resolution and power level, and are applicable to wind power characteristic analysis in different scenarios, with strong engineering practicality.

[0083] Please refer to Figure 3 The second aspect of the present invention provides a quantitative assessment system for wind power fluctuation and intermittency, comprising: The data acquisition module 100 is used to acquire historical active power data of the target wind farm and generate a signal to be decomposed based on the historical active power data. The variational mode decomposition parameter optimization module 200 is used to construct an initial parameter combination for variational mode decomposition of the signal to be decomposed, and then iteratively optimize the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination. The variational mode decomposition module 300 is used to perform variational mode decomposition on the signal to be decomposed based on the target parameter combination, thereby obtaining a number of intrinsic mode function components and the center frequency of each intrinsic mode function component; The dual-threshold mode segmentation module 400 is used to obtain the sample entropy of each intrinsic mode function component; and to determine the mode type of each intrinsic mode function component based on the center frequency and sample entropy of each intrinsic mode function component according to a preset threshold segmentation rule. The signal reconstruction module 500 is used to perform denoising and superposition reconstruction on several intrinsic mode function components based on the mode type of each intrinsic mode function component, thereby obtaining a reconstructed wind power signal; The volatility quantization module 600 is used to obtain the power change rate of the reconstructed wind power signal in adjacent time intervals, and then analyze the volatility index of the reconstructed wind power signal based on the power change rate and the coefficient of variation formula. Intermittent quantization module 700 is used to obtain the rated power of the wind turbines in the target wind farm, and then, based on the power change rate and the rated power of the wind turbines, analyze the wind power ramp-up duty cycle of the reconstructed wind power signal according to the preset ramp-up event judgment rules. The result output analysis module 800 is used to obtain the quantitative evaluation result of wind power volatility based on the volatility index, and to obtain the quantitative evaluation result of wind power intermittency based on the wind power ramp duty cycle.

[0084] Further, the step of acquiring historical active power data of the target wind farm and generating a signal to be decomposed based on the historical active power data includes: Outliers in the historical active power data are removed to obtain the first power time series data; The first power time series data is filled in using linear interpolation to obtain the second power time series data; Gaussian white noise is injected into the second power time series data to obtain the signal to be decomposed.

[0085] Furthermore, the construction of an initial parameter combination for variational mode decomposition of the signal to be decomposed, followed by iterative optimization of the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination, includes: Construct an initial total number of components and an initial penalty factor for variational mode decomposition of the signal to be decomposed, and then use the initial total number of components and the initial penalty factor as the initial parameter combination; A fitness function is constructed based on the signal to be decomposed, the initial total number of components, and the initial penalty factor. A sparrow search population is generated, wherein each individual in the sparrow search population corresponds to a set of random component totals and random penalty factors; Based on the sparrow search algorithm, the initial total number of components and the initial penalty factor are iteratively optimized in several rounds according to the fitness function and the sparrow search population until the final total number of components and the final penalty factor generated after a certain round of iterative optimization meet the preset optimization conditions. Then, the final total number of components and the final penalty factor are used as the target parameter combination. In any round of the aforementioned iterative optimization process: Based on the fitness function, obtain the fitness function value of the total number of random components and the random penalty factor corresponding to each individual in the sparrow search population; Based on the sparrow search algorithm, the two-dimensional position vector of each individual in the population is updated based on the fitness function value corresponding to each individual in the population, thereby obtaining the iterative search population; The sparrow search population is updated based on the iterative search population, and the final total number of components and the final penalty factor are obtained based on the iterative search population, thus ending this round of iterative optimization.

[0086] The method and system for quantitatively evaluating the volatility and intermittency of wind power provided by the present invention have at least the following advantages compared with the prior art: First, this invention adopts a hierarchical collaborative architecture of "parameter optimization - modality segmentation - precise quantization," breaking down the quantization task into three core stages: parameter optimization, signal purification, and characteristic quantization. Each stage is equipped with a dedicated technical solution. Precise modality separation is achieved through SSA adaptive optimization of VMD parameters and dual-threshold modality classification using sample entropy. Two indicators are used to quantify volatility and intermittency respectively, enabling each stage to address its corresponding technical challenges and significantly improving the accuracy and discriminative power of the quantization results. Simultaneously, each stage is sequentially connected according to the "data input - processing - output" flow, avoiding redundant calculations and greatly improving the efficiency of quantitative analysis.

[0087] Secondly, this invention establishes a close correlation between parameter optimization, modal partitioning, and quantization calculation through a multi-stage collaborative and feature-deep fusion mechanism. Specifically, the VMD parameters optimized by SSA provide a high-quality decomposition basis for modal partitioning, and the pure features after modal separation provide reliable data support for quantization index calculation. Conversely, the rationality of the quantization results verifies the effectiveness of parameter optimization and modal partitioning, forming a closed-loop collaboration of "input-processing-feedback." This multi-stage linkage not only enhances the integrity and synergy of the entire system and solves the technical problems of "parameter empiricalization, feature confusion, and index simplification" in traditional quantization methods, but also improves the system's adaptability to different wind farm conditions and noise levels, ensuring the stability and complementarity of the quantization results.

[0088] Finally, this invention proposes a parameter adaptive optimization and threshold dynamic adjustment mechanism, which uses SSA to optimize the number of modes K and the penalty factor of VMD in real time. By flexibly adjusting the sample entropy threshold and center frequency threshold based on the actual data characteristics of wind farms, the system can dynamically adapt to the wind power signal characteristics under different scenarios. This breaks through the static and rigid barriers of fixed parameters and empirically based thresholds in traditional quantification methods, enabling the system to have dynamic optimization capabilities. It can autonomously adjust core parameters according to data characteristics and changes in operating conditions, further improving the universality and engineering applicability of quantification methods.

[0089] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0090] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; however, any combination of these technical features that does not contradict each other should be considered within the scope of this specification.

[0091] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the concept of this application, and these improvements and substitutions should also be considered within the scope of protection of this invention. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A method for quantitatively evaluating the volatility and intermittency of wind power, characterized in that, include: Obtain historical active power data of the target wind farm, and generate a signal to be decomposed based on the historical active power data; An initial parameter combination is constructed for variational mode decomposition of the signal to be decomposed, and then the initial parameter combination is iteratively optimized based on the sparrow search algorithm to obtain the target parameter combination. Based on the target parameter combination, variational mode decomposition is performed on the signal to be decomposed to obtain several intrinsic mode function components and the center frequency of each intrinsic mode function component; Obtain the sample entropy of each of the intrinsic mode function components; According to the preset threshold division rules, the mode type of each intrinsic mode function component is determined based on the center frequency and sample entropy of each intrinsic mode function component; Based on the mode type of each intrinsic mode function component, the intrinsic mode function components are denoised and superimposed to reconstruct the signal, thereby obtaining the reconstructed wind power signal. The power change rate of the reconstructed wind power signal within adjacent time intervals is obtained, and then the volatility index of the reconstructed wind power signal is analyzed based on the power change rate and the coefficient of variation formula. Obtain the rated power of the wind turbines in the target wind farm, and then, based on the power change rate and the rated power of the wind turbines, analyze the wind power ramp-up duty cycle of the reconstructed wind power signal according to the preset ramp-up event judgment rules. The volatility index is used to obtain the quantitative assessment results of wind power volatility, and the wind power ramp duty cycle is used to obtain the quantitative assessment results of wind power intermittency.

2. The method for quantitatively evaluating the volatility and intermittency of wind power according to claim 1, characterized in that, The process of acquiring historical active power data of the target wind farm and generating a signal to be decomposed based on the historical active power data includes: Outliers in the historical active power data are removed to obtain the first power time series data; The first power time series data is filled in using linear interpolation to obtain the second power time series data; Gaussian white noise is injected into the second power time series data to obtain the signal to be decomposed.

3. The method for quantitatively evaluating the volatility and intermittency of wind power according to claim 1, characterized in that, The process involves constructing an initial parameter combination for variational mode decomposition of the signal to be decomposed, and then iteratively optimizing the initial parameter combination based on a sparrow search algorithm to obtain a target parameter combination, including: Construct an initial total number of components and an initial penalty factor for variational mode decomposition of the signal to be decomposed, and then use the initial total number of components and the initial penalty factor as the initial parameter combination; A fitness function is constructed based on the signal to be decomposed, the initial total number of components, and the initial penalty factor. A sparrow search population is generated, wherein each individual in the sparrow search population corresponds to a set of random component totals and random penalty factors; Based on the sparrow search algorithm, the initial total number of components and the initial penalty factor are iteratively optimized in several rounds according to the fitness function and the sparrow search population until the final total number of components and the final penalty factor generated after a certain round of iterative optimization meet the preset optimization conditions. Then, the final total number of components and the final penalty factor are used as the target parameter combination. In any round of the aforementioned iterative optimization process: Based on the fitness function, obtain the fitness function value of the total number of random components and the random penalty factor corresponding to each individual in the sparrow search population; Based on the sparrow search algorithm, the two-dimensional position vector of each individual in the population is updated based on the fitness function value corresponding to each individual in the population, thereby obtaining the iterative search population; The sparrow search population is updated based on the iterative search population, and the final total number of components and the final penalty factor are obtained based on the iterative search population, thus ending this round of iterative optimization.

4. The method for quantitatively evaluating the volatility and intermittency of wind power according to claim 1, characterized in that, The variational mode decomposition of the signal to be decomposed based on the target parameter combination, thereby obtaining several intrinsic mode function components and the center frequency of each intrinsic mode function component, includes: The signal to be decomposed is subjected to variational mode decomposition according to the variational mode decomposition algorithm to obtain the initial mode function components and the center frequency of each initial mode function component; The frequency domain update formula, center frequency update formula, and Lagrange multiplier update formula for modal components are constructed based on the alternating direction multiplier method. Based on the frequency domain update formula, the center frequency update formula, and the Lagrange multiplier update formula, the initial mode function components and the center frequency of each initial mode function component are iteratively updated in several rounds until the difference between the output results of two rounds of iterative updates is less than a preset threshold. Then, based on the output results of the two rounds of iterative updates, several intrinsic mode function components and the center frequency of each intrinsic mode function component are obtained.

5. The method for quantitatively evaluating the volatility and intermittency of wind power according to claim 1, characterized in that, The step of obtaining the sample entropy of each intrinsic mode function component includes: The standard deviation of the original wind power signal is obtained based on historical active power data, and then the tolerance threshold is obtained based on the standard deviation of the original wind power signal. For any of the aforementioned intrinsic mode function components: A phase space is constructed based on the intrinsic mode function component, such that each phase space vector in the phase space corresponds to a sequence value of the intrinsic mode function component; Obtain several spatial distances between any phase space vector and several other phase space vectors, and then analyze the proportion of several spatial distances that are less than the tolerance threshold to obtain the proportion parameter of the phase space vector; The sample entropy of the intrinsic mode function component is calculated based on the scaling parameter of each phase space vector.

6. The method for quantitatively evaluating the volatility and intermittency of wind power according to claim 1, characterized in that, The step of determining the mode type of each intrinsic mode function component based on the center frequency and sample entropy of each intrinsic mode function component according to a preset threshold division rule includes: Set the sample entropy threshold and frequency threshold; For any of the aforementioned intrinsic mode function components: If the sample entropy of the intrinsic mode function component is greater than the sample entropy threshold, and the center frequency of the intrinsic mode function component is greater than or equal to the frequency threshold, then the intrinsic mode function component is a noise-dominant mode type. If the sample entropy of the intrinsic mode function component is less than or equal to the sample entropy threshold, and the center frequency of the intrinsic mode function component is greater than or equal to the frequency threshold, then the intrinsic mode function component is a high-frequency fluctuation mode type. If the center frequency of the intrinsic mode function component is less than the frequency threshold, then the intrinsic mode function component is a low-frequency trend mode type.

7. The method for quantitatively evaluating the volatility and intermittency of wind power according to claim 6, characterized in that, The process of denoising and superimposing several intrinsic mode function components based on the mode type of each intrinsic mode function component to obtain a reconstructed wind power signal includes: The intrinsic mode function components of the mode type that are noise-dominant mode types are taken as noise-dominant function components; the intrinsic mode function components of the mode type that are high-frequency fluctuation mode types are taken as high-frequency fluctuation function components; and the intrinsic mode function components of the mode type that are low-frequency trend mode types are taken as low-frequency trend function components. High-frequency noise in each of the noise-dominant function components is filtered out using a wavelet threshold denoising algorithm, thereby obtaining several denoised function components. The reconstructed wind power signal is obtained by superimposing and reconstructing the signal based on all the denoising function components, all the high-frequency fluctuation function components, and all the low-frequency trend function components.

8. A quantitative evaluation system for wind power fluctuation and intermittency, characterized in that, include: The data acquisition module is used to acquire historical active power data of the target wind farm and generate a signal to be decomposed based on the historical active power data. The variational mode decomposition parameter optimization module is used to construct an initial parameter combination for variational mode decomposition of the signal to be decomposed, and then iteratively optimize the initial parameter combination based on the sparrow search algorithm to obtain the target parameter combination. The variational mode decomposition module is used to perform variational mode decomposition on the signal to be decomposed based on the target parameter combination, thereby obtaining a number of intrinsic mode function components and the center frequency of each intrinsic mode function component; A dual-threshold mode segmentation module is used to obtain the sample entropy of each intrinsic mode function component; According to the preset threshold division rules, the mode type of each intrinsic mode function component is determined based on the center frequency and sample entropy of each intrinsic mode function component; The signal reconstruction module is used to perform denoising and superposition reconstruction on several intrinsic mode function components based on the mode type of each intrinsic mode function component, thereby obtaining the reconstructed wind power signal; A volatility quantization module is used to obtain the power change rate of the reconstructed wind power signal in adjacent time intervals, and then analyze the volatility index of the reconstructed wind power signal based on the power change rate and the coefficient of variation formula. An intermittent quantization module is used to obtain the rated power of the wind turbines in the target wind farm, and then, based on the power change rate and the rated power of the wind turbines, analyzes the wind power ramp-up duty cycle of the reconstructed wind power signal according to a preset ramp-up event judgment rule. The results output analysis module is used to obtain the quantitative assessment results of wind power volatility based on the volatility index, and to obtain the quantitative assessment results of wind power intermittency based on the wind power ramp duty cycle.

9. A quantitative evaluation system for wind power fluctuation and intermittency according to claim 8, characterized in that, The process of acquiring historical active power data of the target wind farm and generating a signal to be decomposed based on the historical active power data includes: Outliers in the historical active power data are removed to obtain the first power time series data; The first power time series data is filled in using linear interpolation to obtain the second power time series data; Gaussian white noise is injected into the second power time series data to obtain the signal to be decomposed.

10. A quantitative evaluation system for wind power fluctuation and intermittency according to claim 8, characterized in that, The process involves constructing an initial parameter combination for variational mode decomposition of the signal to be decomposed, and then iteratively optimizing the initial parameter combination based on a sparrow search algorithm to obtain a target parameter combination, including: Construct an initial total number of components and an initial penalty factor for variational mode decomposition of the signal to be decomposed, and then use the initial total number of components and the initial penalty factor as the initial parameter combination; A fitness function is constructed based on the signal to be decomposed, the initial total number of components, and the initial penalty factor. A sparrow search population is generated, wherein each individual in the sparrow search population corresponds to a set of random component totals and random penalty factors; Based on the sparrow search algorithm, the initial total number of components and the initial penalty factor are iteratively optimized in several rounds according to the fitness function and the sparrow search population until the final total number of components and the final penalty factor generated after a certain round of iterative optimization meet the preset optimization conditions. Then, the final total number of components and the final penalty factor are used as the target parameter combination. In any round of the aforementioned iterative optimization process: Based on the fitness function, obtain the fitness function value of the total number of random components and the random penalty factor corresponding to each individual in the sparrow search population; Based on the sparrow search algorithm, the two-dimensional position vector of each individual in the population is updated based on the fitness function value corresponding to each individual in the population, thereby obtaining the iterative search population; The sparrow search population is updated based on the iterative search population, and the final total number of components and the final penalty factor are obtained based on the iterative search population, thus ending this round of iterative optimization.

Images