A probabilistic assessment method, system, and equipment for broadband oscillation risk in wind power grid-connected systems.

By combining the soft partition membership function and the global node admittance matrix, the spatiotemporal uncertainty problem of broadband oscillation risk assessment in wind power grid-connected systems is solved, and efficient and accurate risk quantification assessment and early warning are achieved.

CN123073787APending Publication Date: 2026-11-10ELECTRIC POWER RES INST OF STATE GRID ZHEJIANG ELECTRIC POWER COMAPNY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611578592.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-13
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing technologies for wind power grid-connected systems, the assessment of broadband oscillation risk suffers from insufficient handling of spatiotemporal uncertainties, leading to assessment results that deviate from reality and fail to provide accurate early warnings. Furthermore, existing methods are computationally complex or inefficient, failing to meet the needs of online assessment.

Method used

We employ a soft-partition membership function to weight and fuse the Copula sub-models, constructing a continuously differentiable wind speed-power joint probability model. We then solve for the dominant oscillation mode using the global node admittance matrix, establishing an analytical mapping from the time-varying power probability density to the dominant oscillation mode, and achieving a dynamic quantitative assessment of broadband oscillation risk.

Benefits of technology

It enables efficient and accurate assessment of broadband oscillation risk in wind power grid-connected systems, provides reliable early warning for the system, reduces computational complexity, and improves assessment accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123073787A_ABST
    Figure CN123073787A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of new energy grid-connected stability analysis technology, and discloses a method, system, and equipment for probabilistic assessment of broadband oscillation risk in wind power grid-connected systems. The method includes: acquiring the future wind speed time series scenario of wind turbines; dividing the wind speed data domain into multiple operating intervals based on the power characteristics of the wind turbines; using a soft partition membership function to weight and fuse the Copula sub-models of each operating interval to obtain a continuously differentiable wind speed-power joint probability model, and deriving the time-varying power probability density of each wind turbine; constructing the global node admittance matrix of the wind power grid-connected system and solving for the dominant oscillation mode; based on the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine, mapping the time-varying power probability density to the time-varying probability density of the real part of the dominant oscillation mode, and integrating it over a preset instability interval to obtain the broadband oscillation instability probability of the wind power grid-connected system at future times.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of new energy grid-connected stability analysis technology, specifically involving a probabilistic assessment method, system and equipment for broadband oscillation risk of wind power grid-connected systems. Background Technology

[0002] In wind power grid-connected systems, the complex dynamic interactions between the wind turbine grid-connected converter and the power grid or converter station lead to high-order nonlinearity and multi-time-scale coupling characteristics, resulting in frequent and prone-to-occurrence broadband oscillations. Broadband oscillations refer to sustained oscillations in the system with frequencies spanning from near the power frequency to hundreds of hertz or even wider. Once they occur, they seriously threaten the safe and stable operation of the power system, and may even trigger large-scale grid disconnections and major engineering accidents. Therefore, accurately assessing the broadband oscillation risks of wind power grid-connected systems is crucial.

[0003] The source-end input of wind power generation exhibits significant spatiotemporal uncertainties, which directly impacts the broadband oscillation risk of grid-connected systems. On one hand, wind speed experiences strong random fluctuations over time; on the other hand, influenced by factors such as wake effects, wind shear, time delays, and the spatial distribution of turbines, the actual output of different wind turbines at the same wind speed measured by a meteorological tower shows significant spatial differences and correlations. If stability assessments rely solely on a single deterministic prediction result, or if the spatiotemporal correlation of the output of each wind turbine is ignored in probabilistic assessments, the assessment results of broadband oscillation risk will deviate significantly from reality (too optimistic or too conservative), failing to provide reliable early warnings for the system.

[0004] In characterizing the correlation between wind speed and turbine output, existing technologies typically rely on a single Copula function to establish a joint probability model. However, this approach has significant drawbacks: the wind speed-output correlation structure varies greatly across different operating ranges, such as the maximum power point tracking (MPPT) zone, the power-limited zone, and the extreme wind speed zone. Using a single global Copula model fails to adequately account for the correlation characteristics of each operating range; and if a hard segmentation (i.e., hard partitioning) is applied according to the operating range, the joint probability density will experience abrupt changes or discontinuities at the partition boundaries. This discontinuity in probability density not only violates the true physical laws governing wind turbine output but also severely compromises the mathematical accuracy of subsequent probability mapping to system oscillation modes.

[0005] Furthermore, existing broadband stability assessment methods, such as the traditional complex torque coefficient method and time-domain simulation method, rely on fixed, deterministic operating points and cannot reflect the dynamic impact of operational uncertainties on stability margins. While the Monte Carlo sampling method offers high accuracy, it relies on massive amounts of repeated simulations, resulting in extremely high computational costs and failing to meet the needs of online assessment or efficient early warning. Probabilistic small-signal analysis based on semi-invariants or series expansions is efficient but relies on distribution assumptions, limiting its ability to characterize the non-Gaussian properties of new energy sources. More critically, current probabilistic stability assessment methods lack an effective means to directly and analytically map the time-varying power probability density of multiple units to the system's oscillation mode probability density. This makes it difficult to continuously, efficiently, and accurately quantify the dynamic process of broadband oscillation instability risk evolving with future wind and operating conditions. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method, system, and device for probabilistic assessment of broadband oscillation risk in wind power grid-connected systems. This method transfers the spatiotemporal uncertainty of wind power output to the probability distribution of the system's dominant oscillation mode, thereby achieving dynamic quantitative assessment of broadband oscillation risk as it evolves with future operating conditions.

[0007] This invention provides the following technical solution: The primary objective of this invention is to provide a probabilistic assessment method for broadband oscillation risk in wind power grid-connected systems, comprising: The future wind speed time series scenario of wind turbines is obtained. Based on the power characteristics of wind turbines, multiple operating intervals are divided in the wind speed data domain. The Copula sub-models of each operating interval are weighted and fused using the soft partition membership function to obtain a continuously differentiable wind speed-power joint probability model. The time-varying power probability density of each wind turbine is derived based on the future wind speed time series scenario. Construct the global node admittance matrix of the wind power grid-connected system, and solve the dominant oscillation mode based on the global node admittance matrix; Based on the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine, the time-varying power probability density is mapped to the time-varying probability density of the real part of the dominant oscillation mode. By integrating the time-varying probability density of the real part of the dominant oscillation mode over a preset instability interval, the broadband oscillation instability probability of the wind power grid-connected system at future times is obtained, so as to realize the probabilistic assessment of the broadband oscillation risk of the system.

[0008] By eliminating the abrupt change in probability density at the boundary of the operating range in traditional hard partitioning modeling through the soft partitioning membership function, the continuous differentiability of the ring wind speed-power joint probability model is ensured throughout the entire domain, providing high-precision probability input for subsequent stability analysis. At the same time, by establishing an analytical mapping from the time-varying power probability density to the time-varying probability density of the real part of the dominant oscillation mode, an efficient quantitative assessment of the dynamic evolution of broadband oscillation instability risk with operating conditions is achieved.

[0009] Preferably, the weighted fusion of Copula sub-models for each operating interval using the soft partition membership function includes: Based on the power curve of the wind turbine, multiple operating intervals are divided in the wind speed data domain; In each operating range, a Copula sub-model is fitted between wind speed and unit active power; A soft partition membership function is constructed from the transition function at the boundary of adjacent operating intervals, so that the sum of the membership degrees of each operating interval under the same wind speed data is 1; The optimal Copula sub-models in each operating range are weighted and fused according to the soft partition membership function to obtain a wind speed-power joint probability model that is continuously differentiable in the wind speed data domain.

[0010] By using a soft partition membership function to weight and fuse the optimal Copula sub-models in each operating interval, the sum of the contributions of each operating interval remains constant under the same observation data. This fundamentally avoids step distortion of probability density at partition switching points and improves the fidelity of the joint probability model in fitting the actual operating characteristics of the equipment.

[0011] Preferably, the step of weighting and fusing the optimal Copula sub-models in each operating interval according to the soft partition membership function to obtain the wind speed-power joint probability model that is continuously differentiable in the wind speed data domain includes: Nonparametric kernel density estimation was used to fit the marginal probability densities of wind speed data and wind turbine power respectively, and the corresponding cumulative distribution function was obtained by integration; Based on the density function, cumulative distribution function, marginal probability density, and soft partition membership function of the optimal Copula sub-model, the joint probability model of wind speed and power is calculated.

[0012] Preferably, the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine, which maps the time-varying power probability density to the time-varying probability density of the real part of the dominant oscillation mode, includes: The dominant oscillation mode is solved by the characteristic equation of the global node admittance matrix, and the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine is obtained according to the implicit function differentiation rule. The change in the real part of the dominant oscillation mode near the reference operating point is expressed as a weighted sum of the power offset of each wind turbine and the corresponding sensitivity coefficient; Based on the weighted sum of the power offset of each wind turbine and the corresponding sensitivity coefficient, the time-varying power probability density is mapped to the real part of the dominant oscillation mode, and the time-varying probability density of the real part of the dominant oscillation mode is obtained by kernel density estimation.

[0013] Preferably, the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine, obtained according to the implicit function differentiation rule, includes: By expanding the derivative of the determinant with respect to the matrix parameters, the trace of the product of the adjoint matrix of the global node admittance matrix and the partial derivative matrix of the global node admittance matrix with respect to the wind turbine power is calculated, so as to obtain the partial derivative of the characteristic equation with respect to the wind turbine power. Based on the partial derivatives of the characteristic equation with respect to the power of the wind turbine and with respect to the complex frequency variable, the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine is calculated.

[0014] By using the implicit function differentiation rule to establish a local linear analytical relationship between the system eigenvalues ​​and the power of the wind turbine, the complex nonlinear stability analysis problem is transformed into a linear transformation problem of probability density, which significantly reduces the computational complexity while ensuring the evaluation accuracy.

[0015] Preferably, the acquisition of the future wind speed time series scenario of the wind turbine includes: Input historical wind speed sequences into the time-series scene generation model to obtain a set of future wind speed time-series scenes, each scene containing wind speed sequences within a preset time period; Calculate the mean wind speed and standard deviation of wind speed for each scenario in the future wind speed time series scenario set; We obtained a comprehensive evaluation index by weighted summation of the normalized mean wind speed and standard deviation of wind speed. The scenario with the largest comprehensive evaluation index was selected as the typical high-fluctuation scenario, and the typical high-fluctuation scenario was used as the future wind speed time series scenario for wind turbines.

[0016] By generating multiple future wind speed time series scenarios based on historical wind speed sequences from meteorological towers, and using a comprehensive evaluation index that takes into account both wind speed level and fluctuation intensity to screen typical high wind speed and high fluctuation wind speed scenarios, this avoids the problem of smoothing out the random fluctuation characteristics of future wind speed caused by using a single deterministic wind speed prediction value, and provides representative input conditions for broadband oscillation risk assessment.

[0017] Preferably, the time-series scene generation model extracts conditional features based on historical wind speed sequences and maps random variables to generate... K Future wind speed time series scenario; k The future wind speed time series scenario is represented as follows:

[0018] in, For the first k The item contains H Wind speed time-series scenario at a future moment. These are conditional features obtained from historical wind speed sequences. For the first k A random variable, For parameters θ A time-series scene generation model.

[0019] Preferably, the broadband oscillation instability probability is the integral of the probability density of the real part of the dominant oscillation mode over the interval where the real part is greater than zero.

[0020] The second objective of this invention is to provide a probabilistic assessment system for broadband oscillation risk in wind power grid-connected systems, used to implement the above-mentioned method, comprising: The unit power probability modeling module is configured to obtain the future wind speed time series scenario of the wind turbine. Based on the power characteristics of the wind turbine, it divides the wind speed data domain into multiple operating intervals. It uses the soft partition membership function to perform weighted fusion of the Copula sub-models of each operating interval to obtain a continuously differentiable wind speed-power joint probability model. Based on the future wind speed time series scenario, it derives the time-varying power probability density of each wind turbine. The oscillation risk assessment module is configured to construct the global node admittance matrix of the wind power grid-connected system, solve the dominant oscillation mode based on the global node admittance matrix, map the time-varying power probability density to the time-varying probability density of the real part of the dominant oscillation mode based on the sensitivity coefficient of the real part of each wind turbine power, and integrate the time-varying probability density of the real part of the dominant oscillation mode over a preset instability interval to obtain the broadband oscillation instability probability of the wind power grid-connected system at future times, so as to realize the probabilistic assessment of the broadband oscillation risk of the system.

[0021] A third objective of this invention is to provide an electronic device comprising at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program that, when executed by the processing unit, causes the processing unit to perform the aforementioned method.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By using the soft partition membership function, the probability density abrupt change at the boundary of the operating range in traditional hard partition modeling is eliminated, ensuring the continuous differentiability of the ring wind speed-power joint probability model over the entire domain, and providing high-precision probability input for subsequent stability analysis; 2. By establishing an analytical mapping from the time-varying power probability density to the time-varying probability density of the real part of the dominant oscillation mode, an efficient quantitative assessment of the dynamic evolution of broadband oscillation instability risk with operating conditions can be achieved.

[0023] 3. Based on historical wind speed sequences from meteorological towers, multiple future wind speed time series scenarios are generated. Typical high-wind-speed and high-fluctuation-speed scenarios are selected using a comprehensive evaluation index that takes into account both wind speed level and fluctuation intensity. This avoids the problem of smoothing out the random fluctuation characteristics of future wind speeds caused by using a single deterministic wind speed prediction value, and provides representative input conditions for broadband oscillation risk assessment.

[0024] 4. In the actual measurement examples, the instability probability obtained by this invention evolves with the operating conditions, and has high evaluation accuracy after Monte Carlo sampling cross-validation. It can provide a quantitative basis for early warning of broadband oscillation risk of wind power flexible DC transmission system. Attached Figure Description

[0025] Figure 1 This invention provides a flowchart of a probabilistic assessment method for broadband oscillation risk in a wind power grid-connected system; Figure 2 A schematic diagram of a flexible DC transmission system for wind power. Figure 3 This invention provides a data flow diagram illustrating a probabilistic assessment method for broadband oscillation risk in a wind power grid-connected system. Figure 4 A schematic diagram of the joint probability density of wind speed and power fitted by weighted Copula for the region below the cut-in wind speed. Figure 5 A schematic diagram of the joint probability density of wind speed and power fitted by weighted Copula for the maximum power tracking region; Figure 6 A schematic diagram of the joint probability density of wind speed and power fitted by weighted Copula for the power-limited region; Figure 7 A schematic diagram of the joint probability density of wind speed and power fitted by weighted Copula for extreme wind speed regions; Figure 8 The power probability evolution curves for each unit are shown. Figure 9 The time series curves represent the probability of system instability. Figure 10 A comparison chart of the evaluated values ​​and the measured values ​​in Monte Carlo is provided for analysis. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings: Example 1: This embodiment uses a real offshore wind farm as the object. The wind farm is connected to the grid via a flexible DC transmission system and includes six wind turbines, each with a rated capacity of 9 MW. The data is based on synchronously acquired 1-minute-level wind speed sequences from meteorological towers and active power sequences from the six turbines. For example... Figure 1 As shown in the figure, this embodiment provides a probabilistic assessment method for broadband oscillation risk in a wind power grid-connected system, specifically including steps S101 to S107. The specific data flow of the method is as follows: Figure 3 As shown.

[0028] Step S101: Generate a set of future wind speed time series scenarios. Obtain the historical wind speed sequence of the anemometer tower, and perform missing value processing, outlier processing, normalization processing, and sliding window segmentation on the historical wind speed sequence to obtain a conditional feature vector c used to characterize the time series features of historical wind speed.

[0029] Connect the conditional feature vector c with the first k random variables Input the pre-trained wind speed time-series scene generation model to obtain the first... k Future wind speed time series scenario :

[0030] In the formula, G θ () represents the parameter θ Wind speed time-series scene generation model; For the first k A future wind speed time series scenario; H The number of moments included in the future prediction time domain; K The number of scenes generated.

[0031] The generated results are denormalized to obtain a set of future wind speed time series scenarios.

[0032] Step S102: Select typical high-wind-speed and high-fluctuation-wind-speed scenarios. This is based on the scenarios generated in step S101. K For each future wind speed time series scenario, calculate the first...k Average wind speed in the scene μ k and wind speed standard deviation σ k :

[0033]

[0034] For the mean wind speed μ k and wind speed standard deviation σ k Perform extreme value normalization:

[0035]

[0036] The comprehensive evaluation index is obtained by weighted summation of the normalized mean wind speed and standard deviation wind speed. :

[0037] in, Select the one that satisfies The future wind speed time series scenario is used as the typical high wind speed and high fluctuation wind speed scenario. The first k The future wind speed time series scenario is taken as a typical high wind speed and high wind speed fluctuation scenario, and is input into the subsequent step S103.

[0038] Step S103: Soft-partition weighted Copula joint probability modeling. The structure of the wind power flexible DC transmission system is as follows: Figure 2 As shown, the system consists of multiple wind turbine generators and a sending-end modular multilevel converter station. Specifically, the wind power flexible DC transmission system includes multiple wind turbine groups in the offshore wind farm, 690V / 35kV and 35kV / 220kV transformers at the turbine end, transmission line impedances Z1 to ZN, and the sending-end MMC-HVDC system; where nodes #1 to #N are the grid connection points for each turbine group, and node #N+1 is the port of the sending-end converter station.

[0039] Considering that the correlation between wind speed and turbine output is not consistent across different operating ranges, and that the transition between adjacent operating ranges is gradual rather than a strict step switch, this step establishes a soft-partition membership-weighted Copula hybrid model: Nonparametric kernel density estimation was used to fit the marginal probability densities of wind speed at the meteorological tower and active power of the generator unit, and the cumulative distribution function was obtained by integration. Wind speed at the wind measurement tower v Marginal probability density The calculation formula is:

[0040] In the formula, K [·] represents the Gaussian kernel function. h For window width parameters, n The total number of wind speed samples; the first j Active power of each wind turbine Marginal probability density Calculate in the same way; By integrating the marginal probability densities of the wind speed at the meteorological tower and the active power of the generating unit, respectively, the cumulative distribution function of the wind speed at the meteorological tower is obtained. Cumulative distribution function of active power of the unit .

[0041] Based on the maximum power point tracking (MPPT) characteristics, power limiting characteristics, and overspeed protection characteristics of the wind turbine power curve, the wind speed domain of the meteorological tower is divided into MPPT intervals, power limiting intervals, and extreme wind speed intervals. Samples below the cut-off wind speed are considered as shutdowns. A soft-partition membership function that continuously varies with wind speed is constructed for each operating interval using a Sigmoid function at the boundaries of adjacent intervals, ensuring that the sum of the membership degrees of all intervals at the same wind speed is one. Then, within each operating interval, Gaussian, t, Gumbel, Clayton, and Frank are used as candidate Copula functions. The optimal Copula sub-model for that interval is selected by comprehensively considering the weighted scores of Spearman rank correlation coefficient error, Kendall rank correlation coefficient error, and squared Euclidean distance.

[0042] Taking six wind turbine units as an example, Copula fitting was performed on the wind speed at the meteorological tower and the active power of the units in the maximum power point tracking (MPPT) zone, the power limiting zone, and the extreme wind speed zone, respectively. The comprehensive score of each candidate Copula function was calculated according to the following formula:

[0043] In the formula e ρ , e τ , e d These represent the Spearman rank correlation coefficient error, the Kendall rank correlation coefficient error, and the squared Euclidean distance, respectively, with weights of α1=α2=1 / 4 and α3=1 / 2, within each operating interval. S (C) The smallest candidate Copula is the optimal Copula submodel for that interval.

[0044] The candidate function with the smallest comprehensive score was selected as the optimal Copula sub-model for the corresponding unit within its respective operating range, and the results are shown in Table 1. Table 1 shows that all units selected Frank Copula within the maximum power tracking region; however, the optimal Copula sub-model differed between units within the power-limited region and the extreme wind speed region. For example, WT6 selected Gumbel Copula in the power-limited region, WT2 selected Gumbel Copula in the extreme wind speed region, and WT3 and WT4 selected t-Copula in the extreme wind speed region. This indicates that the wind speed-power correlation structure is not entirely the same across different operating ranges and at different unit locations.

[0045] Table 1. Optimal Copula sub-models for each wind turbine in different operating ranges

[0046] Finally, the optimal Copula sub-models for each interval are weighted and summed according to the soft partition membership function:

[0047] In the formula, l =1,2, which are the boundary numbers between the three running intervals; b 1. b 2 represents the boundary wind speeds between the maximum power tracking zone and the power limiting zone, and between the power limiting zone and the extreme wind speed zone, respectively. δ l >0 represents the boundary transition width, with units identical to wind speed. Based on this, normalized soft-zone membership degrees are constructed. And satisfying that the sum of the membership degrees of each interval is 1, specifically expressed as:

[0048] Reference Figures 4-7 This yields a continuously differentiable joint probability density of wind speed and power over the entire wind speed domain, thus avoiding abrupt changes in probability density at the boundaries of hard partitions; joint probability density of wind speed and power Represented as:

[0049] In the formula, c l [·] is the first l The density function of the optimal Copula submodel for each operating interval.

[0050] Step S104: Derive the power probability density of each unit. Based on the binary Copula conditional distribution, under the given wind speed of the meteorological tower, calculate the conditional probability density of the active power of each unit using the selected optimal Copula sub-model and the marginal probability density of wind speed and unit power, establishing a mapping relationship from future wind speed scenarios to the probability distribution of unit power. Substitute the wind speed sequence of typical high-fluctuation wind speed scenarios into the time sequence to obtain the synchronous change process of the power probability density of each unit over time. Based on retaining the power probability density of each unit, aggregate the active power of each unit in the wind farm to obtain the probabilistic prediction result of the total output of the wind farm, thereby characterizing the spatiotemporal correlation of the output of different units in the wind farm.

[0051] Given the wind speed at the meteorological tower v Conditions, No. j Active power of typhoon generator conditional probability density Represented as:

[0052] In the formula, the conditional probability density is determined by the weighted fusion of the wind speed-power joint probability density and the wind speed marginal probability density.

[0053] The entire wind farm is constantly t Total output (active power) Represented as:

[0054] In the formula, N This refers to the number of wind turbines in the wind farm. For the first i Typhoon turbine units at all times t The active power.

[0055] Figure 8 This is a graph showing the power probability evolution of each generating unit under a typical high-fluctuation scenario. The horizontal axis represents the prediction delay, and the vertical axis represents the active power of the generating unit. Figure 8 This shows the changes in the average power of each unit and its probability interval over time. (Refer to...) Figure 8 Under the same wind speed conditions at the meteorological tower, the overall trend of power output variation among the units was consistent, but there were significant differences in power output level, fluctuation amplitude, and probability interval width. Figure 8 This reflects the spatial differences in the wind speed-power mapping relationship among various units.

[0056] Step S105: Construct the global node admittance matrix. (Refer to...) Figure 2In this system, each wind turbine grid-connected converter access point corresponds to a node, and the sending-end modular multilevel converter station port corresponds to a node. Considering the frequency coupling characteristics of both the wind turbine grid-connected converter and the sending-end modular multilevel converter station, each node port is dynamically represented as a two-dimensional admittance matrix. The parallel admittance of the equipment at the wind turbine node includes the frequency coupling admittance of the grid-connected converter and the grounding admittance of the turbine-side filter. The parallel admittance of the equipment at the converter station node is determined by the frequency coupling impedance model of the converter station. The nodes are connected by line admittance. Based on this, a global node admittance matrix of the system is constructed, and the number of random groups in its dimension is determined.

[0057] Step S106: Map the real part probability density of the dominant mode. The broadband stability of the system is determined by the characteristic equation in which the determinant of the global nodal admittance matrix is ​​zero. Solving this characteristic equation yields the dominant oscillation mode of the system; the characteristic equation is specifically expressed as:

[0058] In the formula, This is the system's global node admittance matrix. For complex frequency domain variables (complex frequencies). This is the collection of active power from all units in the system.

[0059] The system is in a stable state when the real part of the dominant oscillation mode is less than zero. When the real part moves towards the right half of the complex plane, the system tends towards critical stability or instability. Since the parallel admittance of the wind turbine node equipment is directly related to the operating point of the unit's active power, power fluctuations will affect the distribution of system eigenvalues ​​by changing the node admittance matrix. Based on this, according to the implicit function differentiation rule, the partial derivatives of the characteristic equation with respect to the unit power and complex frequency are obtained, and the sensitivity coefficients of the real part of the dominant oscillation mode to the active power of each unit are obtained by expanding the derivative of the determinant with respect to the matrix parameters. Near the reference operating point, the change in the real part of the dominant mode is approximately linearly expressed as the weighted sum of the power offset of each unit and the corresponding sensitivity coefficient. The power probability density sequence of each unit output in step S104 is sampled, and after the above linear mapping to the eigenvalue domain, a discrete sample set of the real part of the dominant mode is obtained. Then, kernel density estimation is performed on this set to obtain the time-varying probability density function of the real part of the dominant oscillation mode.

[0060] The derivation of the sensitivity coefficient is as follows:

[0061] In the formula, For the first k One dominant oscillation mode; For the first i The active power of the generator unit; f For implicit functions, defined as ; For the firsti The reference operating point power of the generator set.

[0062] Determinant to matrix parameters (here) The derivative expansion of ) is as follows:

[0063] In the formula, Let be the determinant of the global node admittance matrix, tr(·) be the trace of the matrix, and adj(·) be the adjoint matrix. For the global node admittance matrix, the first... i Partial derivatives of the active power of the generator unit.

[0064] The linear mapping of the change in the real part of the dominant mode is as follows:

[0065] In the formula, For the first k The change in the real part of each dominant oscillation mode; Re(·) is the real part of the complex number; The sensitivity coefficient of the dominant oscillation mode to the active power of each unit; Let the real part of the sensitivity coefficient be denoted as ; For the first i The offset of the active power of the generator unit relative to the reference operating point.

[0066] Step S107: Integrate to obtain the broadband oscillation instability probability. Define the broadband oscillation instability probability as the integral of the real part probability density of the dominant oscillation mode in the right half of the complex plane, that is, integrate the real part probability density function of the dominant mode obtained by mapping the power probability density over the interval where the real part is greater than zero to obtain the broadband oscillation instability probability of the system at each future time.

[0067] Wideband oscillation instability probability of the system at various future time points The calculation formula is:

[0068] In the formula, Let be the time-varying probability density function of the real part of the dominant oscillation mode, obtained by mapping the power probability density and estimating the kernel density.

[0069] Figure 9 The graph shows the time series curve of the system instability probability. The horizontal axis represents the prediction delay, and the vertical axis represents the broadband oscillation instability probability. Figure 9 This illustrates the evolution of the instability probability with changing operating conditions and its rapid increase during high-risk periods. (Refer to...) Figure 9In this embodiment, when the power distribution is below the rated power in the previous stage, the instability probability remains at a low level. As the power mean and the upper tail of the fluctuation gradually approach the critical power, the instability probability rises rapidly and reaches a peak of 18.8% at about 187 minutes. This quantitatively describes the dynamic process of instability risk evolving with operating conditions.

[0070] To verify the evaluation results, Monte Carlo sampling was used for cross-validation: a large number of power samples were independently drawn for each future time point according to its wind power probability distribution, the stability of the operating point was determined one by one, and the frequency of the unstable sample was taken as the statistical value of the instability probability at that time.

[0071] Figure 10 This chart compares the analytically estimated values ​​with the Monte Carlo measured values. The top chart shows the comparison between the analytically estimated values ​​and the Monte Carlo measured values ​​of the instability probability, while the bottom chart shows the absolute error and its average value at each time point. (Refer to...) Figure 10 The analytical evaluation value and the Monte Carlo statistical value show a high degree of consistency in their trends across the entire prediction time domain, with an average absolute error of approximately 0.38% across the entire time domain. At the moment of highest instability risk, the error between the analytical value and the statistical value is approximately 0.9%, verifying the effectiveness and accuracy of the method provided in this embodiment.

[0072] Example 2: This embodiment provides a probabilistic assessment system for broadband oscillation risk in wind power grid-connected systems, used to implement the method provided in Embodiment 1, including: The unit power probability modeling module is configured to obtain the future wind speed time series scenario of the wind turbine, divide the wind speed data domain into multiple operating intervals based on the power characteristics of the wind turbine, use the soft partition membership function to perform weighted fusion of the Copula sub-models of each operating interval to obtain a continuously differentiable wind speed-power joint probability model, and derive the time-varying power probability density of each wind turbine based on the future wind speed time series scenario; that is, execute steps S101 to S104. The oscillation risk assessment module is configured to construct the global node admittance matrix of the wind power grid-connected system, solve the dominant oscillation mode based on the global node admittance matrix, map the time-varying power probability density to the time-varying probability density of the real part of the dominant oscillation mode based on the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine, and integrate the time-varying probability density of the real part of the dominant oscillation mode over a preset instability interval to obtain the broadband oscillation instability probability of the wind power grid-connected system at future times, so as to realize the probabilistic assessment of the broadband oscillation risk of the system; that is, to execute steps S105 to S107.

[0073] Example 3: This embodiment provides an electronic device, including at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the program is executed by the processing unit, the processing unit performs the method provided in Embodiment 1.

[0074] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A probabilistic assessment method for broadband oscillation risk in a wind power grid-connected system, characterized in that, include: The future wind speed time series scenario of wind turbines is obtained. Based on the power characteristics of wind turbines, multiple operating intervals are divided in the wind speed data domain. The Copula sub-models of each operating interval are weighted and fused using the soft partition membership function to obtain a continuously differentiable wind speed-power joint probability model. The time-varying power probability density of each wind turbine is derived based on the future wind speed time series scenario. Construct the global node admittance matrix of the wind power grid-connected system, and solve the dominant oscillation mode based on the global node admittance matrix; Based on the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine, the time-varying power probability density is mapped to the time-varying probability density of the real part of the dominant oscillation mode. By integrating the time-varying probability density of the real part of the dominant oscillation mode over a preset instability interval, the broadband oscillation instability probability of the wind power grid-connected system at future times is obtained, so as to realize the probabilistic assessment of the broadband oscillation risk of the system.

2. The method according to claim 1, characterized in that, The weighted fusion of Copula sub-models for each operating interval using the soft partition membership function includes: Based on the power curve of the wind turbine, multiple operating intervals are divided in the wind speed data domain; In each operating range, a Copula sub-model is fitted between wind speed and unit active power; A soft partition membership function is constructed from the transition function at the boundary of adjacent operating intervals, so that the sum of the membership degrees of each operating interval under the same wind speed data is 1; The optimal Copula sub-models in each operating range are weighted and fused according to the soft partition membership function to obtain a wind speed-power joint probability model that is continuously differentiable in the wind speed data domain.

3. The method according to claim 2, characterized in that, The step of weighting and fusing the optimal Copula sub-models in each operating interval according to the soft partition membership function to obtain the wind speed-power joint probability model that is continuously differentiable in the wind speed data domain includes: Nonparametric kernel density estimation was used to fit the marginal probability densities of wind speed data and wind turbine power respectively, and the corresponding cumulative distribution function was obtained by integration; Based on the density function, cumulative distribution function, marginal probability density, and soft partition membership function of the optimal Copula sub-model, the joint probability model of wind speed and power is calculated.

4. The method according to claim 1, characterized in that, The sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine unit maps the time-varying power probability density to the time-varying probability density of the real part of the dominant oscillation mode, including: The dominant oscillation mode is solved by the characteristic equation of the global node admittance matrix, and the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine is obtained according to the implicit function differentiation rule. The change in the real part of the dominant oscillation mode near the reference operating point is expressed as a weighted sum of the power offset of each wind turbine and the corresponding sensitivity coefficient; Based on the weighted sum of the power offset of each wind turbine and the corresponding sensitivity coefficient, the time-varying power probability density is mapped to the real part of the dominant oscillation mode, and the time-varying probability density of the real part of the dominant oscillation mode is obtained by kernel density estimation.

5. The method according to claim 4, characterized in that, The sensitivity coefficients of the real part of the dominant oscillation mode to the power of each wind turbine, obtained according to the implicit function differentiation rule, include: By expanding the derivative of the determinant with respect to the matrix parameters, the trace of the product of the adjoint matrix of the global node admittance matrix and the partial derivative matrix of the global node admittance matrix with respect to the wind turbine power is calculated, so as to obtain the partial derivative of the characteristic equation with respect to the wind turbine power. Based on the partial derivatives of the characteristic equation with respect to the power of the wind turbine and with respect to the complex frequency variable, the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine is calculated.

6. The method according to claim 1, characterized in that, The acquisition of the future wind speed time series scenario for wind turbines includes: Input historical wind speed sequences into the time-series scene generation model to obtain a set of future wind speed time-series scenes, each scene containing wind speed sequences within a preset time period; Calculate the mean wind speed and standard deviation of wind speed for each scenario in the future wind speed time series scenario set; We obtained a comprehensive evaluation index by weighted summation of the normalized mean wind speed and standard deviation of wind speed. The scenario with the largest comprehensive evaluation index was selected as the typical high-fluctuation scenario, and the typical high-fluctuation scenario was used as the future wind speed time series scenario for wind turbines.

7. The method according to claim 6, characterized in that, The time-series scene generation model extracts conditional features based on historical wind speed sequences and maps random variables to generate... K Future wind speed time series scenario; k The future wind speed time series scenario is represented as follows: in, For the first k The item contains H Wind speed time-series scenario at a future moment. These are conditional features obtained from historical wind speed sequences. For the first k A random variable, For parameters θ A time-series scene generation model.

8. The method according to claim 1, characterized in that, The broadband oscillation instability probability is the integral of the probability density of the real part of the dominant oscillation mode over the interval where the real part is greater than zero.

9. A probabilistic assessment system for broadband oscillation risk in a wind power grid-connected system, used to implement the method described in any one of claims 1-8, characterized in that, include: The unit power probability modeling module is configured to obtain the future wind speed time series scenario of the wind turbine. Based on the power characteristics of the wind turbine, it divides the wind speed data domain into multiple operating intervals. It uses the soft partition membership function to perform weighted fusion of the Copula sub-models of each operating interval to obtain a continuously differentiable wind speed-power joint probability model. Based on the future wind speed time series scenario, it derives the time-varying power probability density of each wind turbine. The oscillation risk assessment module is configured to construct the global node admittance matrix of the wind power grid-connected system and solve the dominant oscillation mode based on the global node admittance matrix; Based on the sensitivity coefficient of the real part of the dominant oscillation mode to the power of each wind turbine, the time-varying power probability density is mapped to the time-varying probability density of the real part of the dominant oscillation mode. By integrating the time-varying probability density of the real part of the dominant oscillation mode over a preset instability interval, the broadband oscillation instability probability of the wind power grid-connected system at future times is obtained, so as to realize the probabilistic assessment of the broadband oscillation risk of the system.

10. An electronic device, characterized in that, It includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program that, when executed by the processing unit, causes the processing unit to perform the method as described in any one of claims 1-8.