Data feature and physical mechanism fused wind power generation recurrence period wind speed gridding evaluation method and system and medium

Abstract

The invention discloses a data feature and physical mechanism fused wind power generation recurrence period wind speed gridding evaluation method and system and a medium, and the method comprises the steps: firstly carrying out the spectrum analysis of a gridding historical wind speed sequence, then carrying out the high-frequency compensation, constructing an annual maximum wind speed sequence, and calculating the recurrence period wind speed at the position of a gridding point through the combination of extreme value I-type distribution; performing homogeneity test and correction on the annual maximum wind speed sequence observed by the reference station by using a method of combining RHtests and metadata, and calculating the return period wind speed at the position of the reference station by combining extreme value I-type distribution; completing (reference) station-(wind measurement) tower correction and multi-height layer extrapolation based on a wind shear index according to a strong wind sample; and on the kilometer-level grid, grid points and station tower information are fused through an assimilation framework, and a multi-height-layer return period wind speed grid point data set is generated. According to the method, the precision and the space coverage capability of the return period wind speed evaluation under the complex terrain condition are improved, and a basis is provided for wind power plant site selection, unit type selection and wind resistance design.

Description

Technical Field

[0001] This invention belongs to the field of climate and wind energy engineering and power safety assessment technology, involving wind energy resource assessment, meteorological data processing and extreme wind speed return period calculation, etc. Specifically, it is a gridded assessment method, system and medium for wind power generation return period wind speed that integrates data characteristics and physical mechanisms. It is used to accurately estimate the extreme wind speed of wind power return period in areas without long-term wind measurement data, and provides key data support for wind turbine wind resistance safety design, customized selection and cost reduction and efficiency improvement. Background Technology

[0002] The safety selection of wind turbines is highly sensitive to the quantitative characterization of extreme wind risks. The 50-year return period wind speed (V50) is a key design parameter in the wind power industry, directly impacting the safety and economy of the turbine. On one hand, the maximum 50-year return period wind speed determines the wind turbine's ability to withstand extreme wind loads. If this parameter is set too low, it may lead to serious accidents such as tower breakage, blade damage, or even complete turbine collapse under wind speeds exceeding the design limit. If the value is set too high, it may result in over-design of the tower, foundation, and other structures, increasing unnecessary costs. On the other hand, the maximum 50-year return period wind speed affects the customized design optimization of the turbine. The current wind power industry is trending towards refined development, and this parameter directly affects key design aspects such as turbine selection, tower strength, and foundation load calculation. Therefore, accurate assessment of the return period wind speed can avoid insufficient safety redundancy or material waste, thereby reducing the levelized cost of energy (LCOE).

[0003] However, in practical engineering applications, the proposed wind farm areas are often located in remote regions with sparse meteorological observation stations, lacking long-term and stable wind measurement data. This limits the application of traditional extreme value analysis methods and makes it difficult to meet industry needs. Secondly, complex terrain leads to significant spatial heterogeneity in wind field characteristics, making it difficult for traditional assessment methods based on single-point observations to accurately reflect the wind resource status of the entire wind farm area. Furthermore, the increased frequency and intensity of extreme weather events under the background of climate change also brings new challenges to the accurate assessment of return-time wind speeds.

[0004] To compensate for the lack of observational data, the industry has begun to introduce numerical simulation and reanalysis data. However, due to limitations in computational resolution and parameterization schemes, these model data generally underestimate the energy cutoff of high-frequency turbulence, resulting in insufficient simulation capabilities for extreme wind speed events. Extreme wind speeds calculated directly from their wind speed sequences exhibit systematic biases. Another auxiliary method is to utilize long-term meteorological station data from surrounding areas. However, meteorological stations themselves may exhibit data heterogeneity due to relocation, instrument replacement, or environmental changes, requiring complex homogeneity corrections. More importantly, the geographical location and underlying surface environment of meteorological stations often differ from those of the proposed wind farm, limiting the representativeness of their wind speed characteristics; direct citation can introduce significant errors. Although corrections can be made through correlation analysis between short-term anemometer tower data and meteorological station data, scientifically selecting effective high-wind samples and establishing robust correction models remain challenges.

[0005] In conclusion, scientifically assessing return-time wind speed is not only crucial for the safety and economic viability of individual projects, but also of great significance for the large-scale development of regional wind power and the stable operation of the power grid. However, existing return-time wind speed assessment methods have many shortcomings in terms of data acquisition, accuracy, and applicability. Therefore, how to construct a return-time wind speed assessment method that can comprehensively utilize physical mechanisms and observational data characteristics, overcome the limitations of single data sources, and achieve high-precision gridded output has become an urgent technical problem to be solved. Summary of the Invention

[0006] (a) Purpose of the invention In response to the lack of long-term wind speed observation data in proposed wind farm areas and the aforementioned deficiencies and shortcomings of existing technologies, this invention aims to provide a gridded wind speed assessment method, system, and medium that integrates data characteristics and physical mechanisms for wind power generation with a return period. The invention achieves more accurate and reliable wind speed assessment through the following technical means: (1) developing a gridded return period wind speed dataset using spectrum compensation and numerical simulation techniques; (2) collecting short-term observation data from meteorological towers and long-term wind measurement data from surrounding reference meteorological stations, and conducting data quality control, especially detecting sudden changes in wind speed caused by station relocation, changes in observation instruments, etc. (3) The return period wind speed of the meteorological station is calculated using the extreme value analysis method. Combined with the correlation analysis of the wind measurement tower and the meteorological station wind sample, the return period wind speed of the meteorological station is corrected to the location of the wind measurement tower. (4) By constructing a multi-source data assimilation framework, the observation data of local meteorological stations and wind measurement towers are integrated with the gridded return period dataset to achieve the optimal fusion of multi-source data. This breaks through the adaptation limitations of the traditional extreme value analysis method under complex wind conditions, makes up for the shortcoming that it is difficult to provide the return period wind speed with full coverage due to the sparse observation stations, and alleviates the industry problem of restricting wind power selection and site selection in this region.

[0007] (II) Technical Solution To achieve the objective of this invention and solve its technical problems, the present invention adopts the following technical solution: The first objective of this invention is to provide a gridded assessment method for wind speed during the return period of wind power generation, which integrates data features and physical mechanisms. This method addresses the problems of insufficient accuracy in assessing extreme wind speeds during the return period and the difficulty in conducting wind power development resource assessments due to the sparseness of observation stations in wind farm site selection areas. It enables high-precision spatial assessment of wind speed during the return period required for safe selection of wind turbine units under conditions of sparse observation and complex terrain. The method includes at least the following steps in its implementation: S100. Gridded Spectrum Compensation and AMAX Construction: Gridded historical wind speed sequences covering the sparse target assessment area and its outward buffer zone are obtained through numerical weather prediction models or reanalysis datasets. The frequency domain power spectral density distribution is obtained by spectral analysis of the wind speed sequences using the Fast Fourier Transform (FFT) method. The energy spectrum of the high-frequency band is spectrally compensated according to the "-5 / 3" slope law of turbulent energy spectrum to form an equivalent energy spectrum and then reconstructed in the time domain. The annual maximum wind speed (AMAX) sequence is constructed grid by grid. S200. Construction of a gridded return period wind speed dataset: The extreme value type I distribution is used to perform statistical analysis on the AMAX sequence obtained by spectrum compensation. Based on the empirical distribution linearization or maximum likelihood, the parameters are estimated to calculate the return wind speed value corresponding to the R-year target return period at each grid point. A gridded R-year target return period wind speed basic dataset covering the target assessment area and its outer buffer zone is constructed, and parameter stability and fitting diagnostic indicators are output. S300. Calculation of wind speed during the return period of the reference meteorological station: Long-term historical wind speed observation data of surrounding meteorological stations inside and outside the target assessment area were collected. The homogeneity of the annual maximum wind speed sequence observed by the reference meteorological stations was tested by combining RHtests and metadata. The abrupt change points and non-homogeneity of wind speed data were identified and corrected. The AMAX sequence of each reference meteorological station was constructed. Based on this, the return wind speed value corresponding to the R-year return period of each reference meteorological station was estimated by using extreme value type I distribution, forming a set of return period wind speeds of reference stations. S400. Calculation of return period wind speed at the location of the meteorological tower: Collect hourly wind speed observation data from surrounding meteorological stations and wind measurement towers within the target wind farm area. Based on the causes of annual maximum wind speed, select high wind speed samples with similar weather backgrounds from the wind speed sequences of the meteorological stations and wind measurement towers. Establish a correction relationship from the station to the tower location. Transfer the R-year return wind speed of the reference station to the wind measurement tower location to obtain the R-year return period wind speed value at the wind measurement tower location. S500. Wind speed calculation for the return period of wind power generation: Based on the high-wind-speed samples obtained through screening, the wind shear index under extreme wind conditions is estimated. According to the power law relationship, the R-year return period wind speeds at grid locations, surrounding reference meteorological stations, and anemometer tower locations are uniformly extrapolated to several wind power design height layers (including but not limited to 50m, 100m, 150m, 200m, or target hub height). In accordance with engineering standards, a consistency conversion is performed between the 10-minute average wind and the 3-second gust to obtain the multi-height layer return period wind speeds for wind power generation design.

[0008] The second objective of this invention is to provide a gridded evaluation system for wind speed during the return period of wind power generation, comprising multiple modules for performing the above-mentioned method for gridded evaluation of wind speed during the return period of wind power generation, which integrates data features and physical mechanisms.

[0009] The third objective of this invention is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned wind speed gridding assessment method for wind power generation return period that integrates data characteristics and physical mechanisms.

[0010] (III) Technical Effects Compared with existing technologies, the wind speed gridded assessment method, system, and medium for wind power generation return period based on the integration of data features and physical mechanisms of the present invention have the following beneficial and significant technical effects: (1) This invention effectively solves the technical problem of missing high-frequency information in wind speed data in sparsely observed areas by combining spectrum compensation with physical mechanism constraints. Based on the "-5 / 3" energy spectrum law of turbulence theory, the power spectral density is quantitatively compensated, restoring the complete turbulent wind speed spectrum characteristics and significantly improving the physical rationality of AMAX sequence construction. Compared with the traditional method of relying on limited observation data for extreme value statistics, this invention provides independent extreme value estimation verification through spectral moment integral calculation, reducing the risk of underestimation of return period wind speed due to data loss, and providing a reliable data foundation for wind power development in observation-limited areas.

[0011] (2) The multi-source data optimal assimilation framework constructed in this invention achieves high-precision information fusion from point observations to area grid points. By establishing the background field error covariance matrix, observation error covariance matrix, and representativeness penalty matrix, the uncertainty characteristics of different data sources are quantified to achieve optimal weight allocation. Machine learning such as random forest is introduced to statistically correct systematic deviations under complex terrain. Combined with vertical consistency constraints and wind direction zoning assimilation strategies, the physical coordination of wind speeds at multiple altitude levels is ensured.

[0012] (3) The quality control and verification system established by this invention ensures the engineering reliability of gridded products. Through multi-level cross-validation, independent site inspection and uncertainty quantification assessment, an evaluation framework including multi-dimensional accuracy indicators such as mean deviation, root mean square error and probability distribution fit is constructed, which provides quantitative uncertainty information for wind turbine selection and wind-resistant design, and can effectively meet engineering applications. Attached Figure Description

[0013] Figure 1 This is a flowchart of the wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms, provided in this embodiment of the invention. Figure 2 This is a flowchart illustrating the implementation process of spectrum compensation and AMAX construction in this invention; Figure 3 This involves reanalyzing the spectral distribution of the wind speed sequence; Figure 4 This is a flowchart illustrating the implementation of extreme value type I distribution parameter estimation in this invention; Figure 5 This is a flowchart illustrating the implementation of the method in this invention that combines RHtests and metadata to perform a uniformity test on the annual maximum wind speed sequence observed by the reference meteorological station. Figure 6 This is a comparison chart of wind speed before and after correction in the long sequence of the reference meteorological stations in this invention; Figure 7 This is a flowchart illustrating the implementation of multi-source optimal assimilation and gridded fusion in this invention. Detailed Implementation

[0014] This invention aims to provide a gridded assessment method, system, and medium for wind power generation with return period wind speed that integrates data features and physical mechanisms. It addresses the problems of insufficient accuracy in assessing extreme wind speeds with return periods due to the sparse number of observation stations in wind farm site selection areas, and the difficulty in conducting wind power development resource assessments. It achieves high-precision spatial assessment of return period wind speeds required for safe selection of wind turbine units under conditions of sparse observation and complex terrain. To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. The described embodiments are some, but not all, embodiments of this invention, and are exemplary, intended to explain the invention, and should not be construed as limiting the invention.

[0015] Example 1: Gridded Evaluation Method for Wind Speed ​​During Wind Power Generation Recurrence Period like Figure 1 As shown in the embodiments of the present invention, the wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms, mainly includes the following steps in its implementation: S100. Gridded Spectrum Compensation and AMAX Construction: Gridded historical wind speed sequences covering sparsely observed target assessment areas and their outward buffer zones are obtained using numerical weather prediction models or reanalysis datasets. Frequency domain power spectral density distribution is obtained by performing spectral analysis on the wind speed sequences using FFT. The energy spectrum of high-frequency bands is spectrally compensated according to the "-5 / 3" slope law of turbulent energy spectrum to form an equivalent energy spectrum, which is then reconstructed in the time domain. The annual maximum wind speed (AMAX) sequence is constructed grid by grid.

[0016] As a preferred approach, gridded historical wind speed sequences are used to cover the target assessment area and a geographical buffer zone is established outwards. The width of the buffer zone is 100-200 km or 1-2 degrees of latitude and longitude, depending on the scale of the target area, to reduce the impact of boundary information truncation. The original data is... u , v Component wind speeds are collected hourly or every three hours and unified to the same projection and time reference. Only after synthesizing scalar wind speeds can they enter the spectrum analysis process.

[0017] Preferred data sources include the ERA5 reanalysis dataset, NCEP / NCAR reanalysis data, or high-resolution numerical weather prediction model outputs, ensuring the spatiotemporal continuity and physical consistency of the data. The spectral compensation process strictly follows the "-5 / 3" turbulent energy spectrum theory. By comparing the deviation between the observed spectrum and the theoretical spectrum, high-frequency energy loss is quantitatively compensated, restoring the complete turbulent wind speed spectral characteristics and providing a reliable data foundation with physical mechanism constraints for subsequent extreme value statistical analysis.

[0018] S200. Construction of a gridded return period wind speed dataset: The AMAX sequence obtained by spectrum compensation is statistically analyzed using the extreme value type I distribution. Parameter estimation is performed based on empirical distribution linearization or maximum likelihood. The return wind speed value corresponding to the R-year target return period at each grid point is calculated. A gridded basic dataset of R-year target return period wind speed covering the target assessment area and its outer buffer zone is constructed, and parameter stability and fitting diagnostic indicators are output.

[0019] It should be noted that step S200 is based on extreme value statistics theory. A double logarithmic transformation converts the nonlinear extreme value distribution into a linear one, facilitating parameter estimation and confidence interval calculation. The distribution parameter estimation process includes cross-validation using the method of moments and maximum likelihood estimation to ensure the robustness and reliability of the parameter estimates. The return period wind speed calculation uses an inverse function form, supporting wind speed extrapolation for various standard return periods such as 10 years, 25 years, 50 years, and 100 years. Simultaneously, a goodness-of-fit testing system is established, including statistical testing methods such as the Kolmogorov-Smirnov test and the Anderson-Darling test, to verify the applicability of the Type I extreme value distribution and provide a reliability assessment for engineering applications.

[0020] S300. Calculation of wind speed during the return period of the reference meteorological station: Long-term historical wind speed observation data from surrounding meteorological stations within and outside the target assessment area were collected. The uniformity of the annual maximum wind speed series observed by the reference meteorological stations was tested using a combination of RHtests and metadata. Abrupt changes and non-uniformity of wind speed data were identified and corrected. AMAX series of each reference meteorological station was constructed. Based on this, the return wind speed value corresponding to the R-year return period of each reference meteorological station was estimated using the extreme value type I distribution, forming a set of return period wind speeds for the reference stations.

[0021] As a preferred approach, wind speed data homogeneity testing includes: using the penalized maximum F-test to identify abrupt changes in the wind speed time series; establishing a metadata-based manual testing system, which integrates station metadata and cross-validates observation records to verify key events that may lead to data non-homogeneity, such as station relocation, instrument replacement, and changes in the observation environment; separating trends and abrupt changes using a piecewise regression model, fitting piecewise linear trends, and determining the baseline segment based on the longest segment; and using quantile matching to quantitatively correct the identified non-homogeneities.

[0022] Uniformity testing is a crucial step in ensuring the quality of data from reference stations. The penalized maximum F-test method can effectively identify and correct artificial discontinuities in the observation sequence, ensuring the reliability of extreme value statistical analysis. The selection criteria for reference stations include multiple standards such as the length of the observation sequence (no less than 30 years), spatial representativeness, data integrity, and quality reliability.

[0023] S400. Calculation of return period wind speed at the location of the meteorological tower: Collect hourly wind speed observation data from surrounding reference meteorological stations and anemometer towers within the target wind farm area. Based on the causes of annual maximum wind speed, select high wind speed samples with similar weather backgrounds from the wind speed sequences of reference meteorological stations and anemometer towers. Establish a correction relationship from station to tower location. Transfer the R-year return wind speed of the reference station to the anemometer tower location to obtain the R-year return period wind speed value at the anemometer tower location.

[0024] As a preferred approach, the process of selecting and establishing correction relationships for high-wind-speed samples includes: analyzing the meteorological causes of the annual maximum wind speed at the reference meteorological stations, identifying the dominant weather system types (such as typhoons, cold air, thunderstorms, etc.), wind direction distribution characteristics, and pressure field configuration; based on the principle of weather background consistency, selecting high-wind-speed event samples under similar weather conditions from concurrent observations at the anemometer tower and the reference stations; establishing correction relationships using the ratio method, difference method, or multiple linear regression; evaluating the accuracy of the correction methods through cross-validation, selecting the optimal correction model, and achieving reliable migration of the wind speed at the reference station's return period to the anemometer tower's location. By comprehensively considering the influence of local topography, roughness, and microclimate conditions on extreme wind speeds, establishing a station-tower correction relationship based on physical mechanisms can effectively solve the problem of spatial representativeness differences between the reference stations and wind farm sites.

[0025] S500. Wind speed calculation for the return period of wind power generation: Based on the high-wind-speed samples obtained through screening, the wind shear index under extreme wind conditions is estimated. According to the power law relationship, the R-year return period wind speeds at grid locations, surrounding reference meteorological stations, and anemometer tower locations are uniformly extrapolated to several wind power design height layers (including but not limited to 50m, 100m, 150m, 200m, or target hub height). In accordance with engineering standards, a consistency conversion is performed between the 10-minute average wind and the 3-second gust to obtain the multi-height layer return period wind speeds for wind power generation design.

[0026] As a preferred method, the wind shear index calculation includes: based on the selected high-wind-speed samples, using a power-law formula. Calculate the wind shear index at different observation heights α ,in V z Let z be the wind speed at the target height. V z0 The wind speed at the reference height z0 is used as the reference. The variation law of wind shear index under different wind direction sectors, different seasons, and different atmospheric stability conditions is determined through statistical analysis. A spatial interpolation model of wind shear index is established, taking into account the influence of terrain roughness and surface cover type on wind shear. A logarithmic wind profile model is used for verification and correction to ensure the applicability of wind shear index within the hub height range of wind turbine units.

[0027] S600. Multi-source optimal assimilation and gridded fusion (preferred): A multi-source data assimilation framework is constructed, using gridded return period wind speeds calculated to each target height layer as the initial background field, and return period wind speeds at the locations of surrounding reference meteorological stations and anemometer towers calculated to each target height layer as high-weight observation vectors. Data assimilation methods (such as CRESSMANN, Kriging, inverse distance weighted interpolation, or random forest) are used to fuse point observation information into the areal background field. The background field is then corrected grid-by-grid on a kilometer-level regular grid to generate a multi-height layer R-year return wind speed gridded dataset that maintains the spatial distribution trend and matches the characteristics of the observation location. The analysis variance or equivalent uncertainty index is output simultaneously.

[0028] This assimilation framework, based on optimal interpolation theory or variational data assimilation techniques, achieves optimal weight allocation between the background field and observation data by constructing background field error covariance matrices and observation error covariance matrices. The assimilation process considers spatial representativeness, measurement accuracy, and physical consistency constraints of the observation data. Through a multi-scale nested assimilation strategy, it maintains the spatial continuity of the background field while fully absorbing high-quality observation information. Quality control mechanisms include outlier detection, physical rationality verification, and spatial consistency constraints to ensure the reliability and engineering applicability of the assimilation results.

[0029] S700. Quality Assessment and Result Output (Preferred): Select observation data from some non-integrated wind measurement towers or reference meteorological stations as independent verification samples to perform cross-validation or independent testing on the return period wind speed gridded dataset generated in step S600, including at least extreme value deviation, root mean square error, probability distribution fit and / or uncertainty coverage test; when the indicators do not reach the preset threshold, the assimilation parameters of step S600 are recalculated based on the diagnostic results; after the threshold is met, output multi-height layer R-year return period wind speed gridded data products to provide data support for wind turbine safety selection, wind resistance verification and resource assessment.

[0030] It should be noted that the wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms, effectively solves the problem of missing high-frequency information in wind speed data in sparsely observed areas through spectrum compensation technology. It also applies the "-5 / 3" turbulent energy spectrum law to constrain the power spectral density through physical mechanisms, ensuring the spectral integrity and physical rationality of the constructed wind speed sequence. This method deeply integrates extreme value statistics theory with data assimilation technology, constructing an optimal fusion framework for multi-source information from point observations to gridded areas. This maintains the spatial continuity of the background field while fully absorbing the constraint information from high-quality observation data. In particular, by introducing a terrain correction factor and multi-height layer vertical consistency constraints, the accuracy and reliability of return period wind speed assessment under complex terrain conditions are significantly improved.

[0031] Example 2: Spectrum Compensation and AMAX Construction Based on the above embodiment 1, this embodiment 2 further elaborates on the spectrum compensation and AMAX construction process in step S100. Specifically, as follows: Figure 2 As shown, it mainly includes the following sub-steps: S101. Wind speed data acquisition and preprocessing: Obtain hourly gridded historical wind speed sequences covering the target assessment area and its outer buffer zone using ERA5 reanalysis datasets, NCEP / NCAR reanalysis data, or high-resolution numerical weather prediction models, and perform quality control on the wind speed data.

[0032] S102. FFT Transform Spectrum Analysis: The preprocessed gridded wind speed sequence is segmented according to a fixed length and overlap ratio. Within each segment, a window function and linear detrending are applied to reduce spectral leakage. An FFT transform is performed on each segment to convert the time-domain signal into a frequency-domain signal, and the power spectral density is calculated. S ( ω ), and use overlapping segment averaging to obtain robust estimates, and obtain energy distribution characteristics at different frequencies, such as Figure 3 As shown.

[0033] Specifically, Figure 3 The image shows the spectral distribution of the reanalysis wind speed sequence. The horizontal axis represents the frequency f[day]. -1 The vertical axis represents the power spectral density S(f)[m]. 2 s -2 [day], all using a logarithmic coordinate system. The solid line represents the actual spectral distribution of the reanalysis data, exhibiting typical atmospheric turbulence spectral characteristics: in the low-frequency range (f<10), -1 day -1 The power spectral density is relatively high, reflecting the long-period changes of the weather system; in the high-frequency band (f>10), it is relatively high. 1 day -1 The spectrum decays rapidly but deviates from theoretical expectations. The dashed line represents the baseline conforming to the "-5 / 3" slope, which should follow this power law relationship within the inertial sub-region. Comparison reveals a significant energy deficiency in the high-frequency band of the reanalysis data, primarily due to the spatial resolution limitations and temporal sampling frequency constraints of the numerical model. This energy deficiency directly affects the accurate identification of the annual maximum wind speed; therefore, spectral compensation based on the "-5 / 3" slope law is necessary to restore the complete turbulent energy spectrum characteristics and ensure the physical rationality and engineering reliability of subsequent extreme value statistical analysis.

[0034] S103. High-frequency energy spectrum compensation based on the "-5 / 3" slope: Identify the inertial sub-region range in the power spectrum curve and determine the frequency range that conforms to the "-5 / 3" energy spectrum slope; perform spectral compensation on high-frequency bands that deviate from the "-5 / 3" slope, so that the compensated high spectrum is close to the "-5 / 3" slope characteristic; and improve the spectral density... S ( ω The zero-order moment m0 ​​after compensation is calculated by integral calculation to ensure that the total energy does not change during the compensation process.

[0035] S104. Equivalent power spectrum construction and time domain reconstruction: The compensated high spectrum and the uncompensated low spectrum are continuously and smoothly spliced ​​at the boundary frequency to form an equivalent power spectrum; the phase information consistent with the amplitude is retained or reconstructed, and the equivalent power spectrum in the frequency domain is converted back to the time domain by performing an inverse FFT transformation to construct a wind speed time series with complete spectral characteristics.

[0036] S105. AMAX Sequence Construction: Perform year-by-year statistical analysis on the wind speed time series after spectrum compensation, extract the maximum wind speed value for each year, construct the year-by-year maximum wind speed AMAX sequence point by point, and perform statistical testing on the constructed AMAX sequence.

[0037] Preferably, in step S105, while directly extracting the historical maximum values ​​from the equivalent time-domain sequence, equivalent energy spectrum estimation is introduced to obtain candidate values ​​for the annual maximum wind speed, including: Using the equivalent energy spectrum integral formula Calculate the spectral moments of each order, where m j for j Spectral moments of order j, j=0,2 T a The sampling time interval, S ( ω ) represents the power spectral density function of turbulent wind speed after spectral compensation. ω For frequency; Through the zeroth order spectral moment m 0 and second-order spectral moments m 2. Calculate the average zero crossing rate The average frequency characteristics of wind speed fluctuations are used; the annual maximum wind speed estimation formula is adopted. Calculate the expected value of the maximum wind speed for each year, where For wind speed standard deviation, T 0 The annual effective observation duration, This indicates the expected number of extreme events within the year; Spectral method u maxThe consistency with the annual maximum value of the equivalent time-domain sequence is checked. In cases of time-domain underestimation due to missing data or event clustering, the spectral estimation is used as the candidate value for that year. The final output is the grid-by-grid AMAX and the value derived from (…). m 0, m 2) Confidence intervals obtained from uncertainty propagation.

[0038] This embodiment 2 elaborates on the technical details of spectrum compensation and AMAX construction, establishing a wind speed sequence construction method based on physical mechanism constraints. This method utilizes the "-5 / 3" energy spectrum law of turbulence theory to quantitatively compensate for high-frequency turbulence information missing in numerical models or reanalysis data due to resolution limitations, effectively restoring the complete spectral characteristics of the wind speed sequence. In particular, by introducing equivalent energy spectrum estimation technology, not only is the annual maximum wind speed directly extracted from the time-domain sequence, but independent extreme value estimates are also provided through spectral moment integral calculations, enhancing the reliability and completeness of AMAX sequence construction.

[0039] Example 3: Extreme Value Type I Distribution Parameter Estimation Based on the above embodiment 1, this embodiment 3 further elaborates on the extreme value type I distribution parameter estimation process in step S200. Specifically, as follows: Figure 4 As shown, it mainly includes the following sub-steps: S201. Preprocessing and sorting of annual maximum wind speed sequences: The obtained grid-by-grid AMAX sequences are sorted in ascending order. ,in x i For the first i The annual maximum wind speed observation value, i =1,2,…, n , n Given the sequence length; perform basic statistical analysis of the sequence and calculate the sample mean. Sample standard deviation and coefficient of variation.

[0040] S202. Construction of Empirical Distribution Function and Probability Assignment: Using the empirical distribution function Probability assignment is performed on the sorted annual maximum wind speed sequence; the cumulative probability corresponding to each wind speed value is calculated, and the correspondence between wind speed values ​​and cumulative probabilities is established.

[0041] S203. Linearization Transformation of Type I Extreme Value Distribution: Cumulative Distribution Function Based on Type I Extreme Value Distribution ,in α and μ These represent the scale and location parameters of the distribution, respectively; a double logarithmic transformation is performed on the empirical distribution function to construct a linearized sequence. This transforms the nonlinear extreme value distribution relationship into a linear one. .

[0042] S204. Distribution Parameter Estimation and Calculation: Based on the transformed linear relationship, the method of moments is used to calculate the scaling parameter. and position parameters ,in σ ( y )and σ ( x ) are respectively y sequence sum x Standard deviation of the sequence E ( x )and E ( y ) are the mean values ​​of the corresponding sequences; the maximum likelihood estimation method is used to verify the parameters, and the optimal parameters are solved by the likelihood function.

[0043] S205. Calculation and Verification of Return Period Wind Speed: Based on the estimated distribution parameters, the return period wind speed calculation formula is used. Calculate the return wind speed value corresponding to the target return period R. x R R represents the return period in years; a parameter stability testing system is established, including confidence interval estimation and goodness-of-fit test, to ensure the reliability of parameter estimation and the accuracy of return period wind speed calculation.

[0044] In Example 3, the statistical estimation method for return-time wind speed based on the Type I extreme value distribution was systematically described, establishing a complete technical chain from raw observation data to engineering design wind speed. By transforming the nonlinear extreme value distribution problem into a linear regression problem through a double logarithmic transformation, the parameter estimation process was significantly simplified and the calculation accuracy improved. Cross-validation using the method of moments and the maximum likelihood estimation method ensured the robustness and reliability of the distribution parameters. In particular, the established parameter stability testing system and goodness-of-fit diagnostic mechanism provided a quantitative evaluation basis for the applicability of the extreme value distribution model.

[0045] Example 4: Using a combination of RHtests and metadata to perform homogeneity testing on the annual maximum wind speed series observed by the reference meteorological stations. Based on the above embodiment 1, this embodiment 4 further elaborates on the process of performing a uniformity test on the annual maximum wind speed sequence observed by the reference meteorological station in step S300 using a combination of RHtests and metadata. Specifically, as follows: Figure 5 As shown, it mainly includes the following sub-steps: S301. Wind Speed ​​Time Series Homogeneity Test: The core test algorithm of RHtests, penalizing the maximum F-test, is used to identify abrupt changes in the wind speed time series. For each potential abrupt change location in the entire series, the series is divided into two segments. The sum of squared residuals between the segmented and unsegmented fits is calculated, and the F-test is used to evaluate the improvement in segmented fitting to determine the abrupt change location. The residual sequence is pre-whitened to eliminate autocorrelation interference. Based on the corrected sequence, the reliability of the F-test results is re-verified to ensure the validity of statistical inference. An empirical penalty factor is introduced to differentiate the weighting of the F-statistic at different locations, suppressing spurious significance at marginal locations and in short segments, correcting test bias, and achieving accurate detection of abrupt changes. The same process is used to iteratively search for other abrupt changes in the sub-segments, completing the recursive segmentation detection of multiple abrupt changes and verifying the significance of each abrupt change, forming a complete abrupt change detection process.

[0046] S302. Manual Verification of Abrupt Change Points: Establish a manual verification system based on metadata. By integrating station metadata and cross-validating observation records, verify key events that may lead to data non-uniformity, such as station relocation, instrument replacement, and changes in the observation environment. Eliminate irrelevant events through metadata correlation matching and ensure the timeliness matching of events and abrupt change points through multi-source cross-validation, providing a physically reliable basis for identifying data non-uniformity. S303. Wind Speed ​​Time Series Uniformity Correction: Quantitatively correct identified non-uniformity using quantile matching. Separate trends and abrupt changes using a piecewise regression model, perform piecewise linear trend fitting, and determine the benchmark segment based on the longest segment. For time series with abrupt change points, calculate the quantiles of each segment at the quantile level, calculate the deviation between each segment and the benchmark segment using the cumulative deviation method or direct difference method, establish the quantile mapping relationship with the benchmark segment, and achieve quantitative correction of the deviation segments through cubic spline interpolation.

[0047] Specifically, Figure 6 The figure shows a comparison of the long-series wind speed before and after correction at the reference meteorological station. The solid line in the figure represents the original annual maximum wind speed sequence observed at the reference station. Verification revealed two abrupt change points in 2004 and 2015 (the turning points of the dotted line in the figure), which are related to changes in the wind speed observation instruments and the observation environment at the station. The dashed line in the figure represents the annual maximum wind speed sequence after correction using the quantile matching method, whose uniformity has been significantly improved.

[0048] In summary, Example 4 of this system constructs a complete processing scheme from wind speed time series homogeneity detection to quantitative correction, forming a complete technical framework of "statistical verification - manual verification - deviation correction". By combining penalized maximum F-test with residual pre-whitening and empirical penalty factor weighting, precise location of abrupt change points and recursive detection of multiple abrupt change points are achieved. The manual verification system, utilizing metadata integration and multi-source cross-validation, provides physically reliable support for non-homogeneity identification. Specifically, using the longest segment of the segmented regression separation as a benchmark, a mapping relationship is established through quantile matching to complete quantitative correction, ensuring the statistical rationality of the correction results.

[0049] Example 5: Multi-source optimal assimilation and gridded fusion Based on the above embodiment 1, this embodiment 5 further elaborates on the multi-source optimal assimilation and lattice-based fusion process in step S600. Specifically, as follows: Figure 7 As shown, it mainly includes the following sub-steps: S601. Background-Observation Construction and Mapping: At each target height level, the gridded R-year return period wind speed generated in step S200 is used as the background field. x b The return-time wind speeds of reference meteorological stations and anemometer towers at the same altitude were used as the observation vectors. y 0. Establish the observation operator H to realize the mapping between the observation space and the model grid space, and unify the spatial reference and time reference.

[0050] S602. Error and Representativeness Modeling: Construct the background field error covariance matrix B, the observation error covariance matrix R, and the representativeness penalty matrix P. B is determined by the uncertainty of extreme value parameter estimation, spectral compensation stability, and ensemble deviation. R is determined by the homogeneity correction residual, station-tower correction regression error, and vertical extrapolation error. P increases with topographic relief, roughness difference, wind direction sector deviation, and spatial distance. Define anisotropic correlation and influence radius according to wind direction zoning and height layer.

[0051] S603. Pre-fusion interpolation and feature construction: Deterministic pre-interpolation of observation vectors is performed on a kilometer-level regular grid using the CRESSMANN method, Kriging method, or inverse distance weighted interpolation method to generate candidate fields and spatial smoothing benchmarks; and environmental features are extracted, including at least terrain slope, relative elevation, roughness length, and / or prevailing wind direction, to provide input for subsequent statistical learning and residual correction.

[0052] S604. Assimilation Solution and Machine Learning Residual Correction: Computational Analysis Field Where the gain matrix K=BH T (HBH T +R+P) -1 ; and with random forest pairs ( y0-H x b The spatial residuals are learned from environmental characteristics to obtain the statistical residual field. ,pass This method implements secondary corrections to reduce systematic biases in complex terrain.

[0053] S605. Vertical Consistency and Directional Zoning Constraints: Block joint assimilation is adopted for the state vectors of multiple altitude layers. Inter-layer covariance is established based on historical covariance and wind shear error to maintain vertical consistency. Assimilation is performed separately according to the dominant wind direction sector and continuous splicing is implemented at the sector boundary. If necessary, anisotropic variability functions are introduced to enhance the correlation along the wind direction and weaken the lateral correlation.

[0054] S606. Parameter calibration and robust control: Select the radius of influence, Kriging variogram and parabolic or exponential distance decay parameters through leave-one-out cross-validation or K-fold cross-validation, constrain the upper limit of single observation weights and use robust statistics to suppress high-leverage outliers; when the condition number exceeds the limit or the positive definiteness is not satisfied, regularize B, R and P and perform spectral pruning.

[0055] S607. Result Generation and Uncertainty Output: Form the analysis field and its analysis variance and generate quality labels. Output a multi-height layer R-year return period wind speed gridded dataset and uncertainty raster that maintains the spatial distribution trend of the background field and matches the measured characteristics at the observation point location. This is used for independent verification and threshold callback in step S700.

[0056] In this embodiment, Example 5 constructs a multi-level fusion framework from point observations to area grids. By establishing the background field error covariance matrix, observation error covariance matrix, and representativeness penalty matrix, optimal weight allocation and information fusion among different data sources are achieved. Specifically, by introducing machine learning techniques to statistically correct systematic biases under complex terrain conditions, the accuracy of gridded results in mountainous and hilly areas is effectively improved. Vertical consistency constraints and wind direction zoning assimilation strategies ensure the physical consistency of wind speeds across multiple altitude layers. This fusion method fully leverages the spatial coverage advantages of numerical models and the accuracy advantages of observational data. Through robust quality control and parameter optimization mechanisms, it provides high-precision, high-reliability gridded products for wind farm site selection under complex terrain conditions.

[0057] The objectives of this invention have been fully and effectively achieved through the above embodiments. Those skilled in the art will understand that this invention includes, but is not limited to, the contents described in the accompanying drawings and the specific embodiments described above. Although the invention has been described with reference to what is currently considered the most practical and preferred embodiments, it should be understood that the invention is not limited to the disclosed embodiments, and any modifications that do not depart from the functional and structural principles of the invention will be included within the scope of the claims.

Claims

1. A gridded evaluation method for wind speed during the return period of wind power generation, integrating data features and physical mechanisms, characterized in that, It should include at least the following steps: S100. Obtain gridded historical wind speed sequences covering the target assessment area and its outer buffer zone through numerical models or reanalysis datasets. Perform spectral analysis on the wind speed sequences to obtain the frequency domain power spectral density distribution. Based on the "-5 / 3" slope law of turbulent energy spectrum, perform spectral compensation on the energy spectrum of the high-frequency band to form an equivalent energy spectrum and perform time-domain reconstruction accordingly. Construct an AMAX sequence grid by grid. S200. The extreme value type I distribution is used to perform statistical analysis on the AMAX sequence obtained by spectrum compensation. Based on the empirical distribution linearization or maximum likelihood, the parameters are estimated to calculate the return wind speed value corresponding to the R-year target return period at each grid point location. A gridded R-year target return period wind speed basic dataset covering the target assessment area and its outer buffer zone is constructed, and parameter stability and fitting diagnostic indicators are output. S300. Collect long-term historical wind speed observation data from surrounding meteorological stations within and outside the target assessment area. Use a combination of RHtests and metadata to perform a uniformity test on the annual maximum wind speed sequence observed by the meteorological stations, identify and correct abrupt changes and non-uniformity in the wind speed data, construct the AMAX sequence for each meteorological station, and use the extreme value type I distribution to estimate the return wind speed value corresponding to the R-year return period for each meteorological station, forming a set of return period wind speeds for the meteorological stations. S400. Collect hourly wind speed observation data from the reference meteorological stations and the wind measurement towers in the target wind farm area. Based on the causes of the annual maximum wind speed, select high wind speed samples with similar weather backgrounds from the wind speed sequences of the reference meteorological stations and wind measurement towers. Establish the correction relationship from the station to the tower location. Transfer the R-year return wind speed of the reference station to the wind measurement tower location to obtain the R-year return period wind speed value at the wind measurement tower location. S500. Based on the high wind speed samples obtained from the screening, calculate the wind shear index, and according to the power law relationship, extrapolate the R-year return period wind speeds at grid locations, surrounding reference meteorological stations, and wind measurement tower locations to several wind power design height layers to obtain multi-height layer return period wind speeds for wind power generation design.

2. The wind speed gridded assessment method for the return period of wind power generation according to claim 1, characterized in that, In step S100, the spectrum compensation and AMAX construction include the following sub-steps during implementation: S101. Wind speed data acquisition and preprocessing: Obtain hourly gridded historical wind speed sequences covering the target assessment area and its outer buffer zone using ERA5 reanalysis datasets, NCEP / NCAR reanalysis data, or high-resolution numerical weather prediction models, and perform quality control on the wind speed data. S102. FFT Transform Spectrum Analysis: The preprocessed grid wind speed sequence is segmented into segments with fixed lengths and overlap ratios. Within each segment, window functions and linear detrending are used to reduce spectral leakage. Perform an FFT transform on each sequence segment to convert the time-domain signal into a frequency-domain signal, and calculate the power spectral density. S ( ω The method employs overlapping segment averaging to obtain robust estimates and acquires energy distribution characteristics at different frequencies. S103. High-frequency energy spectrum compensation based on the "-5 / 3" slope: Identify the inertial sub-region range in the power spectrum curve and determine the frequency range that conforms to the "-5 / 3" energy spectrum slope; perform spectral compensation on high-frequency bands that deviate from the "-5 / 3" slope, so that the compensated high spectrum is close to the "-5 / 3" slope characteristic; and improve the spectral density... S ( ω The zeroth moment m0 ​​after compensation is calculated by integral calculation to ensure that the total energy does not change during the compensation process; S104. Equivalent power spectrum construction and time domain reconstruction: The compensated high spectrum and the uncompensated low spectrum are continuously and smoothly spliced ​​at the boundary frequency to form an equivalent power spectrum; the phase information consistent with the amplitude is retained or reconstructed, and the equivalent power spectrum in the frequency domain is converted back to the time domain by performing an inverse FFT transformation to construct a wind speed time series with complete spectral characteristics. S105. AMAX Sequence Construction: Perform year-by-year statistical analysis on the wind speed time series after spectrum compensation, extract the maximum wind speed value for each year, construct the year-by-year maximum wind speed AMAX sequence point by point, and perform statistical testing on the constructed AMAX sequence.

3. The wind speed gridded assessment method for the return period of wind power generation according to claim 2, characterized in that, In step S105, while directly extracting the historical maximum values ​​from the equivalent time-domain sequence, equivalent energy spectrum estimation is introduced to obtain candidate values ​​for the annual maximum wind speed, including: Using the equivalent energy spectrum integral formula Calculate the spectral moments of each order, where m j for j Spectral moments of order j, j=0,2 T a The sampling time interval, S ( ω ) represents the power spectral density function of turbulent wind speed after spectral compensation. ω For frequency; Through the zeroth order spectral moment m 0 and second-order spectral moments m 2. Calculate the average zero crossing rate The average frequency characteristics of wind speed fluctuations are used; the annual maximum wind speed estimation formula is adopted. Calculate the expected value of the maximum wind speed for each year, where For wind speed standard deviation, T 0 The annual effective observation duration, This indicates the expected number of extreme events within the year; Spectral method u max The consistency with the annual maximum value of the equivalent time-domain sequence is checked. In cases of time-domain underestimation due to missing data or event clustering, the spectral estimation is used as the candidate value for that year. The final output is the grid-by-grid AMAX and the value derived from (…). m 0, m 2) Confidence intervals obtained from uncertainty propagation.

4. The wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms according to claim 1, is characterized in that... In step S200, the estimation of the extreme value type I distribution parameters includes at least the following sub-steps: S201. Preprocessing and sorting of annual maximum wind speed sequences: The obtained grid-by-grid AMAX sequences are sorted in ascending order. ,in x i For the first i The annual maximum wind speed observation value, i =1,2,…, n , n Given the sequence length; perform basic statistical analysis of the sequence and calculate the sample mean. Sample standard deviation and coefficient of variation; S202. Construction of Empirical Distribution Function and Probability Assignment: Using the empirical distribution function Probability assignment is performed on the sorted annual maximum wind speed sequence; Calculate the cumulative probability corresponding to each wind speed value and establish the correspondence between wind speed values ​​and cumulative probabilities; S203. Linearization Transformation of Type I Extreme Value Distribution: Cumulative Distribution Function Based on Type I Extreme Value Distribution ,in α and μ These represent the scale and location parameters of the distribution, respectively; a double logarithmic transformation is performed on the empirical distribution function to construct a linearized sequence. This transforms the nonlinear extreme value distribution relationship into a linear one. ; S204. Distribution Parameter Estimation and Calculation: Based on the transformed linear relationship, the method of moments is used to calculate the scaling parameter. and position parameters ,in σ ( y )and σ ( x ) are respectively y sequence sum x Standard deviation of the sequence E ( x )and E ( y ) represent the mean of the corresponding sequence; the maximum likelihood estimation method is used to verify the parameters, and the optimal parameters are solved by the likelihood function; S205. Calculation and Verification of Return Period Wind Speed: Based on the estimated distribution parameters, the return period wind speed calculation formula is used. Calculate the return wind speed value corresponding to the target return period R. x R R represents the return period in years; a parameter stability testing system is established, including confidence interval estimation and goodness-of-fit test, to ensure the reliability of parameter estimation and the accuracy of return period wind speed calculation.

5. The wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms according to claim 1, is characterized in that... In step S300, the homogeneity test of the annual maximum wind speed sequence observed by the reference meteorological stations is performed using a combination of RHtests and metadata. This process includes the following sub-steps: S301. Wind speed time series homogeneity test: The core test algorithm of RHtests, the penalized maximum F-test, is used to identify abrupt changes in the wind speed time series. For each potential abrupt change location in the entire series, the series is divided into two segments. The sum of squared residuals between the segmented and unsegmented fits is calculated. The F-test is used to evaluate the improvement of the segmented fit to determine the location of the abrupt change. The residual series is pre-whitened to eliminate autocorrelation interference. The reliability of the F-test results is re-verified based on the corrected series to ensure the effectiveness of statistical inference. An empirical penalty factor is introduced to differentiate the weighting of the F-statistic at different locations, suppressing false significance at edge locations and short segments, correcting test bias, and achieving accurate detection of abrupt changes. The same process is used to iteratively search for other abrupt changes in the sub-segments, completing the recursive segmentation detection of multiple abrupt changes and verifying the significance of each abrupt change. S302. Manual verification of mutation points: Establish a manual verification system based on metadata. By integrating station metadata and cross-validating observation records, verify key events that may cause data non-uniformity, such as station relocation, instrument replacement, and changes in the observation environment. Eliminate irrelevant events through metadata correlation matching and ensure the timeliness of events and mutation points through multi-source cross-validation. S303. Wind speed time series homogeneity correction: The identified non-homogeneity is quantitatively corrected by quantile matching method. Trends and abrupt changes are separated by piecewise regression model, piecewise linear trend fitting is performed, and the benchmark segment is determined based on the longest segment. For time series with abrupt change points, the quantiles of each segment are calculated at the quantile level. The deviation between each segment and the benchmark segment is calculated using the cumulative deviation method or the direct difference method. A quantile mapping relationship with the benchmark segment is established, and quantitative correction of the deviation segment is achieved through cubic spline interpolation.

6. The wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms according to claim 1, is characterized in that... In step S400, the process of screening high-wind-speed samples and establishing correction relationships includes: analyzing the meteorological causes of the annual maximum wind speed at the reference meteorological stations, identifying the dominant weather system type, wind direction distribution characteristics, and pressure field configuration; based on the principle of weather background consistency, screening high-wind-speed event samples under similar weather conditions from the concurrent observations of the anemometer tower and the reference stations; establishing correction relationships using the ratio method, difference method, or multiple linear regression; evaluating the accuracy of the correction method through cross-validation, selecting the optimal correction model, and achieving reliable migration of the wind speed of the reference station's return period to the location of the anemometer tower.

7. The wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms according to claim 1, is characterized in that... Step S500, the calculation of the wind shear index, includes: Based on the high-wind-speed samples obtained through screening, a power-law formula was used. Calculate the wind shear index at different observation heights α ,in V z Let z be the wind speed at the target height. V z0 The wind speed at the reference height z0 is used as the reference. The variation law of wind shear index under different wind direction sectors, different seasons, and different atmospheric stability conditions is determined through statistical analysis. A spatial interpolation model of wind shear index is established, taking into account the influence of terrain roughness and surface cover type on wind shear. A logarithmic wind profile model is used for verification and correction to ensure the applicability of wind shear index within the hub height range of wind turbine units.

8. The wind speed gridded assessment method for the return period of wind power generation, which integrates data features and physical mechanisms according to claim 1, is characterized in that... It also includes step S600, which involves multi-source optimal assimilation and lattice-based fusion, specifically including: A multi-source data assimilation framework is constructed, using gridded return period wind speeds calculated to each target height layer as the initial background field, and return period wind speeds at the locations of surrounding reference meteorological stations and anemometer towers calculated to each target height layer as high-weight observation vectors. The data assimilation method is used to fuse point observation information into the areal background field, and the background field is corrected grid by grid on a kilometer-level regular grid to generate a multi-height layer R-year return period wind speed gridded dataset covering the target assessment area.

9. The gridded evaluation method for wind speed during the return period of wind power generation, which integrates data features and physical mechanisms according to claim 8, is characterized in that... In step S600, the multi-source optimal assimilation and gridded fusion includes the following sub-steps during implementation: S601. Background-Observation Construction and Mapping: At each target height level, the gridded R-year return period wind speed generated in step S200 is used as the background field. x b The return-time wind speeds of reference meteorological stations and anemometer towers at the same altitude were used as the observation vectors. y 0. Establish the observation operator H to realize the mapping between the observation space and the model grid space, and unify the spatial reference and time reference; S602. Error and Representativeness Modeling: Construct the background field error covariance matrix B, the observation error covariance matrix R, and the representativeness penalty matrix P. B is determined by the uncertainty of extreme value parameter estimation, spectral compensation stability, and ensemble deviation. R is determined by the homogeneity correction residual, station-tower correction regression error, and vertical extrapolation error. P increases with topographic relief, roughness difference, wind direction sector deviation, and spatial distance. Define anisotropic correlation and influence radius according to wind direction zoning and height layer. S603. Pre-fusion interpolation and feature construction: Deterministic pre-interpolation of observation vectors is performed on a kilometer-level regular grid using the CRESSMAN method, Kriging method, or inverse distance weighted interpolation method to generate candidate fields and spatial smoothing benchmarks; and environmental features are extracted, including at least terrain slope, relative elevation, roughness length, and / or prevailing wind direction, to provide input for subsequent statistical learning and residual correction. S604. Assimilation Solution and Machine Learning Residual Correction: Computational Analysis Field Where the gain matrix K=BH T (HBH T +R+P) -1 ; and with random forest pairs ( y 0-H x b The spatial residuals are learned from environmental characteristics to obtain the statistical residual field. ,pass Implement secondary corrections to reduce systematic biases under complex terrain; S605. Vertical Consistency and Directional Zoning Constraints: Block joint assimilation is adopted for the state vectors of multiple height layers. Inter-layer covariance is established based on historical covariance and wind shear error to maintain vertical consistency. Assimilation is performed separately according to the dominant wind direction sector and continuous splicing is implemented at the sector boundary. If necessary, anisotropic variability functions are introduced to enhance the correlation along the wind direction and weaken the lateral correlation. S606. Parameter calibration and robust control: Select the radius of influence, Kriging variogram and parabolic or exponential distance decay parameters through leave-one-out cross-validation or K-fold cross-validation, constrain the upper limit of single observation weights and use robust statistics to suppress high leverage outliers; when the condition number exceeds the limit or the positive definiteness is not satisfied, regularize B, R and P and spectral pruning are performed. S607. Result Generation and Uncertainty Output: Form the analysis field and its analysis variance and generate quality labels. Output a multi-height layer R-year return period wind speed gridded dataset and uncertainty raster that maintains the spatial distribution trend of the background field and matches the measured characteristics at the observation point location. This is used for independent verification and threshold callback in step S700.

10. The wind speed gridded assessment method for the return period of wind power generation based on the fusion of data features and physical mechanisms according to claim 8 or 9, characterized in that, It also includes step S700 for quality assessment and result output, specifically including: Select observation data from some non-integrated wind measurement towers or reference meteorological stations as independent verification samples to perform cross-validation or independent testing on the return period wind speed gridded dataset generated in step S600, including at least extreme value deviation, root mean square error, probability distribution fit and / or uncertainty coverage test; when the indicators do not reach the preset threshold, the assimilation parameters of step S600 are recalculated based on the diagnostic results; after the threshold is met, output multi-height layer R-year return period wind speed gridded data products to provide data support for wind turbine safety selection, wind resistance verification and resource assessment.

11. A gridded wind speed assessment system for the return period of wind power generation, comprising multiple modules, characterized in that, The wind speed gridding assessment method for the return period of wind power generation, which integrates data features and physical mechanisms as described in any one of claims 1 to 10, is used to perform such method.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the wind speed gridding evaluation method for the return period of wind power generation that integrates data features and physical mechanisms as described in any one of claims 1 to 10.

Images