A wind and light resource refined estimation method based on numerical simulation and observation constraint

Abstract

The present application relates to renewable energy assessment and weather forecasting technical field, especially to a kind of wind and light resource fine estimation method based on numerical simulation and observation constraint. Global climate model driving field set is constructed, and the optimal GCM combination is screened out;Regional climate model framework is constructed;High spatio-temporal resolution simulated wind speed and solar radiation data are obtained;Generate kilometer level, hourly wind and light resource data set of history and future. The present application provides a kind of wind and light resource fine estimation method based on numerical simulation and observation constraint, by introducing high resolution, multi-source fusion ground surface wind speed and solar radiation long sequence historical observation data, using quantile mapping method to carry out systematic deviation correction to simulation result, realize statistical downscaling processing under the observation data constraint, finally obtain the wind resource data with 5 kilometers horizontal resolution, so as to significantly improve the high spatio-temporal resolution information capture ability to future wind resource change.

Description

Technical Field

[0001] This invention relates to the field of renewable energy assessment and weather forecasting technology, belonging to the interdisciplinary technical field of computational fluid dynamics (CFD) simulation and weather forecasting devices, and particularly to a refined forecasting method for wind and solar resources based on numerical simulation and observation constraints. Background Technology

[0002] Currently, international research on wind and solar energy resource forecasting primarily relies on multi-model datasets such as the Coupled Model Intercomparison Project (CMIP6) and the Regional Climate Downscaling Project (CORDEX) launched by the World Climate Research Programme (WCRP). These datasets integrate dynamic downscaling, statistical downscaling, and the emerging artificial intelligence (AI) downscaling techniques, providing fundamental data support for global and regional climate research. However, the current technical system still has significant limitations in meeting the needs of refined resource assessment under complex terrain conditions in China. These limitations are mainly reflected in the following aspects: First, there are bottlenecks in data resolution and timeliness—the original resolution of the CMIP6 global climate model is typically only 100-200 kilometers, making it difficult to accurately depict the details of resource distribution under my country's complex terrain; although regional models such as CORDEX have undergone downscaling, their typical resolution (approximately 12-50 kilometers) is still insufficient to meet the needs of kilometer-level assessment, and the data release cycle is relatively long, severely limiting the real-time independent assessment capabilities of domestic operations. Secondly, insufficient data accuracy and model limitations coexist. Existing downscaling products generally exhibit significant systematic biases and lack long-term historical and future projection data at the kilometer-to-hour scale. Most data products only provide daily or monthly average results, failing to capture the short-term fluctuations and local evolution characteristics of wind and solar energy resources. This masks many key spatiotemporal change signals in resource assessment, making it difficult to support detailed planning at the wind farm and photovoltaic power plant level and quantitative assessment of power generation safety risks. Although dynamic and statistical downscaling techniques are relatively mature, the former is limited by high computational costs, while the latter often lacks a sufficient description of physical processes.

[0003] To address the aforementioned issues, this invention aims to resolve three core challenges in wind energy resource forecasting in China: insufficient resolution, limited data accuracy, and weak autonomy. This method can generate long-term historical and future wind and solar resource datasets with kilometer-level resolution (5 km), hourly timescales, and bias corrections. This effectively fills the gap in high-precision forecasting data in this field, providing a crucial data foundation for reliable prediction of wind and solar power generation and risk assessment of system operational safety.

[0004] To address this issue, a refined prediction method for wind and solar resources based on numerical simulation and observation constraints is designed to provide an alternative technical solution. Summary of the Invention

[0005] Therefore, it is necessary to provide a refined prediction method for wind and solar resources based on numerical simulation and observation constraints to address the aforementioned technical problems.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] A refined prediction method for wind and solar resources based on numerical simulation and observation constraints, comprising the following steps:

[0008] S1: Construct a global climate model driving field ensemble, select multiple GCMs with low, medium and high sensitivity from the CMIP6 international model based on the balanced climate sensitivity index, evaluate their simulation performance on near-surface wind speed and solar radiation in the target area, and screen out the optimal GCM combination.

[0009] S2: Construct a regional climate model framework and improve the simulation capabilities for wind speed and radiation through parameterized scheme combination optimization and high-precision surface information assimilation.

[0010] S3: Use the optimized system built in steps S1 and S2 to perform dynamic downscaling simulation and obtain high spatiotemporal resolution simulated wind speed and solar radiation data;

[0011] S4: Based on the quantile mapping method, the simulated values ​​output by dynamic downscaling are statistically downscaled and biased with high-resolution observation data to generate kilometer-level, hourly historical and future wind and light resource datasets;

[0012] S5: Based on the corrected data, calculate wind power density and photovoltaic power generation potential, and identify future trends in wind and solar resources.

[0013] As a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, the construction of the GCM driving field in step S1 is as follows:

[0014] Based on the balanced climate sensitivity index, several GCMs representing low, medium, and high sensitivity were selected from CMIP6.

[0015] By combining historical observation data, a multi-index evaluation system was used to select the GCM combination with the best performance in simulating wind speed and radiation in the target area.

[0016] As a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, the optimization of the regional climate model in step S2 is as follows:

[0017] A combined experiment was conducted on key parameterization schemes for cumulus convection, planetary boundary layer, and radiative transfer.

[0018] High-resolution satellite remote sensing surface information is introduced to replace the original surface data in the model, thereby improving the accuracy of land surface process simulation.

[0019] As a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, the quantile mapping method in step S4 is as follows:

[0020] Interpolate the simulated values ​​from the dynamic downscaling output to the high-resolution observation grid;

[0021] Within the historical reference period, the cumulative probability distribution functions of the observed and simulated values ​​are calculated separately, and the transfer function is constructed.

[0022] The transfer function is used to correct the deviation of future simulated values ​​and retain their relative change signals to generate high-resolution corrected data.

[0023] As a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, in step S5, the formula for calculating wind power density is:

[0024] ;

[0025] in, For wind power density, air density, The wind speed at the turbine height.

[0026] In a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, in step S5, the calculation of photovoltaic power generation potential is based on the following formula:

[0027] ;

[0028] in, For performance ratio, This represents actual solar radiation. Standard solar radiation, Rated power;

[0029] Further determined by battery temperature The calculation shows that:

[0030] ;

[0031] in, , .

[0032] As a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, the observation data is 5 km resolution, hourly surface wind speed and solar radiation data provided by CLDAS-V3.0.

[0033] As a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, the dynamic downscaling adopts the RegCM4.4 regional climate model, and the simulation output is surface wind speed and solar radiation data at 18 km and hourly.

[0034] As a preferred embodiment of the refined wind and solar resource prediction method based on numerical simulation and observation constraints provided by the present invention, the statistical downscaling reduces the simulation data from 18 kilometers to 5 kilometers and simultaneously completes the systematic bias correction.

[0035] It is clear without a doubt that the technical solution described above in this application can solve the technical problem that this application aims to address.

[0036] Meanwhile, through the above technical solutions, the present invention has at least the following beneficial effects:

[0037] 1. This invention provides a refined prediction method for wind and solar resources based on numerical simulation and observation constraints. By introducing high-resolution, multi-source fusion of long-series historical observation data on surface wind speed and solar radiation, and using the quantile mapping method to systematically correct the bias of the simulation results, it achieves statistical downscaling under the constraint of observation data, and finally obtains wind resource data with a horizontal resolution of 5 kilometers. This significantly improves the ability to capture high spatiotemporal resolution information on future wind resource changes. At the same time, based on the optimization of dynamic downscaling numerical simulation, it realizes long-term wind resource simulation and prediction on an hourly scale under the dynamic framework.

[0038] 2. By using a combined approach of “global model optimization + regional model optimization + statistical downscaling”, a high-resolution wind and solar resource dataset with a kilometer (5 km) scale and hourly output is generated, overcoming the limitations of existing CMIP6 and CORDEX data in terms of spatial resolution and release cycle. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of the present invention;

[0041] Figure 2 This is a schematic diagram of the statistical downscaling method of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0043] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0044] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the present invention can be combined with each other.

[0045] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0046] Example 1

[0047] Reference Figure 1 A refined prediction method for wind and solar resources based on numerical simulation and observation constraints.

[0048] The refined wind resource forecasting method based on numerical simulation and observational constraints consists of three parts: global model optimization, regional model dynamic downscaling, and observational-constrained statistical downscaling. The specific steps are as follows: Figure 1 As shown, the steps are as follows:

[0049] Step 11 involves inputting observation data, which is known and provided by CLDAS.

[0050] Step 12 involves using the quantile mapping method to establish a statistical downscaling model. The specific method is as follows:

[0051] Step 12-1: Interpolate the coarse-resolution simulated values ​​onto the grid corresponding to the observed data using a spatial linear interpolation method to achieve the transformation from 18 to 5, forming a new grid;

[0052] Here, "coarse-resolution simulated values" refers to the 18km resolution dynamic downscaling simulation data of surface wind speed and solar radiation (the result of step 23); "high-resolution multi-source fusion observation data grid" refers to the 5km high-resolution CIDAS30 observation data grid (the grid corresponding to the observation data in step 11).

[0053] Step 12-2: On the new grid from Step 12-1, within the selected reference time period, calculate the cumulative probability distribution functions of the observed and simulated values ​​respectively, and construct the transfer function between them.

[0054] Specifically, (step 12-2-1) the formula for the cumulative probability distribution function (CDF) of the observed data is:

[0055] ;

[0056] in, The cumulative probability density function of observations, where o represents observations, c represents time accumulation, and n represents different times. This represents the number of data points in the observed data, and Less than or equal to ;

[0057] (Step 12-2-2) The formula for the cumulative probability distribution function (CDF) of the simulated values ​​is:

[0058] ;

[0059] in, This represents the number of data points in the simulated data, and Less than or equal to x.

[0060] In addition, an inverse CDF is introduced:

[0061] inverse This is the inverse function of the CDF. For a given probability (quantile) p, the inverse CDF gives the corresponding value. , making In empirical CDF, this is typically achieved by sorting the data and taking the value of a specific quantile.

[0062] (Step 12-2-3) The transfer function is a composite function of the observed inverse CDF and the simulated CDF:

[0063] This function The mathematical expression is a composite function of the inverse CDF of the observed values ​​and the CDF of the model simulation values:

[0064] ;

[0065] How to understand, for a specific model simulation value ;

[0066] a. Use the pattern first. Find the quantile position of this value in the pattern distribution. :

[0067] ;

[0068] b. Then, using the inverse of observation Find the position of the same quantile in the observation data The corresponding value:

[0069] ;

[0070] this that is The value after correction by the transfer function is processed by applying this process to all quantiles, thus forming the complete transfer function.

[0071] Step 21 uses known data to construct GCM drivers covering different climate sensitivities;

[0072] The GCM ensemble models used in 21-1, which cover different climate sensitivities, exhibit significant differences in their responses to greenhouse gas forcing, displaying low, medium, and high equilibrium climate sensitivities. This difference is a major source of uncertainty in future climate projections.

[0073] Methodology: Based on the equilibrium climate sensitivity indices of each Global Climate Model (GCM), multiple GCMs representing low, medium, and high sensitivity were systematically selected from international modeling programs such as CMIP6 to form a comprehensive and representative initial set of driving fields. The overall sensitivity of a GCM cannot fully represent its simulation capability for a specific region and specific variables. For wind and solar resource simulation, the model's ability to reproduce key physical processes such as atmospheric circulation, wind speed, and radiation is crucial.

[0074] The global model climate data input from CMIP6 and other multi-model global climate data includes 6-hourly atmospheric variables such as wind speed, temperature, humidity, and the three-dimensional field of surface pressure (at different pressure layers), as well as surface variables such as sea surface temperature (SST) and soil moisture. These are the necessary boundary conditions to drive the dynamic downscaling of climate models.

[0075] The fundamental purpose of dynamic downscaling is to transform the relatively coarse large-scale climate information provided by global climate models (GCMs) into more refined and reliable regional-scale climate information. This compensates for the shortcomings of GCMs in describing regional details. Specifically, it generates physically consistent regional climate details through a higher-resolution dynamic framework and more advanced physical process descriptions. The fundamental purpose of dynamic downscaling is not only to obtain hourly surface wind speed and solar radiation data, but also to compensate for the deficiencies of global models in describing regional physical processes and reduce the uncertainty in wind and solar resource forecasting. However, because dynamic downscaling has high computational and storage requirements, obtaining kilometer-scale results also requires joint statistical downscaling.

[0076] The specific combination method is as follows:

[0077] Dynamic downscaling yields 18 km / hour data, while statistical downscaling yields 5 km / hour data. Essentially, dynamic downscaling achieves high temporal resolution, preserving the multi-temporal-scale physical characteristics of the landscape (dynamic downscaling includes descriptions of physical processes). Statistical downscaling, on the other hand, achieves higher spatial resolution, taking into account observational constraints. One consideration is the dynamic process, and the other is the correction of observational biases.

[0078] 21-2 is used for process-oriented performance evaluation: Based on different sensitivity combinations, the simulation performance of each GCM on near-surface wind speed and surface shortwave radiation in the study area (e.g., China) is further evaluated over a historical reference period. A comprehensive evaluation system including spatial correlation coefficient and root mean square error is used to select the GCM that performs best on key variables.

[0079] High-resolution, multi-source fused surface wind speed observation data is used as input data for the statistical downscaling model.

[0080] The high-resolution multi-source fusion data for statistical downscaling input includes surface wind speed and solar radiation observation data, with a spatial resolution of 5 km and a temporal resolution of 1 hour.

[0081] The observational data is primarily used to calculate the cumulative probability distribution function (CDF), construct the transfer function, and correct the simulated wind speed and solar radiation sequences. The quantile mapping method is employed, with the following steps: coarse-resolution simulated values ​​are spatially interpolated onto a high-resolution multi-source fused observational data grid (specifically, CLDAS 3.0 multi-source fused data with a 5km resolution grid). Then, within a selected reference period (specifically, 1998-2022), the cumulative probability distribution functions of the observed and simulated values ​​are calculated separately, and a transfer function between them is constructed. Using this transfer function, the cumulative probability distribution function of the simulated values ​​within the prediction period is corrected, achieving spatial downscaling while reducing model simulation errors.

[0082] By combining the above-mentioned strategy of "sensitivity coverage" and "process effect assessment", the final selected set of driving fields can not only reflect the overall uncertainty range of future climate change response, but also ensure that the driving data has a more reliable performance basis for key physical processes in regional wind and solar resource simulation.

[0083] Among them, select a few low-sensitivity GCMs, such as a1a2a3; medium-sensitivity GCMs, such as b1b2b3; and high-sensitivity GCMs, such as c1c2c3. These 9 GCMs are combined into multiple low-medium-high combinations, including a1-b1-c1, a1-b2-c1, a1-b3-c1, etc. Then, the performance of these combinations is simulated to select the optimal combination.

[0084] Among them, the driving field is used to perform dynamic downscaling future simulation and prediction. Once the driving field is selected, 22-1 and 22-2 are also completed. The driving field is used to drive the regional model to carry out simulation and prediction experiments of long-term changes, which is dynamic downscaling.

[0085] Step 22 is to determine the combination of parameterization schemes that can most accurately characterize the key physical processes of the regional climate system and construct a high-performance regional simulation framework.

[0086] 22-1: Combinatorial Optimization of Regional Model Parametric Schemes: The simulation results of regional climate models are strongly dependent on their parametric schemes because these schemes represent subgrid-scale physical processes (such as cumulus convection, boundary layer turbulence, and radiative transfer) that cannot be explicitly resolved by the model grid. Different schemes exhibit varying performance across different geographical regions and climatic contexts.

[0087] Key Scheme Identification: Focusing on parameterization schemes that have a direct and significant impact on wind and solar resource simulation, these mainly include: Cumulus convection parameterization scheme: Dominating precipitation and vertical thermal structure, affecting cloud cover and radiation. Planetary boundary layer scheme: Directly determining the vertical exchange of near-surface wind speed, temperature, and humidity. Radiative transfer scheme: Controlling the shortwave solar radiation received by the Earth's surface and the outward longwave solar radiation.

[0088] Combined experimental design: Using the controlled variable method, multiple combinations of different parameterization schemes are designed to conduct long-term historical climate simulation experiments.

[0089] Comprehensive performance evaluation: The simulation results of different combinations are compared and verified with high-resolution observation datasets using multiple variables and multiple indicators to evaluate their simulation capabilities for elements such as temperature, precipitation, wind speed, and radiation. In this way, the parameterization scheme combination with the best simulation performance is selected as the core regional model framework of this study.

[0090] 22-2 Assimilation and Integration of High-Precision Surface Information: Based on the optimized model framework, more realistic underlying surface information is introduced to further reduce the systematic errors of the model in terms of land surface processes.

[0091] Basis: Surface properties (such as vegetation type and albedo) exert crucial control over near-surface meteorological elements by influencing the exchange of energy, momentum, and moisture between the land and the atmosphere. Using potentially outdated or distorted global average surface data embedded in climate models is one of the significant sources of simulation error.

[0092] Method: High-resolution data that is more consistent with the actual situation in China, such as actual surface vegetation cover type and true surface albedo, obtained from satellite remote sensing inversion, will be dynamically integrated or replaced into the selected regional model framework.

[0093] This approach can more accurately characterize the optical properties, roughness, and evapotranspiration efficiency of the Earth's surface, thereby directly improving the model's simulation of the surface energy balance and boundary layer structure, effectively correcting the resulting systematic biases in wind speed and radiation, and allowing step 23 to obtain simulation data from the final "regional model framework" in 22-2.

[0094] Combining steps 21 and 22 forms an optimized system for simulating and predicting wind and solar resources.

[0095] Step 23 is to obtain the simulation results, which are the results of the dynamic downscaling output in step 22. Post-processing is required to extract the two variables: surface wind speed and solar radiation. These two variables are the simulation data.

[0096] The observation data and simulation data in step 12 are labeled as follows: and ;

[0097] Step 31 includes 31-1 bias correction, which uses the historical distribution of the observed data to correct for systematic biases in the historical period of the model.

[0098] The steps for deviation correction are as follows: Step 31-1:

[0099] 31-1 Calculate the simulated future relative change delta:

[0100] ;

[0101] Formula explanation: This ratio It quantifies the future changes predicted by the model. If... >1 indicates that the model predicts that events of this magnitude will strengthen in the future; if A value less than 1 indicates a weakening effect. This signal comes directly from the model's raw output and has not been corrected by the observed data.

[0102] This represents the future simulated value (corresponding to the time dimension f), specifically the variable in step 12-2-2. It corresponds to (but step 12-2-2 is historical, while this includes both historical and future periods, used to calculate the change delta), relative to historical simulation values. For the reference time period in step 12, used for;

[0103] 31-2 Applying transfer functions to future values:

[0104] ;

[0105] Formula explanation: We will use future simulated values Input pass function.

[0106] Output value The meaning is: in real historical observations, compared with future simulated values. The amount of precipitation that has the same probability of occurrence.

[0107] This step completed the... Deviation correction.

[0108] 31-3 Final Synthesis Result

[0109] The "benchmark value" after deviation correction is combined with the "relative change signal" predicted by the model.

[0110] ;

[0111] The complete expansion of the final formula:

[0112] ;

[0113] Step 32 generates the final product;

[0114] Step 33 is the application.

[0115] Example 2

[0116] refer to Figure 2 Based on the above embodiment one, a specific application method is disclosed:

[0117] 1. Regional model dynamics downscaling

[0118] This model is based on dynamic downscaling simulations performed using the RegCM4.4 regional climate model developed by the Italian Research Centre for Theoretical Physics. The RegCM series models use sigma coordinates vertically, an Arakawa B grid difference scheme horizontally, and an exponential relaxation time-varying boundary scheme for the model's side boundaries. The model has now reached its fourth version, RegCM4 (Giorgi et al., 2012). Compared to RegCM3, RegCM4 features significant changes to its architecture. The model code has been rewritten based on the Fortran 2003 standard, and it includes two-dimensional partitioning, parallel output, and other functions, resulting in better parallel efficiency and scalability. Numerous adjustments and improvements have also been made to the physical processes, adding more options for physical parameterization schemes. These include Holtslag and UW schemes for planetary boundary layers, Grell, Emanuel, and Tiedtke schemes for cumulus convection, and BATS and CLM schemes for land surface processes. The model supports various data sources as side boundary forcings, including different reanalysis data and multiple sets of CMIP5 global model results.

[0119] Before conducting long-term simulations, the simulation performance of the RegCM4.4 model in China and East Asia was first optimized and improved (Gao et al., 2017; Wu et al., 2021). Based on this, four global models of CMIP6 (BCC-CSM2-MR, MPI-ESM1-2-LR, IPSL-CM6A-LR, and MIROC-ES2L) were selected as driving fields (the selection of global models was based on the CORDEX Phase III framework). Simulations were completed for low, medium, and high emission scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5) with a horizontal resolution of 18 km and hourly simulations. The land surface process scheme used was CLM3.5, the convection parameterization scheme was Emanuel, the atmospheric radiative transport process used NCAR CCM3, the planetary boundary layer scheme was Holtslag, and the large-scale precipitation scheme used SUBEX. The simulation area is China and its surrounding regions, with the historical simulation period from 1995 to 2014. The simulation period under different future emission scenarios is from 2015 to 2100, thus obtaining hourly dynamic downscaling results for surface wind speed and solar radiation.

[0120] 2. High spatiotemporal resolution, long-sequence surface wind speed and solar radiation observation data

[0121] To better meet the needs of numerical weather prediction, climate research, ecological monitoring, agricultural production and refined meteorological services for high-resolution, multi-element land surface real-time products.

[0122] CLDAS provides high-resolution, long-series observational data for statistical downscaling, and provides constraints for the transition from dynamic downscaling to statistical downscaling.

[0123] 3. Quantile mapping bias correction

[0124] When conducting local-scale wind energy variation prediction studies, results with a resolution higher than 18 km are required. Therefore, based on the collection and organization of multi-source fusion gridded observation data (5 km) developed by the China Meteorological Administration, this project further statistically downscales the dynamic downscaling simulation results (18 km) using the quantile mapping method (QDM) technique. Figure 2The process ultimately yielded high-resolution wind energy resource projection data (5 km). First, the simulated surface wind speed and solar radiation values, dynamically downscaled at 18 km resolution, were spatially interpolated onto a 5 km resolution CLDAS3.0 observation data grid. Then, the simulation results from the regional climate model were statistically downscaled using quantile-mapping (QM) (Cannon et al., 2015; Tong et al., 2017; Han et al., 2018; Han et al., 2019). Quantile mapping was performed on hourly sequences for each grid point, using surface wind speed data from the multi-source fusion analysis product CLDAS3.0 as the reference observation data. Within the selected historical reference period of 1998–2022, the cumulative distribution function (CDF) of the observed and model simulation values ​​was calculated, and a transfer function (TF) was constructed between them. The transfer function was then used to correct the CDF of the simulated values ​​for all simulation periods (including future projections). While correcting errors, the simulation results of the long-term dynamic downscaling of surface wind speed at 18 km resolution were further statistically downscaled to a 5 km resolution grid. Ultimately, refined (5 km, hourly) wind and solar resource forecasts for the Chinese region under low, medium, and high emission scenarios were obtained.

[0125] Figure 2 In statistical downscaling methods, the transfer function is a crucial bridge, a lookup table or function established within a historical reference period to map simulated quantiles to observed quantiles. It is used to quantitatively describe and correct systematic biases in climate models. By combining it with "relative change (Delta)," it perfectly preserves the absolute or relative trends predicted by climate models while correcting biases, thus generating high-resolution data that is both accurate and contains signals of future climate change.

[0126] 4. Methods for identifying future changes in regional landscape resource potential

[0127] High spatiotemporal resolution wind and solar resource forecasting data can identify future changes in the potential of wind and solar resources in China and different regions. Wind power density, which is the amount of wind energy available per square meter, reflects the wind energy potential of a region. Calculating wind power density directly reveals the amount of usable wind energy per square meter in a region, facilitating the assessment of wind energy abundance. It is commonly used to evaluate the abundance of wind energy resources in a given area. The formula for calculating wind power density is:

[0128] (1)

[0129] in, Wind power density (unit: watts per square meter, W / m²) 2 ), It is the air density (usually taken as 1.225 kg / m³). 3 (This depends on factors such as temperature, humidity, and air pressure). This refers to the wind speed at turbine height (unit: meters per second, m / s). Measuring wind speed: First, it is necessary to measure or obtain the wind speed v (unit: m / s) for a specific area. Determining air density: Air density... It will vary depending on local temperature, air pressure, and humidity. Under standard conditions, air density is typically taken as 1.225 kg / m³. 3 However, the actual air density is not constant. This method determines the appropriate wind shear index by classifying the terrain in detail according to the building structure load code, and processes the wind speed under non-standard conditions based on the air density data, and finally outputs the wind speed at the turbine height under standard air density.

[0130] Solar resources can be estimated by calculating photovoltaic (PV) power generation. The calculation uses the method described in (Jerez et al., 2015), which comprehensively considers the effects of temperature, solar radiation, and wind speed on PV power generation. The formula is as follows:

[0131] (2)

[0132] in, and These represent downward shortwave solar radiation (W m⁻²) and solar radiation under standard atmospheric conditions (1000 W m⁻²), respectively. PR is the performance ratio of the photovoltaic cell, which takes into account all losses due to the increase in cell temperature and is estimated using formula (3) (Mavromatakis et al., 2010; Davy and Troccoli, 2012).

[0133] (3)

[0134] In equation (3), It is the power thermal coefficient, taken as -0.005 (Jerez et al., 2015). and These represent the battery temperature and the reference temperature, respectively. The reference temperature is set to 25°C, and the battery temperature is modeled based on Chenni et al. (2007).

[0135] (4)

[0136] In formula (4) It is the ambient temperature (°C) around the battery. V is solar radiation (W m⁻²), and V is wind speed (ms⁻¹). , , and It is the battery temperature coefficient, which depends on the battery characteristics and reflects the battery's heat transfer properties. , , and Generally, 4.3℃, 0.943℃, 0.028℃, and m2 W are used. -1 and -1.528℃sm -1 (Jerez et al, 2015). Under standard temperature and radiation conditions, the power generation reaches full capacity. Finally, the photovoltaic power generation potential can be obtained by combining formulas (2)-(4).

[0137] Example 3

[0138] Based on the above embodiment one, specific application directions are disclosed.

[0139] The refined forecasting method based on numerical simulation and observational constraints, applied to the planning and forecasting optimization of energy and power systems (G06Q 10 / 04, G06Q 50 / 06), deeply couples "climate science" and "energy engineering," and can be applied to the following energy and power planning:

[0140] (1) Energy Project Investment Risk Assessment and Financial Modeling: Before banks or investment companies provide financing for wind farm projects with a lifespan of up to 25 years, a comprehensive assessment of their power generation revenue and financial feasibility throughout their entire life cycle is necessary. Using refined forecasting methods, hourly wind speed and power generation data for the next 25 years can be generated for the wind farm site. By analyzing this long-term series, potential climate change trends can be identified, such as whether the average wind speed in the region is showing a slow decline. Furthermore, this method not only provides a definite power generation estimate but also outputs probabilistic assessment results (such as annual power generation at different confidence levels like P50 and P90), thereby effectively quantifying project return risks and providing a scientific basis for investment decisions.

[0141] (2) Long-term power system planning – ensuring grid reliability: Based on the hourly wind speed and solar radiation data for the next 30 years (e.g., 2025–2055) provided by the method, the output curves of wind power and photovoltaic power plants (i.e., power generation time series data) can be converted. Combining different climate change paths (e.g., SSP2-4.5 medium emission scenario, SSP5-8.5 high emission scenario) and energy policy settings, multiple future renewable energy output scenarios are generated. These long-term time series data are input into the power system planning model (e.g., capacity expansion model), and taking into account factors such as load growth, unit retirement, and technological cost evolution, the optimization solution is performed with the goal of "meeting electricity demand and minimizing total system cost" to determine the optimal power structure configuration (e.g., 50GW wind power, 80GW photovoltaic, 20GW / 80GWh energy storage, etc.) to support the construction of a safe, economical, and low-carbon future power system.

[0142] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A refined prediction method for wind and solar resources based on numerical simulation and observation constraints, characterized in that, The steps are as follows: S1: Construct a global climate model driving field ensemble, select multiple GCMs with low, medium and high sensitivity from the CMIP6 international model based on the balanced climate sensitivity index, evaluate their simulation performance on near-surface wind speed and solar radiation in the target area, and screen out the optimal GCM combination. S2: Construct a regional climate model framework and improve the simulation capabilities for wind speed and radiation through parameterized scheme combination optimization and high-precision surface information assimilation. S3: Use the optimized system built in steps S1 and S2 to perform dynamic downscaling simulation and obtain high spatiotemporal resolution simulated wind speed and solar radiation data; S4: Based on the quantile mapping method, the simulated values ​​output by dynamic downscaling are statistically downscaled and biased with high-resolution observation data to generate kilometer-level, hourly historical and future wind and solar resource datasets; S5: Based on the corrected data, calculate wind power density and photovoltaic power generation potential, and identify future trends in wind and solar resources.

2. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, In step S1, the construction of the GCM-driven field follows these steps: Based on the balanced climate sensitivity index, several GCMs representing low, medium, and high sensitivity were selected from CMIP6. By combining historical observation data, a multi-index evaluation system was used to select the GCM combination with the best performance in simulating wind speed and radiation in the target area.

3. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, In step S2, the optimization of the regional climate model proceeds as follows: A combined experiment was conducted on key parameterization schemes for cumulus convection, planetary boundary layer, and radiative transfer. High-resolution satellite remote sensing surface information is introduced to replace the original surface data in the model, thereby improving the accuracy of land surface process simulation.

4. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, In step S4, the quantile mapping method follows these steps: Interpolate the simulated values ​​from the dynamic downscaling output to the high-resolution observation grid; Within the historical reference period, the cumulative probability distribution functions of the observed and simulated values ​​are calculated separately, and the transfer function is constructed. The transfer function is used to correct the deviation of future simulated values ​​and retain their relative change signals to generate high-resolution corrected data.

5. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, In step S5, the formula for calculating wind power density is: ； in, For wind power density, air density, The wind speed at the turbine height.

6. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, In step S5, the calculation of photovoltaic power generation potential is based on the following formula: ； in, For performance ratio, This represents actual solar radiation. Standard solar radiation, Rated power; Further determined by battery temperature The calculation shows that: ； in, , , It is the power thermal coefficient.

7. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, The observation data are 5 km resolution, hourly surface wind speed and solar radiation data provided by CLDAS-V3.

0.

8. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, The dynamic downscaling was performed using the RegCM4.4 regional climate model, and the simulation output was 18 km hourly surface wind speed and solar radiation data.

9. The method for refined prediction of wind and solar resources based on numerical simulation and observation constraints according to claim 1, characterized in that, The statistical downscaling reduces the simulated data from 18 kilometers to 5 kilometers and simultaneously corrects for systematic biases.

Images