Wind energy resource assessment method and device based on flow matching model and electronic equipment

Abstract

The invention provides a wind energy resource assessment method and device based on a flow matching model. The method comprises the steps of obtaining current earth system mode simulation data of a to-be-assessed region; based on a pre-trained time sequence flow matching model, according to the current earth system mode simulation data, obtaining target wind speed data of the to-be-evaluated region; the time sequence flow matching model is used for modeling a continuous conversion path from the earth system mode simulation data to the meteorological monitoring data based on an ordinary differential equation, and is obtained by training and optimizing according to an earth system mode simulation data sample and a meteorological monitoring data sample; and determining a wind energy resource assessment result of the to-be-assessed region according to the target wind speed data. According to the method, the earth system mode simulation data and the meteorological monitoring data are combined, and the time sequence flow matching model is applied to fit the velocity field of the ordinary differential equation, so that accurate wind speed assessment from the inter-year scale to the inter-year scale is realized, the accuracy and completeness of wind energy resource assessment are remarkably improved, and the computing resource demand is reduced.

Description

Technical Field

[0001] This invention relates to the field of wind energy resource assessment technology, and in particular to a wind energy resource assessment method and apparatus based on a flow matching model. Background Technology

[0002] Driven by the global energy structure transformation and the "dual carbon" goal, wind power, as an important component of clean energy, has shown a rapid growth trend in recent years. Through the construction of large-scale new energy bases, China is accelerating the decarbonization process of its energy sector. However, in the development stage of wind energy resources, conventional meteorological monitoring devices and technologies are insufficient to effectively cover vast areas. Currently, wind energy resource assessment mainly relies on Earth System Model (ESM) simulation data and on-site meteorological monitoring data.

[0003] The Earth's Meteorological Model (ESM) is a mathematical model based on the material-energy exchange patterns between Earth's spheres. It provides historical, current, and future meteorological data through numerical simulations. Reanalysis datasets (such as ERA5 and MERRA) reconstruct long-term meteorological sequences globally by assimilating multi-source monitoring data from satellites and radiosondes. Climate model data are used to predict future climate change scenarios. ESM data has decadal-scale temporal coverage, and combined with established meteorological-electricity conversion mechanism models, it can simulate the wind power potential of any sea area. However, global models typically have a spatial resolution of at least 25 kilometers and a temporal resolution of at least 1 hour, which is insufficient for refined assessments. Higher-resolution simulation data requires dynamic downscaling through regional models.

[0004] On the other hand, ground-based monitoring equipment such as wind towers and lidar can acquire localized high-precision wind speed information, but due to deployment costs and maintenance difficulties, the observation period is mostly limited to interannual scales, making it difficult to reflect long-term climate change trends and the impact of low-probability extreme weather events. Furthermore, the limited number of monitoring sites cannot flexibly support wind energy resource assessments using multiple site selection schemes.

[0005] Therefore, how to solve the problem that the existing macroscopic simulation data for wind energy resource assessment has regional bias and insufficient spatiotemporal coverage of microscopic measured data, resulting in low accuracy and time resolution of wind energy resource assessment, is an important issue that urgently needs to be addressed in the field of wind energy resource assessment. Summary of the Invention

[0006] This invention provides a wind energy resource assessment method and apparatus based on flow matching model, which overcomes the shortcomings of existing wind energy resource assessment methods, such as regional bias in macroscopic simulation data and insufficient spatiotemporal coverage of microscopic measured data, resulting in low accuracy and time resolution of wind energy resource assessment. The method uses meteorological monitoring data to correct the bias in ESM simulation data, thereby improving the accuracy and completeness of wind energy resource assessment.

[0007] On one hand, the present invention provides a wind energy resource assessment method based on a flow matching model, comprising: acquiring current Earth system model simulation data of the area to be assessed; acquiring target wind speed data of the area to be assessed based on a pre-trained time-series flow matching model and the current Earth system model simulation data; wherein the time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained by training and optimization based on samples of Earth system model simulation data and meteorological monitoring data; and determining the wind energy resource assessment result of the area to be assessed based on the target wind speed data.

[0008] Further, training and optimizing the time-series flow matching model specifically includes: constructing an inverse stochastic differential equation based on the Earth system model simulation data sample and the meteorological monitoring data sample; the inverse stochastic differential equation characterizes the meteorological data diffusion path from the Earth system model simulation data to the meteorological monitoring data; reconstructing the inverse stochastic differential equation to obtain an ordinary differential equation; the ordinary differential equation characterizes the deterministic transformation process from the Earth system model simulation data to the meteorological monitoring data; solving the time-varying velocity field based on the ordinary differential equation, and obtaining the predicted output of the Earth system model simulation data sample based on the time-series flow matching model; using the difference between the time-varying velocity field and the predicted output as the training loss, and iteratively optimizing the time-series flow matching model.

[0009] Furthermore, the step of solving the time-varying velocity field based on the ordinary differential equation includes: defining a priori conditional probability distribution and conditional velocity field for the Earth system model simulation data sample and the meteorological monitoring data sample based on the ordinary differential equation; and obtaining the time-varying velocity field by marginalization integration based on the conditional probability distribution and the conditional velocity field.

[0010] Furthermore, the conditional probability distribution is defined based on a Gaussian conditional probability path or a linear conditional probability path.

[0011] Furthermore, the pre-trained time-series flow matching model, based on the current Earth system model simulation data, obtains the target wind speed data of the area to be evaluated, including: inputting the current Earth system model simulation data into the pre-trained time-series flow matching model to obtain the predicted target velocity field; converting the target velocity field into a target probability distribution, and sampling the target wind speed data from the target probability distribution.

[0012] Furthermore, the time-series flow matching model is built based on the Patch Transformer layer.

[0013] Further, determining the wind energy resource assessment result of the area to be assessed based on the target wind speed data includes: determining the power curve model of the target wind turbine, wherein the power curve model represents the correspondence between wind speed and output power; inputting the target wind speed data into the power curve model to obtain the initial output wind power value; correcting the initial wind power value according to the environmental parameters of the target wind turbine to obtain the target wind power value, wherein the target wind power value is the wind energy resource assessment result.

[0014] Secondly, the present invention also provides a wind energy resource assessment device based on a flow matching model, comprising: an Earth system model simulation data acquisition module for acquiring current Earth system model simulation data of the area to be assessed; a target wind speed data acquisition module for acquiring target wind speed data of the area to be assessed based on a pre-trained time-series flow matching model and the current Earth system model simulation data; wherein the time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained by training and optimization based on samples of Earth system model simulation data and meteorological monitoring data; and a wind energy resource assessment result determination module for determining the wind energy resource assessment result of the area to be assessed based on the target wind speed data.

[0015] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the wind energy resource assessment method based on the flow matching model as described above.

[0016] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the wind energy resource assessment method based on the flow matching model as described above.

[0017] The wind energy resource assessment method based on a flow matching model provided by this invention obtains current Earth System Model (ESM) simulation data of the area to be assessed, and, based on a pre-trained time-series flow matching model, obtains target wind speed data for the area based on the current ESM simulation data. The time-series flow matching model models the continuous transformation path from ESM simulation data to meteorological monitoring data using ordinary differential equations. This model is obtained through training and optimization using samples of ESM simulation data and meteorological monitoring data, and then determines the wind energy resource assessment result for the area to be assessed based on the target wind speed data. This method, by combining ESM simulation data and meteorological monitoring data and applying a time-series flow matching model to fit the velocity field of the ordinary differential equations, achieves accurate wind speed assessment from interdecadal to interannual scales, significantly improving the accuracy and completeness of wind energy resource assessment while reducing computational resource requirements. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the wind energy resource assessment method based on a flow matching model provided in an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of the meteorological data diffusion process provided in an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of the structure of the time-series flow matching model provided in an embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of the structure of the wind energy resource assessment device based on the flow matching model provided in an embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] It should be noted that current wind energy resource assessments primarily rely on ESM simulation data and meteorological monitoring data. ESM is a mathematical model based on the material-energy exchange patterns of Earth's various spheres, providing historical, present, and future meteorological data through numerical calculations. Reanalysis data assimilates multi-source monitoring data from satellites, radiosondes, and other sources, and uses ESM to simulate global historical meteorological and climate change processes. Common reanalysis datasets include ERA5 and MERRA. Climate model data uses ESM to simulate various forcing factors on future climate change processes. ESM datasets provide interdecadal scale data coverage, and combined with established meteorological-electricity conversion mechanism models, wind power generation can be simulated in any ocean location. Global models have a spatial resolution ≥25 km and a temporal resolution ≥1 hour, while higher resolution simulation data requires dynamic downscaling using regional models.

[0026] On the other hand, while wind energy resource monitoring can utilize meteorological towers or lidar to collect local meteorological information, the sampling period for meteorological data during the wind farm construction and site selection phase is typically limited to an interannual scale due to equipment deployment costs and maintenance difficulties. Therefore, meteorological monitoring data struggles to capture the impact of interdecadal climate change and low-probability extreme weather events on wind farm power generation capacity and facilities. Furthermore, the limited number of monitoring points restricts the provision of flexible wind energy resource assessments. Overall, wind energy resource assessment suffers from the contradiction of biased macro-level data and incomplete micro-level data. However, the characteristics of these two types of data are complementary, and fusion modeling holds promise for achieving more accurate and comprehensive wind energy resource assessments.

[0027] In view of this, this invention proposes a wind energy resource assessment method based on a flow matching model. This method uses ordinary differential equations to model the continuous transformation path from ESM simulation data to meteorological monitoring data, and achieves high-resolution data distribution modeling by fitting the velocity field during the probability distribution transformation process. Specifically, Figure 1 The diagram shows a flowchart of the wind energy resource assessment method based on the flow matching model provided in an embodiment of the present invention.

[0028] like Figure 1 As shown, the method includes: S110, acquiring current Earth system model simulation data of the area to be evaluated; S120, acquiring target wind speed data based on a pre-trained time-series flow matching model and the current Earth system model simulation data; wherein, the time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained by training and optimization based on samples of Earth system model simulation data and meteorological monitoring data; S130, determining the wind energy resource assessment result of the area to be evaluated based on the target wind speed data.

[0029] The following will provide a detailed description of steps S110-S130 and related steps.

[0030] S110, Obtain the current Earth System Model simulation data (hereinafter referred to as the current ESM simulation data) for the region to be evaluated.

[0031] The area to be evaluated refers to the specific geographical area within which wind energy resources need to be assessed. It is the target area for site selection, planning, and feasibility analysis of wind energy development projects. The area to be evaluated is defined by a set of geographical coordinates. It can be a planned offshore wind power area, a candidate site for onshore wind farms, or the planning scope of a regional new energy base. No specific limitations are made here.

[0032] Obtain current ESM simulation data for the region to be evaluated. Specifically, acquire global or regional reanalysis datasets from publicly available or authorized climate data centers. These datasets are constructed based on Earth system models and include numerical simulation results of coupled multiple spheres, including the atmosphere, ocean, and land surface. Preferably, the reanalysis dataset can be the ERA5 reanalysis dataset released by the European Centre for Medium-Range Weather Forecasts (ECMWF) or the MERRA-2 dataset provided by NASA.

[0033] Subsequently, based on the geographic coordinates of the area to be evaluated, meteorological element data for the corresponding spatial region and time series are extracted from the reanalysis dataset, including but not limited to wind speed, wind direction, temperature, air pressure, and boundary layer height at different altitudes. This yields the current ESM simulation data for the area to be evaluated.

[0034] Based on obtaining the current ESM simulation data of the region to be evaluated in step S110, step S120 is further executed.

[0035] S120, Based on a pre-trained time-series flow matching model, target wind speed data is obtained according to the current Earth system model simulation data; wherein, the time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained by training and optimization based on Earth system model simulation data samples and meteorological monitoring data samples.

[0036] Specifically, this embodiment pre-trains a time-series flow matching model. The training process involves collecting historical meteorological monitoring data from the area to be evaluated or similar climate zones using wind towers, lidar, or buoy stations. This data serves as a high-resolution sample of real wind speed, i.e., the meteorological monitoring data sample. Simultaneously, ESM simulation data within the corresponding spatiotemporal range is extracted as model input, i.e., the ESM simulation data sample. By pairing / aligning the meteorological monitoring data sample and the ESM simulation data sample, a training sample set is constructed. The flow matching objective function / flow matching optimization objective is then used to optimize the model parameters of the time-series flow matching model, enabling the model to learn the optimal transmission path from the ESM wind field distribution to the real wind field distribution (meteorological monitoring wind field distribution).

[0037] In actual wind energy resource assessment, the current ESM simulation data of the area to be assessed is input into a trained and converged time-series flow matching model to obtain a continuous predicted velocity field sequence output by the model. This sequence not only contains a finely detailed spatial wind speed distribution, but also retains the dynamic evolution characteristics over time.

[0038] It is worth mentioning that this embodiment effectively integrates the interdecadal coverage advantage of ESM simulation data with the high-precision characteristics of meteorological monitoring data, which can significantly improve the accuracy and physical consistency of wind energy resource assessment.

[0039] After obtaining the target wind speed data of the area to be evaluated based on the pre-trained time-series flow matching model and the current Earth system model simulation data in step S120, step S130 is further executed.

[0040] S130, Based on the target wind speed data, determine the wind energy resource assessment result for the area to be assessed.

[0041] Specifically, after obtaining the target wind speed data, the target wind speed data is converted into wind power for the area to be evaluated, i.e., the wind energy resource assessment result, based on the power curve model of the wind turbine generator. The power curve model characterizes the nonlinear relationship of the output power of the wind turbine generator under different wind speed conditions, and is usually provided by the wind turbine generator manufacturer or obtained by fitting field measured data.

[0042] In this embodiment, current Earth System Model (ESM) simulation data of the region to be evaluated is acquired, and target wind speed data for the region is obtained based on a pre-trained temporal flow matching model. The temporal flow matching model models the continuous transformation path from ESM simulation data to meteorological monitoring data using ordinary differential equations. This model is trained and optimized using samples of ESM simulation data and meteorological monitoring data. The wind energy resource assessment result for the region is then determined based on the target wind speed data. This method combines ESM simulation data and meteorological monitoring data, and applies a temporal flow matching model to fit the velocity field of the ordinary differential equations. This achieves accurate wind speed assessment from decadal to interannual scales, significantly improving the accuracy and completeness of wind energy resource assessment while reducing computational resource requirements.

[0043] Based on the above embodiments, the following will further describe in detail the overall training and optimization process of the time-series flow matching model.

[0044] The training and optimization of the time-series flow matching model specifically includes: constructing an inverse stochastic differential equation based on Earth system model simulation data samples and meteorological monitoring data samples; the inverse stochastic differential equation characterizes the meteorological data diffusion path from Earth system model simulation data to meteorological monitoring data; reconstructing the inverse stochastic differential equation to obtain an ordinary differential equation; the ordinary differential equation characterizes the deterministic transformation process from Earth system model simulation data to meteorological monitoring data; solving the time-varying velocity field based on the ordinary differential equation, and obtaining the predicted output of the Earth system model simulation data samples based on the time-series flow matching model; using the difference between the time-varying velocity field and the predicted output as the training loss to iteratively optimize the time-series flow matching model.

[0045] It is easy to understand that achieving high-resolution resource data modeling requires the fusion of ESM simulation data and meteorological monitoring data. To train and optimize the time-series flow matching model, it is first necessary to construct a training sample set consisting of ESM simulation data samples and meteorological monitoring data samples.

[0046] Specifically, historical meteorological monitoring data of the area to be evaluated or similar climate areas are collected through wind measurement towers, lidar, or buoy stations as high-resolution real wind speed samples, i.e., meteorological monitoring data samples; at the same time, ESM simulation data within the corresponding spatiotemporal range are extracted as model input, i.e., ESM simulation data samples.

[0047] Subsequently, considering that wind energy resource utilization is related to the height of the wind turbine hub, the ESM simulation data sample and the meteorological monitoring data sample can be converted and aligned by a logarithmic wind speed profile model, where the logarithmic wind speed profile model is as follows (1).

[0048] (1).

[0049] In equation (1), It is the wind speed at the wheel hub height. It refers to the wind speed at reference altitude. It is the height of the wind turbine hub. It is a reference height. It refers to surface roughness.

[0050] The ESM simulation data sample and the meteorological monitoring data sample can be aligned to the same wind turbine hub height using the above formula (1), that is: assuming a certain point The ESM simulation data sample is The meteorological monitoring data sample is ESM simulation data samples were analyzed using a logarithmic wind speed profile model. from Convert to wind turbine hub height ( ), and, by using a logarithmic wind speed profile model to sample meteorological monitoring data from Convert to wind turbine hub height ( ).

[0051] It should be noted that during the alignment process, the temporal resolution of the two types of data describing the same time period may be different, resulting in dimensional differences. This can be converted using linear interpolation to align the two types of data to the same spatiotemporal dimension.

[0052] In this way, a training sample set consisting of aligned ESM simulation data samples and meteorological monitoring data samples can be constructed. In subsequent model training, the ESM simulation data samples and meteorological monitoring data samples used by default are both aligned / paired.

[0053] Furthermore, based on aligned ESM simulation data samples and meteorological monitoring data samples, an inverse stochastic differential equation is constructed. Specifically, the Itō process is used to describe the transformation between the two types of data, and the highly unified ESM simulation data samples and meteorological monitoring data samples are denoted as follows: and A stochastic differential equation (SDE) can be constructed to describe the equation from which the equation originates. to The forward process is as follows (2).

[0054] (2).

[0055] In equation (2), This is the drift coefficient, which represents the trend change, i.e., the deviation correction. The diffusion coefficient represents the degree of random fluctuation. This is the Wiener process.

[0056] SDE(t→1) in equation (1) can be regarded as the sample of meteorological monitoring data. Noise is gradually added to increase information entropy. The process of increasing information entropy is a process that can be preceded, and the inverse SDE(t→0) of equation (2) is the problem that needs to be solved in modeling high-resolution wind energy resource data, as shown in equation (3) below.

[0057] (3).

[0058] In equation (3), It is the logarithmic probability density with respect to the gradient, also known as the score function. In the known forward process equation (2) and Under the premise that both the discretized probability diffusion model and the continuous fractional diffusion model are equivalent to using a neural network to fit the model... This allows for the modeling of reverse SDE.

[0059] However, traditional diffusion models rely on a stepwise Gaussian diffusion process, where the endpoint of the forward noise addition process is noise. rather than having lower information entropy Therefore, in the high-resolution meteorological data mapping process, unlike the complete diffusion process, low-resolution data can be regarded as an intermediate state of the diffusion process. For details, please refer to [link to relevant documentation]. Figure 2 , Figure 2 A schematic diagram of the meteorological data diffusion process provided in an embodiment of the present invention is shown.

[0060] Since diffusion models require sampling the prior distribution that converges to the forward process and reconstructing high-resolution data through the inverse process, it is clear that direct explicit modeling in SDE is not feasible. This prior noise-adding process is extremely difficult. Therefore, existing research generally uses a complete noise-adding process to... Mapping to prior noise Furthermore, conditional inverse SDEs are learned by introducing conditional values ​​as a guide, as detailed in equation (4).

[0061] (4).

[0062] In equation (4), .

[0063] according to Figure 2 It can be seen that, This can be viewed as a continuous process of increasing information entropy; therefore, the ESM simulation data distribution... Distribution of meteorological monitoring data In a higher-dimensional space, there must be a smaller optimal transmission distance. Specifically, it is shown in equation (5).

[0064] (5).

[0065] Therefore, if modeling is possible The transformation process is expected to reduce the computational complexity of the downscaling process. Here, the inverse process of SDE can be reconstructed based on the Fokker-Planck equation, and equation (2) has different diffusion coefficients. The forward processes, and the marginal distributions modeled by these forward processes are the same. When At that time, there exists an equivalent ordinary differential equation (ODE) for the inverse process of SDE, which can be found in equation (6) below.

[0066] (6).

[0067] This can be achieved by constructing this ODE. This deterministic transformation process will greatly simplify the stochastic transformation process of SDE.

[0068] Furthermore, this embodiment proposes a time-series flow matching model to achieve high-resolution wind energy resource data modeling. Flow matching is a generative theory framework based on ODE design, which models the target probability distribution through the vector field of the probability transformation path. When the initial probability distribution is... The transformation process of the probability distribution will be caused by a time-varying velocity field. The decision describes the direction and magnitude of the change in distribution. probability distribution at time step It can be obtained by solving ODE, as shown in equation (7) below.

[0069] (7).

[0070] In equation (7), Therefore, the target probability distribution An approximate value can be calculated through multi-step iterative transformation, as shown in equation (8).

[0071] (8).

[0072] In equation (8), , , This represents the maximum number of iterations.

[0073] Due to the velocity field Since these are unknowns, this embodiment introduces a conditional flow matching strategy to define a prior conditional probability distribution for each sample (ESM simulation data sample and meteorological monitoring data sample). and conditional velocity field ,thus and It can be obtained through marginal integral, as shown in equations (9)-(10) below.

[0074] (9).

[0075] (10).

[0076] Based on the conditional flow matching strategy, the prior probability distribution transformation process for each sample can be designed to obtain the actual velocity field / time-varying velocity field. .

[0077] While solving for the time-varying velocity field, the predicted output corresponding to the ESM simulation data samples is obtained based on the time-series flow matching model. Specifically, the ESM simulation data samples are input into the constructed time-series flow matching model to obtain the corresponding predicted output, which is the predicted velocity field. .

[0078] Furthermore, the time-varying velocity field With predicted output The difference is used as the training loss. The model parameters of the time-series flow matching model are iteratively optimized through a preset loss function to obtain a converged time-series flow matching model. The preset loss function is defined as follows (11).

[0079] (11).

[0080] It should be noted that the prior conditional probability distribution Defined based on Gaussian conditional probability path or linear conditional probability path.

[0081] For example, in a specific embodiment, the prior conditional probability distribution The probability is defined based on the Gaussian conditional probability path, as shown in equation (12).

[0082] (12).

[0083] In equation (12), and These are the prior expectation and variance related to time t, respectively. It must be met at the time , This ensures that all probability paths converge to a standard Gaussian distribution. It must be met at the time , Converging to the target probability distribution The actual velocity field / time-varying velocity field under the Gaussian conditional probability path is defined as follows (13).

[0084] (13).

[0085] In another specific embodiment, conditional probability distribution The conditional probability is defined based on a linear conditional probability path, as detailed in equation (14). Specifically, combined with... Figure 2 It is known that constructing an initial distribution closer to the target distribution during the diffusion process can reduce the probability distance between probability distributions. Therefore, it can be concluded that... The prior distribution of time is constructed as , When using the low-resolution sample distribution as the prior distribution, a linear conditional probability path can be introduced.

[0086] (14).

[0087] In a linear conditional probability path and The pairing is based on the existing reanalysis dataset and the measurement dataset, not random pairing, which is also the difference from the Gaussian conditional probability path. The transformation process of the linear conditional probability path is a linear interpolation transformation, and the actual velocity field / time-varying velocity field under the linear conditional probability path is defined as follows (15).

[0088] (15).

[0089] It is worth mentioning that, unlike the reverse process solution based on SDE, flow matching explicitly constructs the target vector field, avoiding the back-substitution of random noise during the sampling process, thus significantly simplifying the modeling process. Furthermore, in the process of modeling high-resolution wind energy resources, it satisfies the following requirements: to The mapping transformation reduces prior noise. Initial mapping transformation.

[0090] In this embodiment, an inverse stochastic differential equation is constructed based on Earth System Model (ESM) simulation data samples and meteorological monitoring data samples. This inverse stochastic differential equation represents the meteorological data diffusion path from ESM simulation data to meteorological monitoring data. The inverse stochastic differential equation is then reconstructed to obtain an ordinary differential equation, which represents the deterministic transformation process from ESM simulation data to meteorological monitoring data. The time-varying velocity field is then solved based on the ordinary differential equation, and the predicted output of the ESM simulation data samples is obtained based on a time-series flow matching model. The difference between the time-varying velocity field and the predicted output is used as the training loss to iteratively optimize the time-series flow matching model. Thus, based on the pre-trained time-series flow matching model and the current ESM simulation data, the target wind speed data for the area to be evaluated is obtained. This method, by combining ESM simulation data and meteorological monitoring data and applying a time-series flow matching model to fit the velocity field of the ordinary differential equation, achieves accurate wind speed assessment from decadal to interannual scales, significantly improving the accuracy and completeness of wind energy resource assessment while reducing computational resource requirements.

[0091] Based on the above embodiments, the following will further describe in detail the process of fitting the target velocity field with the time-series flow matching model and further obtaining the wind energy resource assessment results.

[0092] Based on a pre-trained time-series flow matching model, target wind speed data for the region to be evaluated is obtained according to current Earth system model simulation data. This includes: inputting current Earth system model simulation data into the pre-trained time-series flow matching model to obtain the predicted target velocity field; converting the target velocity field into a target probability distribution; and sampling the target wind speed data from the target probability distribution.

[0093] It is easy to understand that, in response to the need for high-resolution assessment of wind energy resources, this embodiment proposes a method for fitting the target velocity field. A time-series flow matching network is proposed. To effectively extract the structural features of time-series data, a time-series flow matching model is constructed based on the Patch Transformer layer. Specifically, Figure 3 A schematic diagram of the structure of the time-series flow matching model provided in an embodiment of the present invention is shown.

[0094] like Figure 3As shown, after encoding the current ESM simulation data, the temporal features are extracted through an L-layer Patch Transformer layer, and the predicted target velocity field is finally output. Specifically, to effectively capture the spatiotemporal evolution of wind field data, the temporal flow matching model designs a patch-based temporal attention module. This module divides the input sequence into Patch Tokens and adds temporal position encoding, and calculates the dynamic correlation weights between Patch Tokens through a multi-head self-attention mechanism.

[0095] During iterative sampling, set Step-by-step iterative sampling. Since the mapping process of ODE can satisfy the consistency of edge distribution, but in order to increase the diversity of high-resolution samples, this embodiment designs a gradually converging time-varying perturbation mechanism for sampling, with the time step being the reverse process t→0, as shown in the following equation (16).

[0096] (16).

[0097] In equation (16), This is a random perturbation term, which can be obtained by sampling from a normal distribution.

[0098] After predicting the target velocity field, the target velocity field is mapped and transformed into a conditional target probability distribution. Then, target wind speed data is randomly sampled from the target probability distribution, and the wind energy resource assessment results of the area to be assessed are determined based on the target wind speed data.

[0099] Specifically, the target wind turbine is determined based on a preset or user-specified wind turbine model, and the corresponding power curve model is obtained. This power curve model represents the relationship between wind speed and output power. The target wind speed data is input into the power curve model, and the corresponding wind power value, i.e., the initial wind power value, is determined by looking up a table or interpolation method.

[0100] To improve assessment accuracy, the initial wind power value is corrected based on the actual environmental parameters of the target wind turbine to obtain the target wind power value, i.e., the wind energy resource assessment result, thereby achieving the assessment of wind energy resources. The actual environmental parameters include, but are not limited to, air density, wind shear index, yaw alignment error, and turbulence intensity.

[0101] It is worth mentioning that, compared with the prior art, the wind energy resource assessment method based on flow matching provided in this embodiment of the invention has the following advantages: (1) Long-term, high-resolution wind energy resource assessment: By combining ESM simulation data and meteorological monitoring data, and applying a time-series flow matching model for deviation correction and resolution improvement, accurate wind speed and wind power assessment from the interdecadal scale to the interannual scale is achieved; this method can not only capture macro-climate change trends, but also effectively reflect the impact of local extreme weather events on the power generation capacity of wind farms; (2) Reduced computational complexity of high-resolution resource assessment: Ordinary differential equations (ODE) are used instead of stochastic differential equations (SDE) to simplify the diffusion path, and a more efficient linear conditional probability path is designed, which significantly reduces the demand for computational resources.

[0102] Corresponding to the wind energy resource assessment method based on the flow matching model described in the above embodiments, this invention also proposes a wind energy resource assessment device based on the flow matching model. Specifically, Figure 4 A schematic diagram of the wind energy resource assessment device based on the flow matching model provided in an embodiment of the present invention is shown.

[0103] like Figure 4 As shown, the device includes: an Earth system model simulation data acquisition module 410, used to acquire current Earth system model simulation data of the area to be evaluated; a target wind speed data acquisition module 420, used to acquire target wind speed data of the area to be evaluated based on a pre-trained time-series flow matching model and the current Earth system model simulation data; wherein, the time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained by training and optimization based on samples of Earth system model simulation data and meteorological monitoring data; and a wind energy resource assessment result determination module 430, used to determine the wind energy resource assessment result of the area to be evaluated based on the target wind speed data.

[0104] In this embodiment, the Earth System Model (ESM) simulation data acquisition module 410 acquires the current ESM simulation data of the area to be evaluated, and the target wind speed data acquisition module 420, based on a pre-trained time-series flow matching model, acquires the target wind speed data of the area to be evaluated according to the current ESM simulation data. The time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from ESM simulation data to meteorological monitoring data, and is obtained through training and optimization using samples of ESM simulation data and meteorological monitoring data. Finally, the wind energy resource assessment result determination module 430 determines the wind energy resource assessment result of the area to be evaluated based on the target wind speed data. This device, by combining ESM simulation data and meteorological monitoring data and applying the time-series flow matching model to fit the velocity field of the ordinary differential equations, achieves accurate wind speed assessment from decadal to interannual scales, significantly improving the accuracy and completeness of wind energy resource assessment while reducing computational resource requirements.

[0105] It should be noted that the wind energy resource assessment device based on the flow matching model provided in this embodiment of the invention can be referred to in correspondence with the wind energy resource assessment method based on the flow matching model described in the above embodiments, and will not be repeated here.

[0106] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 510, a communications interface 520, a memory 830, and a communication bus 540, wherein the processor 510, communications interface 520, and memory 830 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a wind energy resource assessment method based on a flow matching model. This method includes: acquiring current Earth system model simulation data of the area to be assessed; acquiring target wind speed data of the area to be assessed based on a pre-trained time-series flow matching model and the current Earth system model simulation data; wherein the time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained through training and optimization based on samples of Earth system model simulation data and meteorological monitoring data; and determining the wind energy resource assessment result of the area to be assessed based on the target wind speed data.

[0107] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0108] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the wind energy resource assessment method based on the flow matching model provided by the above methods. This method includes: acquiring current Earth system model simulation data of the area to be assessed; acquiring target wind speed data of the area to be assessed based on a pre-trained time-series flow matching model and the current Earth system model simulation data; wherein the time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained through training and optimization based on samples of Earth system model simulation data and meteorological monitoring data; and determining the wind energy resource assessment result of the area to be assessed based on the target wind speed data.

[0109] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A wind energy resource assessment method based on a flow matching model, characterized in that, include: Obtain current Earth system model simulation data for the region to be evaluated; Based on a pre-trained time-series flow matching model, target wind speed data for the region to be evaluated is obtained according to the current Earth system model simulation data. The time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained by training and optimization based on Earth system model simulation data samples and meteorological monitoring data samples. Based on the target wind speed data, the wind energy resource assessment result for the area to be assessed is determined.

2. The wind energy resource assessment method based on the flow matching model according to claim 1, characterized in that, Training and optimizing the time-series flow matching model specifically includes: Based on the Earth system model simulation data sample and the meteorological monitoring data sample, an inverse stochastic differential equation is constructed; the inverse stochastic differential equation characterizes the meteorological data diffusion path from the Earth system model simulation data to the meteorological monitoring data. The inverse stochastic differential equation is reconstructed to obtain an ordinary differential equation; the ordinary differential equation characterizes the deterministic transformation process from Earth system model simulation data to meteorological monitoring data; The time-varying velocity field is solved based on the ordinary differential equation, and the predicted output of the Earth system model simulation data sample is obtained based on the time-series flow matching model. The difference between the time-varying velocity field and the predicted output is used as the training loss to iteratively optimize the time-series flow matching model.

3. The wind energy resource assessment method based on the flow matching model according to claim 2, characterized in that, The solution of the time-varying velocity field based on the ordinary differential equation includes: Based on the ordinary differential equation, a priori conditional probability distribution and conditional velocity field are defined for the Earth system model simulation data sample and the meteorological monitoring data sample; The time-varying velocity field is obtained by marginalization integration based on the conditional probability distribution and the conditional velocity field.

4. The wind energy resource assessment method based on the flow matching model according to claim 3, characterized in that, The conditional probability distribution is defined based on a Gaussian conditional probability path or a linear conditional probability path.

5. The wind energy resource assessment method based on the flow matching model according to claim 1, characterized in that, The pre-trained time-series flow matching model obtains target wind speed data for the region to be evaluated based on the current Earth system model simulation data, including: The current Earth system model simulation data is input into a pre-trained time-series flow matching model to obtain the predicted target velocity field; The target velocity field is converted into a target probability distribution, and the target wind speed data is obtained by sampling from the target probability distribution.

6. The wind energy resource assessment method based on the flow matching model according to claim 5, characterized in that, The time-series stream matching model is built on the Patch Transformer layer.

7. The wind energy resource assessment method based on a flow matching model according to any one of claims 1-6, characterized in that, The step of determining the wind energy resource assessment result of the area to be assessed based on the target wind speed data includes: Determine the power curve model of the target wind turbine, wherein the power curve model represents the correspondence between wind speed and output power; The target wind speed data is input into the power curve model to obtain the output initial wind power value; The initial wind power value is corrected based on the environmental parameters of the target wind turbine to obtain the target wind power value, which is the wind energy resource assessment result.

8. A wind energy resource assessment device based on a flow matching model, characterized in that, include: The Earth system model simulation data acquisition module is used to acquire current Earth system model simulation data for the region to be evaluated. The target wind speed data acquisition module is used to acquire target wind speed data of the area to be evaluated based on the current Earth system model simulation data, using a pre-trained time-series flow matching model. The time-series flow matching model is based on ordinary differential equations to model the continuous transformation path from Earth system model simulation data to meteorological monitoring data, and is obtained by training and optimization based on samples of Earth system model simulation data and meteorological monitoring data. The wind energy resource assessment result determination module is used to determine the wind energy resource assessment result of the area to be assessed based on the target wind speed data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the wind energy resource assessment method based on the flow matching model as described in any one of claims 1 to 6.

10. A non-transitory 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 energy resource assessment method based on the flow matching model as described in any one of claims 1 to 6.

Images