Digital twinning-based wind power plant wake effect coupling power prediction system

Abstract

The invention discloses a wind power plant wake effect coupling power prediction system based on digital twinning, and belongs to the technical field of electric digital data processing, and the system comprises a data collection module which is used for obtaining real-time operation data and geographic information data of a wind power plant, and generating multi-source wind power plant data; the data processing module is used for processing the multi-source wind field data to generate wind field state feature vectors; the dual-core wake flow simulation module is used for generating a fine-tuned low-fidelity wake flow model based on the wind field state feature vector; the wake flow influence parameter generation module is used for performing wake flow simulation on the wind field state feature vector by using the fine-tuned low-fidelity wake flow model to generate a dynamic wake flow influence parameter; and the power prediction module is used for inputting the dynamic wake flow influence parameters into the digital twinborn model to generate a wind power plant power prediction value. According to the invention, a high-low dual-underwriting true wake flow model and a digital twinning technology are adopted, and real-time data driving is combined, so that dynamic simulation of wake flow interaction between fans can be realized.

Description

Technical Field

[0001] This invention relates to the field of electrical digital data processing technology, and in particular to a wind farm wake effect coupled power prediction system based on digital twins. Background Technology

[0002] Wind power, as a clean and renewable energy source, plays a crucial role in the global energy structure transformation. Wind farms typically consist of dozens or even hundreds of wind turbines, and their total output power is characterized by fluctuations and intermittency. To ensure the safe and stable operation of the power grid and optimize electricity market transactions, accurate prediction of the future output power of wind farms using computer data processing systems is essential. Among these, the wake effect is a key physical phenomenon affecting the actual power generation of wind farms. It refers to the low-speed, high-turbulence region formed downstream after upstream wind turbines absorb wind energy, negatively impacting the power generation performance of subsequent turbines.

[0003] In related technologies, Chinese invention patent CN120145904B discloses a method for predicting the wake of a wind farm in a scenario with terrain changes, including: step a, determining the evaluation time period based on the actual engineering situation, determining the site terrain range, turbine layout, and basic information of the turbine units; step b, performing large eddy simulation based on the modeling information of the selected area and the given boundary conditions to establish a high-precision CFD background wind field with terrain effects; step c, combining the background wind field calculated by CFD simulation, interpolating and reconstructing the flow field, and using the engineering wake model to predict the wake effect of the wind farm and the total power generation of the wind farm turbine units.

[0004] However, the aforementioned existing technical solutions have the following technical drawbacks: High computational resource consumption, making real-time dynamic prediction difficult: Existing technologies rely on high-precision large eddy simulations of background wind fields using computational fluid dynamics (CFD). Such simulations are computationally complex and time-consuming, failing to meet the real-time prediction requirements for power fluctuations during wind farm operation. Especially in scenarios with frequent changes in wind direction and speed, frequent initiation of high-fidelity CFD simulations imposes a significant computational burden, making dynamic tracking and online optimization difficult in practical engineering. Poor model adaptability and lack of adaptive mechanisms: Existing technologies use a fixed engineering wake model combined with the CFD background field, without introducing online model calibration or adaptive update mechanisms. When wind field conditions such as wind speed, wind direction, and atmospheric stability change, the model cannot actively adjust parameters to match the actual physical processes, leading to decreased prediction accuracy. Neglecting the spatiotemporal dynamic evolution characteristics of the wake: Existing solutions focus on static or quasi-static wake simulation, failing to fully consider the inertial effects and historical evolution patterns of the wake over time. The lack of fusion processing between historical wake sequences and real-time simulation results makes the prediction results prone to drastic fluctuations due to sudden changes in wind conditions or model noise, affecting the stability of power output. Insufficient interpretability and granularity: Existing methods directly predict total power generation through engineering wake models, failing to calculate the corrected wind speed and individual power loss under the influence of wake on a turbine-by-turbine basis, and lacking a fine-grained characterization of the energy transfer mechanism within the wind farm. This limits its ability to support advanced applications such as wind farm layout optimization and fault diagnosis. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a wind farm wake effect coupled power prediction system based on digital twins. It employs a high-low dual-core fidelity wake model and digital twin technology, combined with real-time data-driven operation, to achieve dynamic simulation of wake interactions between wind turbines.

[0006] The above objectives can be achieved through the following approach:

[0007] A wind farm wake effect coupled power prediction system based on digital twin includes a data acquisition module for acquiring real-time operational data and geographic information data of the wind farm to generate multi-source wind farm data; a data processing module for processing the multi-source wind farm data to generate wind farm state feature vectors; a dual-core wake simulation module for generating a fine-tuned low-fidelity wake model based on the wind farm state feature vectors; a wake influence parameter generation module for using the fine-tuned low-fidelity wake model to simulate the wake of the wind farm state feature vectors and generate dynamic wake influence parameters; and a power prediction module for inputting the dynamic wake influence parameters into the digital twin model to generate wind farm power prediction values.

[0008] Furthermore, the data acquisition module includes: a sensor data acquisition unit, a geographic information acquisition unit, a data preprocessing unit, and a data fusion unit; wherein, the sensor data acquisition unit is used to acquire real-time data from sensors to generate raw sensor data; the geographic information acquisition unit is used to acquire geographic information data describing the physical environment of the wind farm; the data preprocessing unit is connected to the sensor data acquisition unit and performs noise filtering and missing value processing on the raw sensor data to generate preprocessed sensor data; the data fusion unit is connected to both the geographic information acquisition unit and the data preprocessing unit and is used to fuse the preprocessed sensor data with the geographic information data to generate multi-source wind farm data.

[0009] Furthermore, the data processing module includes a feature extraction unit and a normalization processing unit; wherein, the feature extraction unit is connected to the data fusion unit and is used to perform feature extraction and dimensionality reduction processing on the multi-source wind field data to generate an initial feature set; the normalization processing unit is connected to the feature extraction unit and is used to normalize the initial feature set to generate a wind field state feature vector.

[0010] Furthermore, the dual-core wake simulation module includes: a condition monitoring unit, a high-fidelity simulation unit, and a model online fine-tuning unit; wherein, the condition monitoring unit, connected to the normalization processing unit, is used to monitor the changes in the wind field state feature vector, and when the change exceeds a preset threshold representing a change in wind field conditions, it generates a high-fidelity simulation trigger signal; the high-fidelity simulation unit, connected to the condition monitoring unit, is used to start the high-fidelity wake model according to the high-fidelity simulation trigger signal, simulate based on the wind field state feature vector, and generate high-precision wake field data; the model online fine-tuning unit, connected to the high-fidelity simulation unit, is used to use the high-precision wake field data as training data to train the low-fidelity wake model online, update the model parameters, and generate a fine-tuned low-fidelity wake model.

[0011] Furthermore, the wake influence parameter generation module includes: a basic wake parameter generation unit and a dynamic correction unit; wherein, the basic wake parameter generation unit is connected to the online model fine-tuning unit, and is used to simulate the wake of the wind field state feature vector using the fine-tuned low-fidelity wake model to generate basic wake parameters; the dynamic correction unit is connected to the basic wake parameter generation unit, and is used to perform correction calculations on the basic wake parameters to generate dynamic wake influence parameters.

[0012] Furthermore, the power prediction module includes: a historical sequence generation unit, a weighted fusion unit, and a power calculation unit; wherein, the historical sequence generation unit is used to acquire historical wake data within a time window and generate a wake historical sequence; the weighted fusion unit is connected to the dynamic correction unit and the historical sequence generation unit respectively, and is used to perform weighted fusion processing on the dynamic wake influence parameters and the wake historical sequence to generate spatiotemporal fused wake parameters; the power calculation unit is connected to the weighted fusion unit and is used to input the spatiotemporal fused wake parameters into the digital twin model to generate wind farm power prediction values.

[0013] Furthermore, the step of inputting the spatiotemporal fusion wake parameters into the digital twin model to generate wind farm power prediction values ​​includes: calculating the wind speed deficit value of each turbine based on the spatiotemporal fusion wake parameters, and generating corrected wind speed parameters by combining the obtained incoming wind speed; calculating the output power of each turbine according to the corrected wind speed parameters and the preset power curve function of each turbine to generate individual power parameters; and aggregating the individual power parameters of all turbines to generate wind farm power prediction values.

[0014] Furthermore, after generating the wind farm power prediction value, the process further includes: sending the wind farm power prediction value to the wind farm monitoring system according to a standard data exchange protocol; obtaining the grid planned power generation curve and performing comparative analysis based on the wind farm power prediction value to generate power dispatch suggestions.

[0015] Furthermore, the step of performing noise filtering and missing value processing on the original sensor data to generate preprocessed sensor data includes: performing time-window-based moving average filtering on the original sensor data to obtain denoised sensor data; and performing linear interpolation on the denoised sensor data to generate preprocessed sensor data.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] This invention constructs a dual-core wake simulation mechanism that coordinates high-fidelity and low-fidelity models and introduces an online fine-tuning strategy. When wind field conditions change, the high-fidelity model can be used to calibrate the low-fidelity model. This ensures that the power prediction results closely follow the actual physical process while avoiding the huge computational resource consumption caused by continuously running a high-precision model, thus achieving an effective balance between prediction accuracy and computational efficiency.

[0018] The prediction system of this invention possesses excellent dynamic adaptability and robustness. By setting up an operating condition monitoring unit, the system can intelligently identify substantial changes in wind field conditions and selectively trigger costly, accurate simulations and model updates. This allows the prediction model to proactively adapt to constantly changing external conditions such as wind speed, wind direction, and atmospheric stability, ensuring stable and reliable prediction performance in complex and ever-changing real-world operating environments.

[0019] This invention constructs a spatiotemporal fusion wake parameter by introducing a weighted fusion process of historical wake sequences and current dynamic wake influence parameters. This method not only considers the spatial distribution of the wake but also incorporates its evolutionary inertia in the temporal dimension, making the input parameters used for power calculation smoother and more in line with physical laws. It effectively suppresses drastic fluctuations in predicted values ​​caused by sudden changes in wind conditions or model noise, thus improving the stability of the prediction results.

[0020] The digital twin model constructed in this invention possesses high physical interpretability. By calculating the corrected wind speed of each turbine under the influence of wake vortex and combining it with its independent power curve, the individual power is predicted, and finally, the total power of the entire wind farm is obtained. This refined calculation method can accurately quantify the power loss caused by wake vortex, clearly reflecting the energy transfer and loss mechanisms within the wind farm, and providing a clear physical basis for subsequent wind farm optimization control and fault diagnosis.

[0021] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a framework diagram of a wind farm wake effect coupled power prediction system based on digital twin according to an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of a wind farm wake effect coupled power prediction system based on digital twin according to an embodiment of the present invention.

[0025] Figure 3 This is a flowchart of the dual-core wake simulation module according to an embodiment of the present invention. Detailed Implementation

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

[0027] Reference Figure 1 One embodiment of the present invention proposes a wind farm wake effect coupled power prediction system based on digital twins. It adopts a high-low dual-core fidelity wake model and digital twin technology, combined with real-time data driving, which can realize dynamic simulation of wake interaction between wind turbines.

[0028] like Figure 2 As shown, the system in this embodiment specifically includes: a data acquisition module, a data processing module, a dual-core wake simulation module, a wake influence parameter generation module, and a power prediction module; wherein,

[0029] S1. The data acquisition module is used to acquire real-time operation data and geographic information data of the wind farm and generate multi-source wind farm data.

[0030] Furthermore, the data acquisition module includes: a sensor data acquisition unit, a geographic information acquisition unit, a data preprocessing unit, and a data fusion unit; wherein,

[0031] The sensor data acquisition unit is used to acquire real-time data from the sensor and generate raw sensor data;

[0032] The geographic information acquisition unit is used to acquire geographic information data describing the physical environment of the wind farm;

[0033] The data preprocessing unit is connected to the sensor data acquisition unit and performs noise filtering and missing value processing on the raw sensor data to generate preprocessed sensor data.

[0034] The data fusion unit is connected to the geographic information acquisition unit and the data preprocessing unit respectively, and is used to fuse the preprocessed sensor data with the geographic information data to generate multi-source wind field data.

[0035] Specifically, the sensor data acquisition unit obtains real-time operating data from sensors within the wind farm, such as anemometers, wind vanes, thermometers, and generator power sensors, and generates raw sensor data. The geographic information acquisition unit is used to acquire geographic information data describing the physical environment of the wind farm, including but not limited to wind turbine coordinates, terrain information, and surface roughness.

[0036] The processing begins with noise filtering of the raw sensor data. This is achieved through a time-window-based moving average filter to obtain denoised sensor data. The moving average filter averages the values ​​of all raw sensor data within a continuous time window to obtain the denoised sensor data value at that time point. After denoising, missing value processing is performed on the denoised sensor data using linear interpolation to generate preprocessed sensor data. Linear interpolation involves performing a linear fit between known data points before and after the missing data point to estimate the missing value.

[0037] The data fusion unit is connected to both the geographic information acquisition unit and the data preprocessing unit. It is used to fuse the preprocessed sensor data with geographic information data to generate multi-source wind field data. This fusion involves associating and integrating the preprocessed real-time sensor data with geographic information data describing the physical environment of the wind field to form a comprehensive dataset that includes the real-time status of the wind field and environmental characteristics.

[0038] For example, to obtain wind speed data of a wind turbine at a certain moment in a wind farm: the sensor data acquisition unit obtains the raw wind speed readings at that moment and nearby time points from the wind speed sensor of the wind turbine, thus generating raw sensor data. The geographic information acquisition unit obtains the geographic information data of the wind turbine, such as its coordinates and the surface roughness of the area. The data preprocessing unit first performs a time-window-based moving average filtering on the raw sensor data to remove noise, obtaining denoised sensor data. The window size of the moving average filtering is set according to actual needs. Next, if there are missing values ​​in the denoised sensor data, a linear interpolation method is used to estimate and fill them, finally generating preprocessed sensor data. Finally, the data fusion unit integrates and correlates this preprocessed sensor data, i.e., the cleaned and interpolated real-time operating data, with the geographic information data of the wind turbine, i.e., environmental feature data, to jointly constitute multi-source wind farm data, which can be used for subsequent feature extraction and power prediction.

[0039] Furthermore, the step of performing noise filtering and missing value processing on the original sensor data to generate preprocessed sensor data includes:

[0040] The original sensor data is subjected to a time window-based moving average filtering process to obtain denoised sensor data.

[0041] The denoised sensor data is subjected to linear interpolation to generate preprocessed sensor data.

[0042] Specifically, moving average filtering is a time-domain smoothing technique designed to effectively filter out random high-frequency noise or transient outliers generated during sensor measurements. Its core idea is to set a continuous time window, calculate the arithmetic mean of all raw sensor data values ​​within that window, and use this average as the denoised sensor data value at the center time point of the window. This effectively smooths the data sequence.

[0043] Linear interpolation is used to compensate for missing values ​​that occur during sensor data transmission or acquisition. This method assumes a linear relationship between missing data points and their adjacent known denoised data points. By performing a linear fit between these two known points, the value of the missing point is estimated and filled in, ensuring the continuity and integrity of the data sequence.

[0044] For example, taking the preprocessing of raw wind speed data collected from a certain wind turbine as an example: The raw sensor data undergoes a time-window-based moving average filtering process: Assume the time window length is set to 5 data points. The raw wind speed reading at the current moment and the raw wind speed readings at moments before and after it are 8.0 m / s, 8.1 m / s, 15.0 m / s, 7.9 m / s, and 8.0 m / s. 15.0 m / s is assumed to be a noise outlier. The moving average filtering process involves adding these 5 raw wind speed readings together and dividing by 5, resulting in an average value of 9.4 m / s, which is the denoised sensor data at that moment. Linear interpolation is then performed on the denoised sensor data: Assume that data is missing at a certain moment in the filtered denoised sensor data sequence. The adjacent known data points before and after this missing point are 9.0 m / s and 9.2 m / s, respectively. Using the principle of linear interpolation, assuming the missing point is located in the middle, the estimated preprocessed sensor data is 9.1 m / s. Through the above steps, noise in the raw data is smoothed, missing values ​​are properly filled, and finally reliable preprocessed sensor data is generated.

[0045] S2. The data processing module is used to process the multi-source wind field data and generate a wind field state feature vector.

[0046] Furthermore, the data processing module includes: a feature extraction unit and a normalization processing unit; wherein,

[0047] The feature extraction unit is connected to the data fusion unit and is used to perform feature extraction and dimensionality reduction on the multi-source wind field data to generate an initial feature set.

[0048] The normalization processing unit is connected to the feature extraction unit and is used to normalize the initial feature set to generate a wind field state feature vector.

[0049] Specifically, the feature extraction unit performs feature extraction and dimensionality reduction on these multi-source wind field data to generate an initial feature set. Feature extraction refers to identifying, selecting, combining, and transforming key information that affects power prediction from the original multi-source wind field data. Dimensionality reduction, while preserving the main information of the data, reduces the number of features in the feature set through methods such as principal component analysis or linear discriminant analysis, thereby reducing computational complexity and improving the model's generalization ability. The initial feature set is a set of key features that has been screened and compressed.

[0050] The normalization unit normalizes the initial feature set. Normalization is a data preprocessing technique designed to map feature data of different dimensions and orders of magnitude to a unified numerical range, such as [0,1] or [-1,1]. Commonly used normalization methods include min-max normalization or... Fraction normalization. After normalization, a wind field state feature vector is generated. This vector is a unified, optimized input data that comprehensively represents the current wind field state and can be directly used in the subsequent dual-core wake simulation module.

[0051] For example, consider processing multi-source wind farm data, including real-time wind speed, wind direction, temperature, and turbine coordinates at a specific moment: The feature extraction unit receives the multi-source wind farm data. First, feature extraction is performed, selecting, for example, real-time wind speed, wind direction, ambient temperature, turbine coordinates, and surface roughness as initial features. Then, through dimensionality reduction, these initial feature sets are compressed, removing highly correlated redundant information to generate a lower-dimensional initial feature set. Next, the normalization processing unit receives this initial feature set. For each feature in the initial feature set, such as wind speed, min-max normalization is performed. First, the minimum and maximum values ​​of this feature in historical data are determined. Then, the current real-time feature value is substituted into the normalization process to calculate the normalized feature value. Similar normalization is performed on all features in the initial feature set. Finally, all normalized feature values ​​are combined in a predetermined order to form a wind farm state feature vector. This feature vector is used as a standardized input and fed into the dual-core wake simulation module for wake simulation.

[0052] S3. The dual-core wake simulation module is used to input the wind field state feature vector into a preset high-low dual-core fidelity wake model, generate high-precision wake field data through the high-precision wake model, and fine-tune the low-precision wake model online based on the high-precision wake field data to generate a fine-tuned low-precision wake model.

[0053] Furthermore, the workflow of the dual-core wake simulation module is as follows: Figure 3 As shown, the dual-core wake simulation module includes: a condition monitoring unit, a high-fidelity simulation unit, and a model online fine-tuning unit; wherein,

[0054] The operating condition monitoring unit is connected to the normalization processing unit and is used to monitor the changes in the wind field state feature vector. When the changes exceed the preset threshold representing the change in wind field operating conditions, a high-fidelity simulation trigger signal is generated.

[0055] The high-fidelity simulation unit is connected to the operating condition monitoring unit and is used to start the high-fidelity wake model according to the high-fidelity simulation trigger signal, perform simulation based on the wind field state feature vector, and generate high-precision wake field data.

[0056] The online fine-tuning unit of the model is connected to the high-fidelity simulation unit and is used to train the low-fidelity wake model online using the high-precision wake field data as training data, update the model parameters, and generate a fine-tuned low-fidelity wake model.

[0057] Specifically, the operating condition monitoring unit is connected to the normalization processing unit to receive and monitor changes in the wind field state feature vector. This unit continuously checks the difference between the current value of the wind field state feature vector and the previous value or a reference value. When the change in the wind field state feature vector exceeds a preset threshold representing a change in wind field operating conditions, the operating condition monitoring unit immediately generates a high-fidelity simulation trigger signal. This threshold is used to define whether the wind field operating state has undergone a change sufficient to affect the wake effect, such as drastic fluctuations in wind speed, wind direction, or turbulence intensity.

[0058] The high-fidelity simulation unit is connected to the operating condition monitoring unit and is used to activate a preset high-fidelity wake model upon receiving a high-fidelity simulation trigger signal. The high-fidelity wake model is typically a complex model built using high-precision computational fluid dynamics (CFD) methods, capable of accurately simulating the wake field details of wind turbines within a wind farm. The high-fidelity simulation unit performs wake simulation based on the received wind farm state feature vectors, generating high-precision wake field data.

[0059] The online model fine-tuning unit is connected to the high-fidelity simulation unit to receive high-precision wake field data. This high-precision wake field data is used as training data to train a pre-defined low-fidelity wake model online. The low-fidelity wake model is typically a simplified model with higher computational efficiency, such as an analytical or engineering model. Through online training, the online model fine-tuning unit updates the parameters of the low-fidelity wake model, calibrating its simulation results towards the accurate results of the high-fidelity wake model, thereby generating a fine-tuned low-fidelity wake model.

[0060] For example, taking a sudden change in wind direction triggering a wake simulation as an example: The operating condition monitoring unit continuously monitors the wind field state feature vector, such as the feature value representing wind direction. Assuming a preset threshold of wind direction change exceeding ten degrees, when the wind direction feature value is detected to change from one direction to another within a short period, exceeding the preset threshold, the operating condition monitoring unit generates a high-fidelity simulation trigger signal. Upon receiving the trigger signal, the high-fidelity simulation unit immediately starts a preset CFD-based high-fidelity wake model. It uses the latest wind field state feature vector as input to perform a high-precision simulation of the complex wake field of the entire wind farm, generating high-precision wake field data containing detailed information such as wind speed and turbulence intensity. Subsequently, the online model fine-tuning unit receives this high-precision wake field data. This unit uses this data as accurate "ground truth" or training samples to train or calibrate the parameters of a pre-existing low-fidelity wake model online. After the training process is complete, the parameters of the low-fidelity model are updated, generating a new fine-tuned low-fidelity wake model. Subsequently, the wake influence parameter generation module will use this finely tuned low-fidelity model, which combines the advantages of accuracy and speed, to perform routine wake simulations.

[0061] S4. The wake influence parameter generation module is used to simulate the wake of the wind field state feature vector using the fine-tuned low-fidelity wake model to generate dynamic wake influence parameters.

[0062] Furthermore, the wake influence parameter generation module includes: a basic wake parameter generation unit and a dynamic correction unit; wherein,

[0063] The basic wake parameter generation unit is connected to the online model fine-tuning unit and is used to simulate the wake of the wind field state feature vector using the fine-tuned low-fidelity wake model to generate basic wake parameters.

[0064] The dynamic correction unit is connected to the basic wake parameter generation unit and is used to perform correction calculations on the basic wake parameters to generate dynamic wake influence parameters.

[0065] Specifically, the basic wake parameter generation unit is connected to the online model fine-tuning unit in the dual-core wake simulation module, and is used to receive the fine-tuned low-fidelity wake model. This unit inputs the wind field state feature vector received from the data processing module into the fine-tuned low-fidelity wake model to perform wake simulation. Wake simulation calculates the impact of the wind turbine wake on downstream wind turbines using the low-fidelity model based on the wind field state feature vector, such as wind speed loss and increased turbulence intensity. The result of the wake simulation is the basic wake parameter. The basic wake parameter is a preliminary quantification of the wake effect by the low-fidelity model under the current operating conditions.

[0066] The dynamic correction unit is connected to the basic wake parameter generation unit and receives the basic wake parameters. This unit performs correction calculations on the basic wake parameters. The correction calculation further calibrates and optimizes the basic wake parameters to improve their accuracy and dynamic adaptability. This correction may be based on historical wake data, real-time turbulence information, or other correction algorithms to compensate for simplification errors in the low-fidelity model or the effects of rapidly changing operating conditions. The result of the correction calculation is the dynamic wake influence parameter. The dynamic wake influence parameter is a final, real-time, optimized quantitative indicator used to characterize the interactions between wind turbines within a wind farm.

[0067] For example, taking wake simulation of the wind field state feature vector at a certain moment as an example: the basic wake parameter generation unit receives the fine-tuned low-fidelity wake model generated by the dual-core wake simulation module. Simultaneously, it receives the wind field state feature vector generated by the data processing module. This unit inputs the feature vector into the fine-tuned low-fidelity wake model. Based on features such as wind speed, wind direction, and turbine location, the model calculates the wind speed deficit and turbulence increase value generated by each turbine on downstream turbines, generating basic wake parameters. Next, the dynamic correction unit receives these basic wake parameters. This unit may dynamically adjust the basic wake parameters based on real-time monitored turbulence intensity information or preset correction coefficients. For example, if the real-time turbulence intensity is higher than the value predicted by the basic model, an additional correction is applied to the wind speed deficit value to better reflect the actual situation of wake mixing and recovery. Through correction calculations, dynamic wake influence parameters are finally generated. Finally, these dynamic wake influence parameters are sent to the power prediction module as one of the key inputs for predicting the wind farm's output power.

[0068] S5. The power prediction module is used to input the dynamic wake influence parameters into a preset digital twin model to generate wind farm power prediction values.

[0069] Furthermore, the power prediction module includes: a historical sequence generation unit, a weighted fusion unit, and a power calculation unit; wherein,

[0070] The historical sequence generation unit is used to acquire historical wake data within a time window and generate a historical wake sequence.

[0071] The weighted fusion unit is connected to the dynamic correction unit and the historical sequence generation unit respectively, and is used to perform weighted fusion processing on the dynamic wake influence parameters and the wake historical sequence to generate spatiotemporal fusion wake parameters.

[0072] The power calculation unit is connected to the weighted fusion unit and is used to input the spatiotemporal fusion wake parameters into the digital twin model to generate wind farm power prediction values.

[0073] Specifically, the historical sequence generation unit is used to acquire historical wake data within a time window. This unit retrieves wake influence parameter records for a specific past period by accessing a historical database or storage system. The time window is designed to ensure that the historical data reflects the temporal evolution patterns under current operating conditions. After processing, the historical wake data generates a wake historical sequence, which contains empirical information about the wake effect over time.

[0074] The weighted fusion unit is connected to both the dynamic correction unit and the historical sequence generation unit in the wake influence parameter generation module. It receives real-time spatial information of the dynamic wake influence parameters and temporal empirical information of the wake historical sequence. This unit performs weighted fusion processing on these two sets of parameters to generate spatiotemporally fused wake parameters. The core of the weighted fusion processing lies in balancing the accuracy of real-time simulation and the stability of historical data by assigning different weights to the dynamic parameters and the historical sequence. The calculation formula for weighted fusion is as follows:

[0075] ,

[0076] in, These are the generated spatiotemporal fusion wake parameters, which are a comprehensive wake influence quantity, and the units are consistent with the input quantity. It is a dynamic wake influence parameter, representing the current real-time wake influence amount, which is obtained through the wake influence parameter generation module. It is a historical wake sequence, representing the average or a representative value of the historical wake influence within a time window. It is the weight of the dynamic wake influence parameter, a dimensionless weight coefficient obtained through model training or expert experience. It represents the weights of the wake history sequence, a dimensionless weight coefficient, also obtained through model training or expert experience, and satisfies... The constraints ensure the rationality of the dimensions and numerical ranges.

[0077] The power calculation unit receives the generated spatiotemporal fusion wake parameters. It uses these parameters as one of the core inputs to a pre-defined digital twin model. This digital twin model is a precisely calibrated virtual model that reflects the complex physics and operational status of the wind farm. Based on the spatiotemporal fusion wake parameters and other necessary environmental and operational inputs, the model calculates and outputs predicted wind farm power values.

[0078] For example, taking the spatiotemporal fusion of wake influence parameters of all wind turbines in a wind farm at a certain moment as an example: the historical sequence generation unit first extracts historical data similar to the current operating conditions from the historical database over a past period, and calculates the historical average value of the wake influence parameters within the time window. Assume the historical wake influence parameter average value at this moment... This refers to the average power loss percentage. The weighted fusion unit receives this historical wake influence parameter. The current dynamic wake influence parameters output by the wake influence parameter generation module. Assuming the weights for real-time dynamic information are determined through model training. Weighting of historical information The weighted fusion unit performs weighted fusion processing to calculate the spatiotemporal fusion wake parameters: Finally, the power calculation unit receives the spatiotemporal fusion wake parameters. The power calculation unit inputs this parameter into the digital twin model. The digital twin model, combined with other inputs such as incoming wind speed and ambient temperature, uses internal physics and data-driven algorithms to calculate the individual power of all wind turbines in the wind farm, and aggregates them to generate the final predicted power value of the wind farm.

[0079] Furthermore, the step of inputting the spatiotemporal fusion wake parameters into the digital twin model to generate wind farm power prediction values ​​includes:

[0080] The wind speed loss value of each turbine is calculated based on the spatiotemporal fusion wake parameters, and the corrected wind speed parameters are generated by combining the obtained incoming wind speed.

[0081] Based on the corrected wind speed parameters and the preset power curve function for each turbine, the output power of each turbine is calculated, and individual power parameters are generated.

[0082] By aggregating the individual power parameters of all turbines, a wind farm power prediction value is generated.

[0083] Specifically, the wind speed deficit for each turbine is calculated based on the spatiotemporally fused wake parameters, and corrected wind speed parameters are generated by combining these parameters with the acquired incoming wind speed. First, the power calculation unit obtains the spatiotemporally fused wake parameters output by the weighted fusion unit; these parameters characterize the spatiotemporal coupling effect of wake influence in the wind farm. Based on these parameters, a corrected wind speed parameter is generated for each turbine in the wind farm. Calculate the total wind speed loss it suffers. This deficit is the result of the combined effect of the wakes from all upstream turbines. Next, the power calculation unit obtains the incoming wind speed from the wind field sensors at the current moment. Then, calculate the wind speed loss value. Wind speed Subtract from the middle to get the first. Corrected wind speed parameters of the turbine The corrected wind speed parameter represents the effective wind speed actually experienced by the turbine blades after the wake effect. The calculation formula is as follows:

[0084] ,

[0085] It is the first Corrected wind speed parameters for the turbine; It is the first The incoming wind speed at the turbine location is obtained through the data acquisition and processing module. It is the first The total wind speed deficit of the turbines is calculated based on spatiotemporal fusion wake parameters. The output power of each turbine is calculated based on the corrected wind speed parameters and the power curve function of each turbine, generating individual power parameters. The power calculation unit then uses the corrected wind speed parameters obtained in the first step. As input, the preset or calibrated power curve function of the turbine is substituted into it. The power curve function is a nonlinear function describing the relationship between the effective wind speed of the turbine and its output power. The calculated result is the [number of curves]. Individual power parameters of the turbine This represents the real-time output power of the turbine considering the wake effect. The calculation formula is as follows:

[0086] ,

[0087] It is the first Individual power parameters of the turbine; It is the first The power curve function of the turbine is a preset or calibrated functional relationship. It is the first Corrected wind speed parameters for the turbine.

[0088] The individual power parameters of all turbines are aggregated to generate a predicted power value for the wind farm. The power calculation unit combines the power parameters of all turbines in the wind farm. Individual power parameters calculated from a turbine By summing and aggregating, we can obtain the predicted total output power of the wind farm under the current operating conditions. The calculation formula is as follows:

[0089] ,

[0090] This is the predicted power output of the wind farm. It is the first Individual power parameters of the turbine. It represents the total number of turbines in the wind farm.

[0091] For example, in a containing and Taking a wind farm with two turbines as an example, among which... lie in Upstream: Step 1: Calculate the wind speed deficit and generate corrected wind speed parameters. Current inflow wind speed. Both are 10.0 m / s. Without upstream wake influence, total wind speed loss value =0 m / s. Calculated based on spatiotemporal fusion wake parameters. by Wake effect, total wind speed loss =3.0m / s. Corrected wind speed parameters The calculation is as follows: , Corrected wind speed parameters The calculation is as follows: , It is the first Corrected wind speed parameters for the turbine. It is the first The incoming wind speed of the turbine. It is the first The total wind speed loss of the turbine. It is the first Corrected wind speed parameters for the turbine. It is the first The incoming wind speed of the turbine. It is the first The total wind speed loss of each turbine. Calculate the output power of each turbine to generate individual power parameters. Assume... and The power curve function is the same. (Within the range of 3m / s≤V≤12m / s). Individual power parameters The calculation is as follows: , Individual power parameters The calculation is as follows: , It is the first Individual power parameters of the turbine. It is the first Individual power parameters of each turbine. Aggregate the individual power parameters of all turbines to generate a wind farm power forecast. Wind farm power forecast. The calculation is as follows: .

[0092] Furthermore, after generating the predicted wind farm power value, the process also includes:

[0093] The predicted power value of the wind farm is sent to the wind farm monitoring system according to the standard data exchange protocol;

[0094] The power grid's planned power generation curve is obtained and compared with the wind farm's power forecast value to generate power dispatch suggestions.

[0095] Specifically, the wind farm power prediction values ​​generated by the power prediction module The data is encapsulated and transmitted according to a preset standard data exchange protocol via a data communication interface. The purpose of this transmission is to integrate the predicted values ​​into the wind farm monitoring system in real time. During this process... This is a numerical value representing the total output power of a wind farm. Standard data exchange protocols ensure consistency in data format, transmission mechanisms, and data reliability between the forecasting and monitoring systems.

[0096] It utilizes predicted values ​​for decision support. The system obtains the planned power generation curves from the power grid dispatch center. This curve represents the power output target value required by the power grid from the wind farm at a given time. Then, the system predicts the wind farm's power output. and grid planned power generation curve The core of the comparative analysis is to calculate the power deviation between the two. :

[0097] ,

[0098] It is the power deviation, which represents the difference between the predicted value and the planned value. It is the predicted power value of the wind farm, generated by the power calculation unit. This is the power value of the planned power generation curve of the power grid, obtained by the power grid dispatching system. Based on this power deviation... The system uses preset control strategies or decision models to generate power scheduling suggestions. If If the value is greater than zero, the suggestion might be to "reduce power" or "adjust the pitch angle of some wind turbines"; if If the value is less than zero, the suggestion might be to "notify the dispatch center of insufficient power generation" or "check the operating status of the wind turbines," etc.

[0099] For example, consider the predicted power of a wind farm and the grid plan at a certain moment: Assume the predicted power of the wind farm... =105 This value will be encapsulated into a measurement data unit, including a timestamp and data quality markers, and sent to the real-time database of the wind farm monitoring system via a network interface. The system then obtains the current grid planned power generation curve. =100 Calculate the power deviation. ,because =5 And since it is a positive value, it indicates that the predicted power generation exceeds the grid's planned requirements. The system determines 5 based on the preset control logic. The deviation exceeded the allowable error range. Therefore, the system generates power scheduling recommendations, such as: "It is recommended to schedule power for the specified..." A total of 5 typhoon machines were implemented. The power reduction can be achieved by adjusting the pitch angle or yaw angle. This recommendation was then sent to the wind farm operators or automatic control systems to guide the wind farm in adjusting its operation to bring its output power closer to 100. The planned value.

[0100] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0101] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A wind farm wake effect coupled power prediction system based on digital twin, characterized in that, The system includes: a data acquisition module, a data processing module, a dual-core wake simulation module, a wake influence parameter generation module, and a power prediction module; wherein... The data acquisition module is used to acquire real-time operation data and geographic information data of the wind farm and generate multi-source wind farm data. The data processing module is used to process the multi-source wind field data and generate a wind field state feature vector. The dual-core wake simulation module is used to input the wind field state feature vector into a preset high-low dual-core fidelity wake model, generate high-precision wake field data through the high-precision wake field model, and perform online fine-tuning of the low-precision wake model based on the high-precision wake field data to generate a fine-tuned low-precision wake model. The wake influence parameter generation module is used to simulate the wake of the wind field state feature vector using the fine-tuned low-fidelity wake model, and generate dynamic wake influence parameters. The power prediction module is used to input the dynamic wake influence parameters into a preset digital twin model to generate wind farm power prediction values.

2. The wind farm wake effect coupled power prediction system based on digital twin as described in claim 1, characterized in that, The data acquisition module includes: a sensor data acquisition unit, a geographic information acquisition unit, a data preprocessing unit, and a data fusion unit; wherein, The sensor data acquisition unit is used to acquire real-time data from the sensor and generate raw sensor data; The geographic information acquisition unit is used to acquire geographic information data describing the physical environment of the wind farm; The data preprocessing unit is connected to the sensor data acquisition unit and performs noise filtering and missing value processing on the raw sensor data to generate preprocessed sensor data. The data fusion unit is connected to the geographic information acquisition unit and the data preprocessing unit respectively, and is used to fuse the preprocessed sensor data with the geographic information data to generate multi-source wind field data.

3. The wind farm wake effect coupled power prediction system based on digital twin according to claim 2, characterized in that, The data processing module includes: a feature extraction unit and a normalization processing unit; wherein... The feature extraction unit is connected to the data fusion unit and is used to perform feature extraction and dimensionality reduction on the multi-source wind field data to generate an initial feature set. The normalization processing unit is connected to the feature extraction unit and is used to normalize the initial feature set to generate a wind field state feature vector.

4. The wind farm wake effect coupled power prediction system based on digital twin according to claim 3, characterized in that, The dual-core wake simulation module includes: a condition monitoring unit, a high-fidelity simulation unit, and a model online fine-tuning unit; wherein... The operating condition monitoring unit is connected to the normalization processing unit and is used to monitor the changes in the wind field state feature vector. When the changes exceed the preset threshold representing the change in wind field operating conditions, a high-fidelity simulation trigger signal is generated. The high-fidelity simulation unit is connected to the operating condition monitoring unit and is used to start the high-fidelity wake model according to the high-fidelity simulation trigger signal, perform simulation based on the wind field state feature vector, and generate high-precision wake field data. The online fine-tuning unit of the model is connected to the high-fidelity simulation unit and is used to train the low-fidelity wake model online using the high-precision wake field data as training data, update the model parameters, and generate a fine-tuned low-fidelity wake model.

5. A wind farm wake effect coupled power prediction system based on digital twin according to claim 4, characterized in that, The wake influence parameter generation module includes: a basic wake parameter generation unit and a dynamic correction unit; wherein... The basic wake parameter generation unit is connected to the online model fine-tuning unit and is used to simulate the wake of the wind field state feature vector using the fine-tuned low-fidelity wake model to generate basic wake parameters. The dynamic correction unit is connected to the basic wake parameter generation unit and is used to perform correction calculations on the basic wake parameters to generate dynamic wake influence parameters.

6. The wind farm wake effect coupled power prediction system based on digital twin according to claim 5, characterized in that, The power prediction module includes: a historical sequence generation unit, a weighted fusion unit, and a power calculation unit; wherein... The historical sequence generation unit is used to acquire historical wake data within a time window and generate a historical wake sequence. The weighted fusion unit is connected to the dynamic correction unit and the historical sequence generation unit respectively, and is used to perform weighted fusion processing on the dynamic wake influence parameters and the wake historical sequence to generate spatiotemporal fusion wake parameters. The power calculation unit is connected to the weighted fusion unit and is used to input the spatiotemporal fusion wake parameters into the digital twin model to generate wind farm power prediction values.

7. A wind farm wake effect coupled power prediction system based on digital twin according to claim 6, characterized in that, The step of inputting the spatiotemporal fusion wake parameters into the digital twin model to generate wind farm power prediction values ​​includes: The wind speed loss value of each turbine is calculated based on the spatiotemporal fusion wake parameters, and the corrected wind speed parameters are generated by combining the obtained incoming wind speed. Based on the corrected wind speed parameters and the preset power curve function for each turbine, the output power of each turbine is calculated, and individual power parameters are generated. By aggregating the individual power parameters of all turbines, a wind farm power prediction value is generated.

8. A wind farm wake effect coupled power prediction system based on digital twin according to claim 7, characterized in that, After generating the predicted wind farm power values, the following is also included: The predicted power value of the wind farm is sent to the wind farm monitoring system according to the standard data exchange protocol; The planned power generation curve of the power grid is obtained and compared with the predicted power value of the wind farm to generate power dispatch suggestions.

9. A wind farm wake effect coupled power prediction system based on digital twin according to claim 2, characterized in that, The step of performing noise filtering and missing value processing on the original sensor data to generate preprocessed sensor data includes: The original sensor data is subjected to a time window-based moving average filtering process to obtain denoised sensor data. The denoised sensor data is subjected to linear interpolation to generate preprocessed sensor data.

Images