Wind field-unit two-stage power characteristic prediction method under cold-wave weather

Abstract

The invention discloses a wind field-unit two-stage power characteristic prediction method for a cold wave meteorological condition. The method comprises the following steps: establishing a cold-wave weather identification model to realize accurate identification of cold-wave weather; a wind speed-power characteristic curve with corrected environmental parameters is constructed, and a CFD simulation model of each fan set is established; then, a unit power prediction model is established in combination with flow field characteristics and meteorological data, and the accuracy and robustness of the model are improved by adopting an optimization algorithm; and constructing an overall power prediction model of the wind field through a machine learning algorithm LightGBM, and obtaining predicted power of the overall wind field and the unit. According to the method, the precision and reliability of wind power prediction can be remarkably improved, and the operation risk of a wind power plant and unstable factors of a power system caused by prediction errors are reduced. The method has the functions of accurately predicting and evaluating the performance of the wind turbine generator under different cold weather conditions, and providing powerful guarantee for optimized operation of a wind power plant and stable operation of a power system.

Description

Technical Field

[0001] The present invention discloses a method for predicting the two - stage power characteristics of a wind field - turbine under cold wave weather conditions, and particularly relates to the technology of predicting the two - stage power of a wind field under cold wave weather in the field of wind power, which can effectively ensure the stable power generation of wind power under cold wave weather and maximize the reliability and economy of the power system. Background Art

[0002] With the continuous growth of the global demand for renewable energy, wind power has become one of the important clean energies. However, the power generation characteristics of wind power have strong uncertainties. Especially under cold wave weather conditions, the changes in meteorological parameters such as wind speed, temperature, and humidity will have a profound impact on the power output of wind turbines. These changes may not only affect the operating state of the wind turbines but also lead to large fluctuations in power output, thus posing a severe challenge to the stable operation of the power grid. How to accurately predict wind power has become a key factor in ensuring the efficient operation of wind farms.

[0003] Under cold wave weather conditions, meteorological parameters such as wind speed and temperature will change violently. The fluctuations of wind speed, the sharp drop in temperature, and the deposition of ice and snow will directly affect the working efficiency and power output of wind turbines. For example, extreme low temperature may cause changes in the viscosity of the lubricating oil of wind turbines, thereby affecting the normal operation of the wind turbines; strong winds may cause additional mechanical loads on the turbines, affecting the power generation performance of the wind turbines. These factors increase the instability of power generation in wind farms under cold wave weather, so the accurate prediction of wind power becomes particularly important for the reasonable dispatching and load distribution of wind farms.

[0004] Existing prediction methods have some limitations under cold wave weather conditions. First, due to the rapid changes in meteorological conditions under cold wave weather, the prediction methods that only rely on the conventional wind speed - power characteristic curve often cannot effectively cope with the rapid fluctuations of wind speed and temperature, resulting in insufficient prediction accuracy. Second, these methods are difficult to capture the non - linear and complex effects of cold wave weather on the power output of wind turbines. Especially under the combined action of factors such as the viscosity of the lubricating oil of wind turbines and the layout of wind turbines, traditional models cannot adjust the prediction results in a timely manner. In addition, existing wind power prediction methods usually have poor adaptability to abnormal meteorological patterns, resulting in large errors in the power prediction of wind farms under extreme weather conditions such as cold waves. By adopting a new method for predicting the two - stage power characteristics of a wind field - turbine under cold wave weather, the problems of insufficient accuracy, poor adaptability, and poor interpretability in traditional methods can be effectively solved, and more stable and accurate prediction results can be provided. Summary of the Invention

[0005] The object of the present invention is to provide a method for predicting the two - level power characteristics of the wind field - unit under cold wave weather. By combining, including but not limited to, meteorological data, wind turbine operation status information, and wind turbine lubricating oil status information, and applying the advanced machine learning LightGBM and ensemble learning Stacking, a two - level power characteristic model of the wind field - unit is established, which can provide more accurate and stable wind power prediction under cold wave weather. Compared with the prior art, the present invention can effectively cope with the meteorological changes under cold wave weather, improve the accuracy and robustness of wind power prediction, and provide strong technical support for the efficient operation of wind farms and the stable operation of power grids.

[0006] In view of the technical problems existing above, the present invention provides the following technical solution: A method for predicting the two - level power characteristics of the wind field - unit under cold wave weather. This method conducts power characteristic analysis at the wind farm level and the unit level respectively. By first predicting the power of a single wind turbine and then deriving the power of the entire wind field, a multi - level prediction structure from micro to macro is formed. The specific steps are as follows:

[0007] S1 Input the obtained meteorological data into a pre - trained cold wave weather identification model to obtain the prediction results of cold wave weather and the corresponding meteorological conditions. If the prediction result is cold wave weather, proceed to the subsequent steps; S1 Based on past meteorological data and wind turbine power data, considering different characteristics of cold wave weather and the time - dynamic characteristics of meteorological changes, the mean - shift clustering algorithm is used to achieve a fine division of working conditions for cold wave weather, and correspondingly construct a data set of cold wave weather, which contains meteorological characteristics and their corresponding time - dynamic change characteristics. Combining the clustering results and cold wave weather data, a machine learning - based cold wave weather identification model is established to achieve accurate identification of cold wave weather.

[0008] Further elaborate on S1. S1 includes S11: Collect historical meteorological data and perform normalization or standardization cleaning on the collected data; S12: Extract cold wave characteristics from the meteorological data; Mark the time period when the cold wave occurs and its corresponding meteorological characteristics based on the historical meteorological data, classify the cold wave meteorology in each time period according to the cold wave characteristics, and create cold wave labels; S13: After standardizing the selected cold wave characteristics in the meteorological data, use the Mean Shift clustering algorithm to divide different meteorological conditions under cold wave meteorology; S14: Combine the annotation of the cold wave event with the meteorological data to generate a cold wave meteorological dataset containing cold wave labels, and divide the dataset into a training set, a validation set, and a test set for the training and evaluation of a pre-constructed cold wave meteorological identification model; The cold wave meteorological identification model takes meteorological data as input and outputs cold wave meteorology and its corresponding meteorological conditions; S15: Use the cold wave meteorological dataset to train the cold wave meteorological identification model, adjust the hyperparameters to improve the accuracy of the model, use the validation set to evaluate the model effect, and obtain the cold wave meteorological identification model to achieve the identification of cold wave meteorology. S1 collects past meteorological data including historical wind speed, temperature, humidity, etc. and wind turbine power output data, and performs normalization or standardization cleaning on the data; Extract temperature, humidity, wind speed, air pressure, and other cold wave meteorological characteristic data, mark the time period when the cold wave occurs and its corresponding meteorological characteristics based on the past meteorological data, classify according to cold wave characteristics such as extreme low temperature, high wind speed, etc., and create cold wave labels; Standardize the selected cold wave-related characteristic data in the meteorological data, and then use the Mean Shift clustering algorithm to divide different conditions of cold wave meteorology; Combine the annotation of the cold wave event with the meteorological data to generate a dataset containing cold wave labels, and divide the dataset into a training set, a validation set, and a test set for subsequent model training and evaluation; According to the cold wave dataset, select a suitable machine learning model, such as DNN, RF, use the cold wave meteorological dataset to train the model, adjust the hyperparameters to improve the accuracy of the model, use the validation set to evaluate the model effect, and obtain the cold wave meteorological identification model to achieve the identification of cold wave meteorology.

[0009] In S2, since temperature and humidity can comprehensively affect air density, and the change of air density will directly affect the power curve of the fan. The relationship between the fan power and air density is as follows: Where P is the fan output power, C P is the power coefficient, which is an inherent characteristic of the wind turbine generator set and is closely related to the design and manufacture of the unit; S is the swept area of the wind turbine rotor; ρ is the air density, and V is the wind speed. The relationship between air density ρ and temperature t is as follows: ρ = 1.2761 + 0.0036t(p - 0.378p w) / 1000 Therefore, based on numerical weather prediction, obtain the wind field temperature, wind speed, humidity and related parameters. Combine with the actual working state of the wind turbine to construct the wind speed-power characteristic curve with environmental parameter correction. At the same time, use the oil detection system to obtain the state information of the lubricating oil of the wind turbine in real time to assist in the construction of the subsequent CFD model.

[0010] S3 Based on the wind speed-power characteristic curve corrected by environmental parameters, combine the layout of the wind turbine, such as the position coordinates, spacing, number arrangement of the wind turbines in the wind farm, hub height, etc., the lubricating oil state information, and the real-time operating parameters of the wind turbine, such as cut-in and cut-out wind speeds, rated wind speed, number and diameter of blades, and environmental factors, to establish a CFD simulation model of the unit under cold wave weather. Analyze and simulate the flow field characteristics around the wind turbine under cold wave weather through the CFD simulation model. By establishing the CFD model, obtain the flow field changes during the operation of the wind turbine, and provide further data support for the establishment of the unit power prediction model.

[0011] S4 Based on the integrated learning Stacking, establish the corresponding unit power characteristic prediction model according to the flow field characteristics and numerical weather prediction under different working conditions. Use the past data, including power, wind speed, wind direction, humidity, temperature, air pressure meteorological data, and flow field characteristics including turbulence intensity, wind speed distribution, and wake effect, as input features, and use the SFOA and / or HO optimization algorithms to optimize the model parameters to improve accuracy and robustness.

[0012] S5 Based on the power prediction characteristic model of the unit, conduct a preliminary power prediction. Combine the historical power data, wind speed, and humidity-related data, and use the Transformer model or MLP model to correct the power prediction error, and finally obtain the predicted power of the unit.

[0013] S6 Obtain the topographic and geomorphic feature data of the wind farm through GIS or satellite remote sensing technology. Based on multi-source heterogeneous data, that is, the prediction results of the unit power characteristic prediction model, the numerical weather prediction of the wind farm, the topographic and geomorphic data obtained from the wind farm planning dataset of the National Satellite Meteorological Center, GIS or satellite remote sensing technology, and the past data of the wind farm, use machine learning LightGBM to construct the overall power prediction model of the wind farm. The numerical weather prediction of the wind farm includes basic meteorological parameters, precipitation-related parameters, cloud visibility, radiation, and light and other parameters; the wind farm planning set includes geographical location information such as latitude, longitude, and altitude of the wind farm, meteorological parameters, wind energy resource assessment, wind farm planning data, etc.

[0014] S7 Use the Transformer model to adaptively adjust the model parameters during the training process, and finally obtain the overall predicted power of the wind farm.

[0015] Furthermore, S5 and S7 are used to obtain the predicted power of each wind turbine and the overall predicted power of the wind farm. The two-level power prediction method for the wind farm and wind turbines aims to improve the accuracy and stability of power prediction in the wind farm under cold snap weather, optimize the operation and management of the wind farm, and is of great significance for the operation and management of the wind farm and the stable operation of the power grid. Brief Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions implemented in the present invention, the following briefly introduces the drawings required in the embodiments. Obviously, the following drawings are only partial embodiments of the present invention, and those skilled in the art can derive other implementation solutions based on these drawings without creative labor.

[0017] Figure 1 This is the overall process architecture of a two-level power characteristic prediction method for a wind farm and wind turbines under cold snap weather according to the present invention.

[0018] Figure 2 This is the core part architecture of a two-level power characteristic prediction method for a wind farm and wind turbines under cold snap weather, that is, the architecture of a two-level power characteristic prediction model for a wind farm and wind turbines under cold snap weather.

[0019] Figure 3 It shows the process of cold snap condition division and the establishment of a cold snap weather identification model. Detailed Embodiments

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. The embodiments described in the present invention are only partial embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present invention.

[0021] As Figure 1 shown, the embodiments of the present invention propose an overall process architecture of a two-level power prediction analysis method for a wind farm and wind turbines under cold snap conditions, including the following:

[0022] Input real-time numerical weather forecast data and data related to the wind turbines and wind farm required for prediction into the two-level prediction model for the wind farm and wind turbines to perform power characteristic prediction for the wind farm and wind turbines.

[0023] As Figure 2 shown, the core of the two-level power characteristic prediction method for a wind farm and wind turbines under cold snap conditions proposed by the present invention is to establish a two-level power prediction model for the wind farm and wind turbines, and the specific content is as follows:

[0024] The division of the cold wave meteorological conditions and the establishment process of the cold wave meteorological model are shown in Figure 3 . Clean the data by normalizing or standardizing it through collecting historical meteorological data and the power output data of wind turbines; extract data such as temperature, humidity, wind speed, air pressure, and other cold wave meteorological characteristic data, mark the time period when the cold wave occurs and its corresponding cold wave characteristics according to the historical meteorological data, classify according to cold wave characteristics such as extreme low temperature and high wind speed, and create cold wave labels; for the cold wave-related characteristic data selected from the meteorological data, after standardizing, use the Mean Shift clustering algorithm to divide different meteorological conditions under cold wave meteorology; combine the annotation of cold wave events with the meteorological data to generate a data set containing cold wave labels, and divide the data set into a training set, a validation set, and a test set for subsequent model training and evaluation; according to the cold wave data set, select a suitable machine learning model, such as DNN, RF, use the cold wave meteorological data set for model training, adjust the hyperparameters to improve the accuracy of the model, and use the validation set to evaluate the model effect to obtain the cold wave meteorological identification model.

[0025] In S2, since temperature and humidity can comprehensively affect air density and the change of air density will directly affect the power curve of the fan. The relationship between the fan power and air density is as follows: where P is the output power of the fan, C P is the power coefficient which is an inherent characteristic of the wind turbine generator set and is closely related to the design and manufacture of the unit; S is the swept area of the wind wheel; ρ is the air density and V is the wind speed. The relationship between the air density ρ and the temperature t is as follows: ρ = 1.2761 + 0.0036t(p - 0.378p w ) / 1000 Therefore, combine the numerical weather forecast to obtain parameters such as wind field temperature, wind speed, and humidity, combine with the actual working state of the wind turbine generator set, construct the wind speed-power characteristic curve corrected by environmental parameters, and at the same time use the oil detection system to obtain the state information of the lubricating oil of the wind turbine generator set in real time to assist the construction of the subsequent CFD model.

[0026] Based on the wind speed-power characteristic curve corrected by environmental parameters, establish a CFD simulation model for each wind turbine group under cold wave meteorology with the wind turbine layout, lubricating oil state information, real-time operation parameters of the fan, and environmental factors, simulate and analyze the flow field characteristics around the wind turbine under cold wave weather, and obtain the flow field change during the operation of the fan by establishing a CFD model to provide further input data support for the establishment of the power prediction model.

[0027] S4 is based on the integrated learning Stacking. According to the flow field characteristics under different working conditions and numerical weather forecasts, a corresponding prediction model for the power characteristics of the unit is established. Using historical data including power, wind speed, wind direction, humidity, temperature, barometric meteorological data and flow field characteristics including turbulence intensity, wind speed distribution, and wake effect as input features, and using the SFOA and / or HO optimization algorithms to optimize the model parameters to improve accuracy and robustness.

[0028] S5 makes a preliminary power prediction based on the power prediction characteristic model of the unit. Combining historical data, wind speed, humidity, the lubricating oil state information of each component of the unit and related factors, uses the Transformer model or the MLP model to correct the power prediction error, and finally obtains the predicted power of the unit.

[0029] S6 obtains the topographic and geomorphic feature data of the wind farm through GIS or satellite remote sensing technology. Based on multi-source heterogeneous data, namely the prediction results of the power characteristic prediction model of the unit, the numerical weather forecast of the wind farm, the terrain factors obtained from the wind farm planning dataset of the National Satellite Meteorological Center, GIS or satellite remote sensing technology, and the historical data of the wind farm, uses machine learning LightGBM to construct the overall power prediction model of the wind farm. Among them, the numerical weather forecast of the wind farm includes basic meteorological parameters, precipitation-related parameters, cloud visibility, radiation and illumination and other parameters; the wind farm planning set includes geographical location information such as latitude, longitude and altitude of the wind farm, meteorological parameters, wind energy resource assessment, wind farm planning data, etc.

[0030] S7 uses the Transformer model to adaptively adjust the model parameters during the training process, and finally obtains the overall predicted power of the wind farm.

[0031] Furthermore, S5 and S7 obtain the predicted power of each wind turbine and the overall predicted power of the wind farm. The two-level power prediction method for the wind farm-unit aims to improve the accuracy and stability of power prediction in the cold wave weather of the wind farm, optimize the operation and management of the wind farm, and is of great significance for the operation and management of the wind farm and the stable operation of the power grid.

[0032] The above solutions are elaborated in detail below: In S1, the meteorological data is normalized or standardized and cleaned by collecting historical meteorological data and wind turbine power output data; temperature, humidity, wind speed, air pressure, and other cold wave meteorological characteristic data are extracted, and the time period of cold wave occurrence and its corresponding meteorological characteristics are labeled according to historical meteorological data; classification is carried out according to cold wave characteristics such as extreme low temperature and high wind speed, and cold wave labels are created; for the cold wave-related characteristic data selected from the meteorological data, after standardization processing, the Mean Shift clustering algorithm is used to divide different meteorological conditions under cold wave meteorology. First, the window radius R is selected, and a data point is selected as the window center. The weight α of each data point is calculated using the Gaussian kernel function. i , and the weight calculation formula is as follows: Among them, the meanings represented by the parameters are as follows: α i : is the weight of the i-th data point G(·): is the Gaussian kernel function x i : is the position of the i-th data point, c: is the position of the current window center, σ: is the standard deviation of the Gaussian kernel function, σ = R, ||x i - c|| 2 : is the square of the distance between the data point x i and the current window center c. After obtaining the weight value of each data point, substitute it into the position update formula Among them, the meanings represented by the parameters are as follows: c new : the new window center position, ω i : the weight of the i-th data point, x i : is the position of the i-th data point, n: is the total number of data points. The center position of the window is updated by calculating the weighted average position of all data points, and the weight is determined by the Gaussian kernel function according to the distance between the data point and the current window center. Repeatedly calculate the weight and update the position. After multiple iterations, until all data points are clustered into different clusters, establish the corresponding relationship between these clusters and different working conditions, which can be specifically divided into normal operation conditions, low-temperature operation conditions, blade icing conditions, and extreme cold freezing conditions; combine the annotation of cold wave events with meteorological data to generate a data set containing cold wave labels, and divide the data set into training set, validation set, and test set for subsequent model training and evaluation; according to the cold wave data set, select a suitable machine learning model, such as DNN, RF, use the cold wave meteorological data set for model training, adjust the hyperparameters to improve the accuracy of the model, and use the validation set to evaluate the model effect to obtain the cold wave meteorological identification model.

[0033] In S2, since temperature and humidity can comprehensively affect air density, and the change in air density will directly affect the power curve of the fan. The relationship between the fan power and air density is as follows: Where P is the output power of the fan, C P is the power coefficient, which is an inherent characteristic of the wind power generation unit and is closely related to the design and manufacture of the unit; S is the swept area of the wind turbine; ρ is the air density, and V is the wind speed. The relationship between air density ρ and temperature t is as follows: ρ = 1.2761 + 0.0036t(p - 0.378p w ) / 1000 Where p is the atmospheric pressure, and p w is the water vapor pressure. Therefore, by combining numerical weather prediction to obtain parameters such as wind field temperature, wind speed, and humidity, and combining the actual working state of the wind turbine, a wind speed-power characteristic curve corrected by environmental parameters is constructed. At the same time, the oil detection system is used to obtain the status information of the lubricating oil of the wind turbine in real time, which helps to construct the subsequent CFD model.

[0034] S3 Based on the wind speed-power characteristic curve corrected by environmental parameters, combined with the wind turbine layout, lubricating oil status information, real-time operating parameters of the fan, and environmental factors, a CFD simulation model of the unit under cold wave weather is established. A CFD simulation model is established through ANSYS Fluent or OpenFOAM to simulate and analyze the flow field characteristics around the wind turbine under cold wave weather. By establishing a CFD model, it can help us understand the flow field change law during the operation of the fan under cold wave weather, especially the phenomena such as airflow disorder, vortex, and wind speed change that may occur in a low-temperature environment. Through CFD simulation, we can also obtain detailed flow field parameters, such as wind speed distribution, airflow direction, pressure change, temperature gradient, etc. These data provide important references for the real-time operating state of the fan. At the same time, these simulation results can further support the establishment of the power prediction model. In the CFD simulation process, the K-ε or K-ω model can be selected to simulate the turbulent characteristics of the wind field. The standard K-ε equation of the turbulent model is as follows: Among them, the meanings of the parameters are as follows: k: turbulent kinetic energy, t: time, U: velocity vector, gradient operator, P k : turbulent production term, ε: turbulent dissipation rate, μ t : turbulent viscosity, σ k 、σ ε 、C ε1 、C ε2: Turbulence model constant. When setting boundary conditions, input the wind speed and temperature under cold snap meteorological conditions, consider the influence of low temperature on the performance of lubricating oil, and set the rotational speed and power output of the fan as real-time parameters. For the wind speed under cold snap meteorology, the boundary condition is: U = U wind where U wind is the wind speed. For temperature, use the temperature boundary condition: T wind = T ambient where T wind is the temperature brought by the wind speed, and T ambient is the ambient temperature. In addition, the influence of low temperature on the viscosity of lubricating oil is usually characterized by a temperature-dependent viscosity model. The relationship between temperature and viscosity can be expressed by the Arrhenius formula: where the meanings of the parameters are as follows: μ(T): viscosity at temperature T, μ0: viscosity at the reference temperature, E a : activation energy, R: gas constant, T: temperature, T0: reference temperature

[0035] S4 is based on the integrated learning Stacking. According to the flow field characteristics and numerical weather forecasts under different working conditions, a corresponding unit power characteristic prediction model is established. The past data including power, wind speed, wind direction, humidity, temperature, air pressure meteorological data and flow field characteristics including turbulence intensity, wind speed distribution, and wake effect are used as input features. Since the SFOA and HO optimization algorithms can effectively explore the hyperparameter space and improve the accuracy and robustness of the model, the SFOA and / or HO optimization algorithms are used to optimize the model parameters to improve the accuracy and robustness.

[0036] S5 Based on the power prediction characteristic model of the unit, a preliminary power prediction is carried out. Combining historical data, wind speed, humidity and related factors, a Transformer model or an MLP model is used to correct the power prediction error. Here, only taking the Transformer model as an example, based on time series data and learning the temporal pattern of the error based on the self-attention mechanism, through training, the Transformer model can capture the law of the error in power prediction, so as to adjust the original prediction result. During the training process of the Transformer model, the power prediction error in historical data is used for training to optimize the hyperparameters of the model (such as the number of attention layers, the dimension of the hidden layer, etc.). Through continuous adjustment and optimization, it is ensured that the Transformer model can accurately correct the deviation in power prediction. Finally, the corrected result output by the Transformer model is applied to the preliminary predicted power of the power prediction characteristic model of the unit to obtain the final predicted power of the unit.

[0037] S6 To construct the overall power prediction model of the wind farm, we first obtain the topographic and geomorphic feature data of the wind farm through Geographic Information System (GIS) or satellite remote sensing technology. These data include information such as the terrain, landform, elevation, slope of the wind farm, which can help us understand the wind speed distribution in different areas of the wind farm and its impact on the power output of the wind turbines. Next, based on multi-source heterogeneous data, including the results of the power characteristic prediction model of the unit, the numerical weather forecast of the wind farm, the wind farm planning dataset, the historical data of the wind farm, and the terrain factors, the machine learning algorithm LightGBM is used to construct the overall power prediction model of the wind farm. Specifically, first collect and preprocess all relevant data: including the past power data of the wind turbines, the numerical weather forecast data, and the terrain data. Then, since LightGBM can handle large-scale and multi-dimensional data and can improve the prediction ability of the model through the way of ensemble learning, these multi-source heterogeneous data can be input into the LightGBM model for training, thus constructing the wind farm power prediction model. S7 The Transformer model is adopted to adaptively adjust the model parameters during the training process, and finally the overall predicted power of the wind farm is obtained. This process can effectively improve the accuracy and robustness of the wind farm power prediction, providing more reliable data support for the operation optimization of the wind farm.

Claims

1. A prediction method for the two - level power characteristics of the wind field - unit under cold wave weather, characterized in that, It includes the following steps: S1: Input the obtained meteorological data into a pre-trained cold wave meteorological identification model to obtain the prediction results of cold wave meteorology and the corresponding meteorological conditions. If the prediction result is a cold wave meteorology, proceed to the subsequent steps; S2: Obtain environmental parameters based on numerical weather prediction, construct a wind speed-power characteristic curve corrected by environmental parameters under the actual working state of the wind turbine, and simultaneously use the oil detection system to obtain the lubricating oil state information of the wind turbine in real time; S3: Based on the wind speed-power characteristic curve corrected by environmental parameters, establish a CFD simulation model for each wind turbine under cold wave meteorology using the wind turbine layout information, lubricating oil state information, real-time operating parameters of the fan, and environmental parameters to simulate the flow field characteristics around the wind turbine under cold wave meteorology, and obtain the flow field changes during the operation of the fan through CFD simulation; S4: Construct a power characteristic prediction model for the unit. The unit power characteristic prediction model takes meteorological data and flow field characteristics under different meteorological conditions as inputs and the unit power as the output; S5: Based on the power prediction characteristic model, predict the preliminary predicted power; combine historical power data, environmental parameters, and lubricating oil state information, and use the Transformer model or MLP model to correct the preliminary predicted power to obtain the predicted power of the wind turbine unit; S6: Use machine learning LightGBM to construct a power prediction model for the entire wind farm. The power prediction model for the entire wind farm takes the predicted power of the unit obtained in S5, as well as the pre-obtained wind farm numerical weather prediction data, wind farm planning data, the terrain and landform characteristic data, and wind farm SCADA data as inputs and the power of the entire wind farm as the output; S7: Adopt the Transformer model to adaptively adjust the model parameters of the power prediction model for the entire wind farm during the training process to obtain the corrected predicted power of the entire wind farm.

2. The cold wave weather wind field - unit two - stage power characteristic prediction method according to claim 1, characterized in that, The establishment of the cold wave meteorological identification model in S1 includes the following steps: S11: Collect historical meteorological data and perform normalization or standardization cleaning on the collected data; S12: Extract cold wave characteristics from the meteorological data; label the time periods when cold waves occur and their corresponding cold wave characteristics based on historical meteorological data, classify the cold wave meteorology in each time period according to the cold wave characteristics, and create cold wave labels; S13: After standardizing the selected cold wave characteristics in the meteorological data, use the Mean Shift clustering algorithm to divide different meteorological conditions under cold wave meteorology; S14: Combine the annotation of cold wave events with the meteorological data to generate a cold wave meteorological data set containing cold wave labels, and divide the data set into a training set, a validation set, and a test set for the training and evaluation of the pre-constructed cold wave meteorological identification model; the cold wave meteorological identification model takes meteorological data as input and cold wave meteorology and the corresponding meteorological conditions as output; S15: Use the cold wave meteorological data set to train the cold wave meteorological identification model, adjust the hyperparameters to improve the accuracy of the model, use the validation set to evaluate the model effect, and obtain the cold wave meteorological identification model to achieve the identification of cold wave meteorology.

3. The method for predicting the two-stage power characteristics of the wind field and the unit under cold wave weather according to claim 1, wherein: In S2, the environmental parameters include the temperature, wind speed, and humidity of the wind field.

4. The method for predicting the two-stage power characteristics of the wind field-unit under cold wave weather according to claim 3, wherein: In S3, the wind turbine layout information includes the position coordinates, spacing, number arrangement method of wind turbines in the wind field, and hub height; the lubricating oil state information includes lubricating oil viscosity, temperature, and moisture content; the real-time operation parameters of the wind turbine include cut-in and cut-out wind speeds, rated wind speed, number and diameter of blades.

5. The prediction method for the two-stage power characteristics of the wind field-turbine under cold wave weather according to claim 2, characterized in that: In S4, the unit power characteristic prediction model is established by using the Stacking algorithm in the ensemble learning method according to the flow field characteristics and meteorological data under different meteorological conditions; the meteorological data includes wind speed, wind direction, humidity, temperature, and air pressure; the flow field characteristics include turbulence intensity, wind speed distribution, and wake effect.

6. The cold snap weather wind field - unit two - stage power characteristic prediction method according to claim 5, characterized in that: The unit power characteristic prediction model uses SFOA (Starfish Optimization Algorithm) and / or HO (Hippopotamus Optimization Algorithm) to optimize the model parameters.

7. The method for predicting the two-stage power characteristics of the wind field and the unit under cold wave weather according to claim 1, wherein: In S6, the terrain and geomorphic feature data is obtained through GIS or satellite remote sensing technology, and the terrain and geomorphic feature data includes the terrain, landform, elevation, and slope of the wind field; The wind field numerical weather forecast data includes basic meteorological parameters, precipitation-related parameters, cloud visibility, radiation, and illumination; The wind field planning data includes geographical location information such as latitude, longitude, and altitude of the wind field, meteorological parameters, wind energy resource assessment, and wind farm planning data.

Images