A method, system, equipment and medium for predicting the output of a wind turbine under icing conditions.
By optimizing wind speed data using WRF simulation data and machine learning algorithms under icing conditions, the problem of predicting the nonlinear relationship between wind speed and wind power was solved, enabling precise minute-level prediction of wind turbine output and supporting refined grid scheduling.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUIZHOU POWER GRID CO LTD
- Filing Date
- 2025-12-31
- Publication Date
- 2026-06-02
AI Technical Summary
In existing technologies, numerical weather prediction models lack sufficient accuracy in wind speed prediction under complex terrain conditions, making it difficult to accurately capture the nonlinear relationship between wind speed and wind power under icing conditions. Furthermore, the time resolution of the prediction results is low, failing to meet the grid dispatching requirements for refined prediction.
By acquiring historical operating data of wind turbines, spatial interpolation is performed using WRF simulation forecast data, and a wind speed optimization model is trained using machine learning algorithms to optimize wind speed data. Finally, wind turbine output prediction results are calculated by combining temperature and air pressure, thereby improving prediction accuracy and time resolution.
It has achieved accurate prediction of wind turbine output under icing conditions, with the prediction results having a time resolution of minutes, meeting the grid dispatching requirements for refined prediction and improving the reliability of wind farm energy efficiency management and grid safe and stable operation.
Smart Images
Figure CN122136792A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wind field prediction technology, and in particular to a method, system, equipment and medium for predicting wind turbine output under icing conditions. Background Technology
[0002] With the intensification of global climate change, extreme freezing weather has become more frequent in southern my country in recent years, severely impacting the safe and stable operation of the power grid. The Yunnan-Guizhou region possesses abundant wind resources and has already built numerous wind farms; however, most of these wind farms are located in mountainous and plateau areas, making them susceptible to icing in winter, leading to turbine shutdowns. Statistics show that annual energy losses due to icing can reach 10% to 20% of total power generation. When the ice thickness on the leading edge of the turbine blades reaches a certain level, the maximum output power of the wind turbine can decrease by more than 50% or even lead to shutdown. Therefore, the problem of wind turbine blade icing has become a significant factor restricting wind power investment and development and the stable operation of the power grid in icy regions.
[0003] However, wind turbine output forecasting is limited by the accuracy of weather forecasts. Traditional numerical weather prediction models have low accuracy in wind speed prediction under complex mountainous terrain conditions, which directly affects the accuracy of wind turbine output forecasting. Existing forecasting methods are difficult to effectively capture the nonlinear relationship between wind speed and wind power under icing conditions, and the time resolution of the forecast results is low, usually only reaching the hourly level. Therefore, they cannot meet the grid dispatching requirements for refined forecasting. Summary of the Invention
[0004] In view of the aforementioned existing problems, the present invention is proposed.
[0005] Therefore, this invention provides a method, system, device, and medium for predicting wind turbine output under icing conditions to solve the problems of insufficient wind speed prediction accuracy in existing numerical weather prediction models under complex terrain conditions, difficulty in accurately capturing the nonlinear relationship between wind speed and wind power under icing conditions, and low temporal resolution of prediction results.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for predicting wind turbine output under icing conditions, comprising the following steps: acquiring historical wind turbine operation data of a target wind farm during the icing period, and processing the historical wind turbine operation data using a wind power calculation method to obtain wind power data; performing WRF simulation forecast data obtained by simulating the target wind farm during the icing period, and performing spatial interpolation processing on the WRF simulation forecast data to obtain interpolated wind speed data; matching the wind power data and the WRF simulation forecast data according to the time dimension to form a training dataset; training the training dataset using a machine learning algorithm to obtain a wind speed optimization model; optimizing the interpolated wind speed data using the wind speed optimization model to obtain optimized wind speed data; and substituting the optimized wind speed data and the temperature and air pressure in the WRF simulation forecast data into the wind power calculation formula to obtain the wind turbine output prediction result.
[0007] As a preferred embodiment of the wind turbine output prediction method under icing conditions described in this invention, the step of spatial interpolation processing of the WRF simulation forecast data includes: obtaining gridded wind speed data of the WRF simulation forecast data and the geographical coordinates of each wind turbine in the target wind farm; determining four adjacent grid points around each wind turbine in the grid of the gridded wind speed data based on the geographical coordinates of each wind turbine; calculating the second wind speed value of the location of each wind turbine using the first wind speed value of the four grid points; summarizing the second wind speed values of the location of each wind turbine, and using the summed data as the interpolated wind speed data.
[0008] The beneficial effects of this preferred technical solution are: by using spatial interpolation, the grid wind speed data output by the WRF mode is accurately mapped to the actual location of each wind turbine, which improves the spatial resolution and positional accuracy of the wind speed data and makes the prediction results more consistent with the actual operating status of the wind turbine.
[0009] As a preferred embodiment of the wind turbine output prediction method under icing conditions described in this invention, the step of matching the wind power data with WRF simulation forecast data according to the time dimension includes: unifying the time resolution of the wind power data to X minutes; unifying the time resolution of the WRF simulation forecast data to X minutes; matching the wind power data and WRF simulation forecast data at the same time according to the unified time resolution; and combining the matched wind power data and WRF simulation forecast data to form the training dataset.
[0010] As a preferred embodiment of the wind turbine output prediction method under icing conditions described in this invention, the step of obtaining the wind speed optimization model includes: dividing the training dataset into a training subset and a test subset according to a preset ratio; training the training subset using N sets of algorithms to obtain multiple candidate wind speed optimization models; evaluating the performance of each candidate wind speed optimization model using the test subset to obtain the evaluation index value corresponding to each candidate wind speed optimization model; selecting the model with the best performance from the multiple candidate wind speed optimization models based on the evaluation index value, and using the selected candidate wind speed optimization model as the wind speed optimization model.
[0011] The beneficial effects of this preferred technical solution are: by using N sets of algorithms for parallel training and comparison, the most suitable optimization algorithm can be selected for different wind farm terrain features and meteorological conditions, thereby improving the model's adaptability and prediction accuracy.
[0012] As a preferred embodiment of the wind turbine output prediction method under icing conditions described in this invention, the step of selecting the best-performing model from multiple candidate wind speed optimization models based on the evaluation index values includes: calculating the average relative error of wind power, wind power deviation, root mean square error of wind power, and average percentage error of wind power symmetry for each candidate wind speed optimization model; adding weights to the calculated evaluation index values and then calculating the final evaluation value; and selecting the candidate wind speed optimization model corresponding to the smallest final evaluation value as the wind speed optimization model.
[0013] The beneficial effects of this preferred technical solution are: it can comprehensively evaluate the model performance from multiple dimensions, ensuring that the selected model has balanced and stable predictive performance.
[0014] As a preferred embodiment of the wind turbine output prediction method under icing conditions described in this invention, the wind power calculation method is expressed as follows: ; In the formula, n is the number of records within the specified time period. air density; Let the cube of the i-th wind speed record be... This refers to wind power data; the formula for calculating air density is: ; in, Atmospheric pressure, This is the dry air specific gas constant. The temperature is in Kelvin.
[0015] As a preferred embodiment of the wind turbine output prediction method under icing conditions described in this invention, the wind power calculation formula is as follows: ; in, This refers to the output power of the fan. The wind power coefficient, air density, This represents the cross-sectional area of the wind turbine hub. The wind speed value is the wind speed value in the optimized wind speed data.
[0016] Secondly, the present invention provides a wind turbine output prediction system under icing conditions, comprising: a wind turbine data acquisition module, used to acquire historical operating data of wind turbines in a target wind farm during the icing period, and to process the historical operating data of wind turbines through a wind power calculation method to obtain wind power data; The WRF simulation module is used to simulate the target wind farm during the icing period to obtain WRF simulation forecast data, and to perform spatial interpolation processing on the WRF simulation forecast data to obtain interpolated wind speed data. The data matching module is used to match the wind power data with the WRF simulation forecast data according to the time dimension to form a training dataset; The model training module is used to train the training dataset using machine learning algorithms to obtain a wind speed optimization model; The wind speed optimization module is used to optimize the interpolated wind speed data using the wind speed optimization model to obtain optimized wind speed data. The power output prediction module is used to substitute the optimized wind speed data and the temperature and air pressure in the WRF simulation forecast data into the wind power calculation formula to obtain the wind turbine power output prediction result and output it.
[0017] Thirdly, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the wind turbine output prediction method under icing conditions are implemented.
[0018] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the wind turbine output prediction method under icing conditions.
[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: By combining the WRF numerical weather prediction model with N sets of algorithms, the wind speed data output by the WRF model is optimized using N sets of algorithms, thereby capturing the nonlinear relationship between wind speed and wind power under icing conditions, improving the accuracy of wind turbine output prediction. Secondly, the time resolution of the prediction results can reach the minute level, which is a qualitative leap compared with the traditional hourly prediction accuracy. It can meet the grid dispatching requirements for refined prediction and provide reliable data support for the energy efficiency management of wind farms in icing areas and the safe and stable operation of the grid. Attached Figure Description
[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 This is a schematic diagram of the overall process of the wind turbine output prediction method under icing conditions according to an embodiment of the present invention. Detailed Implementation
[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0023] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for predicting the output of a wind turbine under icing conditions is provided, comprising the following steps: S100. Obtain historical operating data of wind turbines in the target wind farm during the icing period, and process the historical operating data of wind turbines using a wind power calculation method to obtain wind power data.
[0024] S200. The WRF simulation forecast data obtained by simulating the target wind farm during the icing period is used to obtain the interpolated wind speed data by performing spatial interpolation on the WRF simulation forecast data.
[0025] S300. Match the wind power data with the WRF simulation forecast data according to the time dimension to form a training dataset.
[0026] S400. The training dataset is trained using a machine learning algorithm to obtain a wind speed optimization model.
[0027] S500. The interpolated wind speed data is optimized using the wind speed optimization model to obtain optimized wind speed data.
[0028] S600. Substitute the optimized wind speed data and the temperature and air pressure in the WRF simulation forecast data into the wind power calculation formula to obtain the wind turbine output prediction result.
[0029] It should be noted that most wind farms in southern my country are located in mountainous and plateau areas, where they are prone to icing in winter, leading to turbine shutdowns. Annual power losses due to icing can reach 10% to 20% of total power generation. Under icing conditions, traditional numerical weather prediction models have low accuracy in wind speed forecasting in complex mountainous terrain. The grid data output by the WRF model has spatial discrepancies with the actual location of the wind turbines. Furthermore, the relationship between wind speed and wind power under icing conditions is complex and nonlinear, making it difficult for traditional linear methods to accurately capture this characteristic. In addition, the time resolution of existing prediction methods is typically only at the hourly level, which cannot meet the grid dispatching requirements for refined forecasting.
[0030] Therefore, to address the aforementioned issues of prediction accuracy and time resolution, the following steps (S100-S600) are employed: First, historical wind turbine operating data is acquired and wind power data is calculated. Simultaneously, a meteorological simulation is performed using the WRF model, and spatial interpolation is used to improve the location accuracy of the wind speed data. Then, the wind power data and the WRF simulation forecast data are matched along the time dimension to form a training dataset. N sets of algorithms are used to train a wind speed optimization model, which optimizes and corrects the wind speed data output by the WRF model. Finally, the optimized wind speed data, along with temperature and air pressure parameters, are combined to calculate the wind turbine output prediction results, achieving accurate prediction of wind turbine output under icing conditions. The time resolution of the prediction results can reach the minute level, providing reliable data support for wind farm energy efficiency management and the safe and stable operation of the power grid.
[0031] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the above embodiment, a method for predicting the output of a wind turbine under icing conditions is provided.
[0032] In this embodiment of the application, S100 obtains the historical operating data of the wind turbines of the target wind farm during the icing period, and processes the historical operating data of the wind turbines through a wind power calculation method to obtain wind power data.
[0033] Specifically, the historical operating data of the wind turbines includes wind speed, temperature, air pressure, and power generation data recorded by the wind turbines during the icing period. Taking the Meihuashan Wind Farm as an example, the historical operating data of the wind turbines from 12:00 on February 2, 2024 to 08:00 on February 4, 2024 were obtained, which is a typical period of icing conditions.
[0034] The wind power calculation method is expressed as follows: ; In the formula, n is the number of records within the specified time period. air density; Let the cube of the i-th wind speed record be... This is wind power data; The formula for calculating air density is as follows: ; in, Atmospheric pressure, This is the dry air specific gas constant. The temperature is in Kelvin.
[0035] As can be seen from the above formula, the calculation of wind power data needs to take into account three parameters: wind speed, air pressure, and temperature. Among them, wind speed has the most significant impact on wind power, showing a cubic relationship. Air density is determined by both air pressure and temperature. The higher the air pressure and the lower the temperature, the greater the air density, and the greater the wind power data under the same wind speed conditions.
[0036] In one optional implementation, step S100, which involves acquiring historical wind turbine operation data of the target wind farm during the icing period, may also include quality control processing of the historical wind turbine operation data, removing outliers and missing values caused by sensor failures or communication interruptions, and filling in missing data using interpolation methods to ensure the integrity and reliability of the historical wind turbine operation data and avoid the impact of abnormal data on the subsequent wind power data calculation results.
[0037] In this embodiment of the application, S200, the target wind farm is simulated during the icing period to obtain WRF simulation forecast data, and the WRF simulation forecast data is spatially interpolated to obtain interpolated wind speed data.
[0038] Specifically, the WRF simulation forecast data includes wind speed, temperature, and air pressure data. Taking the Meihuashan Wind Farm as an example, the simulation period was selected from 12:00 on February 2, 2024 to 08:00 on February 4, 2024. The longitude range of the simulation area was 26.68°N to 26.69°N, and the latitude range was 104.54°E to 104.68°E. The WRF model simulation was performed using the MYJ boundary layer parameterization scheme, with a grid resolution of 111m, to obtain grid wind speed, temperature, and air pressure data covering the target wind farm area.
[0039] The steps for spatial interpolation processing of the WRF simulated forecast data include A1 to A4: A1. Obtain the gridded wind speed data of the WRF simulation forecast data, as well as the geographical coordinates of each wind turbine in the target wind farm.
[0040] Specifically, gridded wind speed data is extracted from the WRF mode output file, with each grid point corresponding to a latitude and longitude coordinate and the wind speed value at the corresponding time. At the same time, the geographical coordinates of each wind turbine, including longitude and latitude information, are obtained from the wind farm ledger information.
[0041] A2. Based on the geographical coordinates of each wind turbine, determine the four adjacent grid points around each wind turbine in the grid of the grid wind speed data.
[0042] Specifically, for any wind turbine, its location is determined in the regular grid output by the WRF mode based on its latitude and longitude coordinates, and the four adjacent grid points surrounding the wind turbine's location are determined, located in the lower left, lower right, upper left and upper right directions of the wind turbine's location, respectively.
[0043] A3. Calculate the second wind speed value at the location of each fan using the first wind speed value at the four grid points.
[0044] Specifically, a bilinear interpolation method is used to calculate the second wind speed value at the location of the wind turbine by weighting the first wind speed value at the four adjacent grid points based on the distance relationship between the wind turbine location and the four adjacent grid points. The bilinear interpolation first performs two linear interpolations in the longitude direction and then one linear interpolation in the latitude direction to obtain the estimated wind speed value at the wind turbine location.
[0045] A4. Summarize the second wind speed values at the locations of each fan, and use the summed data as the interpolated wind speed data.
[0046] Specifically, the steps A1 to A3 above are repeated for all wind turbines in the target wind farm to obtain the second wind speed value of each wind turbine at each time. The second wind speed values of all wind turbines are summarized and organized according to the wind turbine number and time sequence to form the interpolated wind speed data.
[0047] In an optional implementation, step S200, which involves simulating the target wind farm during the icing period to obtain WRF simulation forecast data, may also include selecting a boundary layer parameterization scheme for the WRF model. Based on the terrain features and climate conditions of the area where the target wind farm is located, a suitable boundary layer parameterization scheme may be selected from the YSU scheme, MYJ scheme, and ACM2 scheme to improve the wind speed simulation accuracy of the WRF model under complex mountainous terrain conditions.
[0048] In this embodiment of the application, S300 matches the wind power data with the WRF simulation forecast data according to the time dimension to form a training dataset.
[0049] The steps for matching the wind power data with the WRF simulation forecast data according to the time dimension include B1 to B4: B1. The time resolution of the wind power data is unified to X minutes.
[0050] Specifically, taking the Meihuashan Wind Farm as an example, the time resolution of wind power data is unified to 10 minutes. For wind power data with an original acquisition frequency higher than 10 minutes, downsampling is performed by averaging within a time window; for wind power data with an original acquisition frequency lower than 10 minutes, upsampling is performed using a linear interpolation method.
[0051] B2. The time resolution of the WRF simulation forecast data is unified to X minutes.
[0052] Specifically, the time resolution of the WRF simulation forecast data is standardized to 10 minutes, consistent with the time resolution of the wind power data. The wind speed, temperature, and air pressure data output by the WRF model are all processed according to the same time resolution.
[0053] B3. Match the wind power data and WRF simulation forecast data at the same time according to the unified time resolution.
[0054] Using timestamps as matching keywords, wind power data at the same moment is matched one-to-one with the corresponding interpolated wind speed, temperature, and air pressure data. Taking data from 12:00 on February 2, 2024 to 08:00 on February 4, 2024 as an example, with a time resolution of 10 minutes, a total of 264 matching records are generated.
[0055] B4. Combine the matched wind power data and WRF simulation forecast data to form the training dataset.
[0056] The interpolated wind speed, temperature, and air pressure data from each matched record are used as input features, and the corresponding wind power data is used as the output label, forming the training dataset. Each sample in the training dataset contains five fields: timestamp, wind speed, temperature, air pressure, and wind power.
[0057] In an optional implementation, step S300 involves matching the wind power data with the WRF simulation forecast data according to the time dimension. It may also include outlier detection processing on the matched data to remove abnormal matching records caused by equipment failure or extreme weather, ensuring the rationality and validity of each sample data in the training dataset and avoiding the impact of abnormal samples on the training effect of subsequent machine learning models.
[0058] In this embodiment of the application, S400 trains the training dataset using a machine learning algorithm to obtain a wind speed optimization model.
[0059] The steps to obtain the wind speed optimization model include C1 to C4: C1. Divide the training dataset into a training subset and a test subset according to a preset ratio.
[0060] Specifically, the training dataset is divided in a 7:3 ratio, with 70% of the data used as the training subset for model training and 30% used as the test subset for model performance evaluation. Taking Meihuashan Wind Farm as an example, 19 sets of wind turbine wind speed data are selected as the training subset, and 8 sets of wind turbine wind speed data are selected as the test subset.
[0061] C2. Train the training subset using N sets of algorithms respectively to obtain multiple candidate wind speed optimization models.
[0062] For example, the training subset is trained using the random forest algorithm, gradient boosting decision tree algorithm, support vector regression algorithm, and neural network algorithm, respectively, to obtain four candidate wind speed optimization models. Specifically, the random forest algorithm has 100 decision trees, a maximum tree depth of 10, and a maximum number of leaf nodes of 50; the gradient boosting decision tree algorithm has 100 base learners, a learning rate of 0.1, and a maximum tree depth of 10; the support vector regression algorithm has a penalty coefficient of 1, a linear kernel function, and a maximum number of iterations of 1000; and the neural network algorithm has 100 hidden layer neurons, an identity function as the activation function, and 1000 iterations.
[0063] C3. Use the test subset to evaluate the performance of each candidate wind speed optimization model and obtain the evaluation index value corresponding to each candidate wind speed optimization model.
[0064] Specifically, the wind speed, temperature, and air pressure data from the test subset are input into each candidate wind speed optimization model to obtain the wind speed prediction results of each model. Then, the wind power prediction value is calculated based on the wind speed prediction results. The wind power prediction value is compared with the actual wind power data in the test subset, and the average relative error of wind power, wind power deviation, root mean square error of wind power, and average percentage error of wind power symmetry of each candidate wind speed optimization model are calculated respectively.
[0065] C4. Select the best-performing model from multiple candidate wind speed optimization models based on the evaluation index value, and use the selected candidate wind speed optimization model as the wind speed optimization model.
[0066] Specifically, candidate wind speed optimization models with an average relative error of wind power less than a preset error threshold are selected to form a set of candidate models. Among these candidate models, the candidate wind speed optimization model with the smallest root mean square error of wind power is determined as the best-performing model and used as the wind speed optimization model. Taking the Meihuashan wind farm as an example, the random forest algorithm has an average relative error of 0.17, a wind power deviation of 0.11, a root mean square error of 31.37, and a symmetric average percentage error of 0.02, showing the best overall performance among the four algorithms. Therefore, the model trained using the random forest algorithm is selected as the wind speed optimization model.
[0067] In this embodiment of the application, S500 uses the wind speed optimization model to optimize the interpolated wind speed data to obtain optimized wind speed data.
[0068] The interpolated wind speed data obtained in S200 is input into the wind speed optimization model trained in S400. The wind speed optimization model corrects the interpolated wind speed data, eliminating the systematic bias of the WRF model under complex terrain conditions, and outputs optimized wind speed data. Taking Meihuashan Wind Farm as an example, the wind speed optimization model trained by the random forest algorithm optimizes the interpolated wind speed data to obtain the optimized wind speed data for each wind turbine at each time point.
[0069] The step of selecting the best-performing model from multiple candidate wind speed optimization models by evaluating the index values includes steps D1 to D3: D1. Calculate the average relative error of wind power, wind power deviation, root mean square error of wind power, and average percentage error of wind power for each candidate wind speed optimization model.
[0070] Specifically, the average relative error of wind power is used to measure the average deviation between the predicted and actual values; the wind power deviation is used to measure the direction and magnitude of the systematic shift of the predicted value relative to the actual value; the root mean square error of wind power is used to measure the overall magnitude of the prediction error and is more sensitive to larger errors; the symmetrical average percentage error of wind power is used to measure the percentage magnitude of the prediction error relative to the actual value, avoiding evaluation bias caused by differences in dimensions.
[0071] D2. Add weights to the calculated evaluation index values, and then calculate the final evaluation value.
[0072] Based on the importance of each evaluation index in wind turbine output prediction, corresponding coefficients are set for the average relative error of wind power, wind power deviation, root mean square error of wind power, and symmetrical average percentage error of wind power. The final evaluation value of each candidate wind speed optimization model is obtained by multiplying the value of each evaluation index with the corresponding coefficient and summing the results.
[0073] D3. The candidate wind speed optimization model corresponding to the smallest final evaluation value is taken as the wind speed optimization model.
[0074] The final evaluation values of each candidate wind speed optimization model are compared, and the candidate wind speed optimization model with the smallest final evaluation value is selected as the wind speed optimization model. Taking Meihuashan Wind Farm as an example, the average relative error of wind power obtained by the random forest algorithm is 0.17, the wind power deviation is 0.11, the root mean square error of wind power is 31.37, and the average percentage error of wind power symmetry is 0.02. The calculated final evaluation value is the smallest among the four algorithms. Therefore, the model trained by the random forest algorithm is selected as the wind speed optimization model.
[0075] In this embodiment of the application, S600 substitutes the optimized wind speed data and the temperature and air pressure from the WRF simulation forecast data into the wind power calculation formula to obtain the wind turbine output prediction result.
[0076] The power calculation formula is: ; in, This refers to the output power of the fan. The wind power coefficient, air density, This represents the cross-sectional area of the wind turbine hub. The wind speed value is the wind speed value in the optimized wind speed data.
[0077] Specifically, firstly, based on the temperature and air pressure data in the WRF simulation forecast data, the air density is calculated using the formula... The air density values at each moment are calculated. Then, the air density values, the wind turbine hub cross-sectional area, the wind power coefficient, and the wind speed values from the optimized wind speed data are substituted into the wind power calculation formula to calculate the predicted output power of each wind turbine at each moment. Finally, the predicted output power values of each wind turbine are summarized to form the wind turbine output prediction results. Taking the Meihuashan Wind Farm as an example, the prediction results have a time resolution of 10 minutes, and the average relative error of wind power is 0.17, which is better than the expected 0.25, meeting the grid dispatching requirements for refined prediction.
[0078] In an optional implementation, step S600, which substitutes the optimized wind speed data and the temperature and air pressure from the WRF simulation forecast data into the wind power calculation formula, may also include a reasonableness verification process for the wind turbine output prediction result. This involves comparing the wind turbine output prediction result with the wind turbine's rated power. If the wind turbine output prediction result exceeds the wind turbine's rated power, the wind turbine output prediction result is corrected to the wind turbine's rated power. If the wind turbine output prediction result is negative, the wind turbine output prediction result is corrected to zero, ensuring that the wind turbine output prediction result is within a reasonable power output range.
[0079] In summary, by combining the WRF numerical weather prediction model with N sets of algorithms, the wind speed data output by the WRF model is optimized using N sets of algorithms. This allows for the capture of the nonlinear relationship between wind speed and wind power under icing conditions, improving the accuracy of wind turbine output prediction. Furthermore, the time resolution of the prediction results can reach the minute level, representing a qualitative leap compared to the traditional hourly prediction accuracy. This meets the grid dispatching requirements for refined prediction and provides reliable data support for the energy efficiency management of wind farms in icing areas and the safe and stable operation of the power grid.
[0080] Example 3 illustrates a method for predicting the output of a wind turbine under icing conditions. It should be noted that the technical solution of this system for predicting the output of a wind turbine under icing conditions is based on the same concept as the technical solution of the method described above. Details not described in detail in this example can be found in the description of the technical solution of the method described above.
[0081] This embodiment also provides a wind turbine output prediction system for icing conditions, including: The wind turbine data acquisition module is used to acquire historical operating data of wind turbines in the target wind farm during the icing period, and to process the historical operating data of wind turbines through wind power calculation methods to obtain wind power data. The WRF simulation module is used to simulate the target wind farm during the icing period to obtain WRF simulation forecast data, and to perform spatial interpolation processing on the WRF simulation forecast data to obtain interpolated wind speed data. The data matching module is used to match the wind power data with the WRF simulation forecast data according to the time dimension to form a training dataset; The model training module is used to train the training dataset using machine learning algorithms to obtain a wind speed optimization model; The wind speed optimization module is used to optimize the interpolated wind speed data using the wind speed optimization model to obtain optimized wind speed data. The power output prediction module is used to substitute the optimized wind speed data and the temperature and air pressure in the WRF simulation forecast data into the wind power calculation formula to obtain the wind turbine power output prediction result and output it.
[0082] This embodiment also provides an electronic device suitable for predicting the output of wind turbines under icing conditions, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the wind turbine output prediction method under icing conditions as proposed in the above embodiment.
[0083] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for predicting wind turbine output under icing conditions as proposed in the above embodiments.
[0084] The storage medium proposed in this embodiment and the method for predicting wind turbine output under icing conditions proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.
[0085] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.
[0086] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for predicting the output of a wind turbine under icing conditions, characterized in that, Includes the following steps: Historical operating data of wind turbines in the target wind farm during the icing period are obtained, and the historical operating data of wind turbines is processed by the wind power calculation method to obtain wind power data; The WRF simulation forecast data of the target wind farm during the icing period is obtained, and the WRF simulation forecast data is spatially interpolated to obtain the interpolated wind speed data. The wind power data and WRF simulation forecast data are matched along the time dimension to form a training dataset; A wind speed optimization model is obtained by training the training dataset using machine learning algorithms; The interpolated wind speed data is optimized using the aforementioned wind speed optimization model to obtain optimized wind speed data. Substituting the optimized wind speed data and the temperature and air pressure from the WRF simulation forecast data into the wind power calculation formula, the predicted wind turbine output is obtained.
2. The method for predicting wind turbine output under icing conditions as described in claim 1, characterized in that, The steps for spatial interpolation processing of the WRF simulation forecast data include: Obtain gridded wind speed data from the WRF simulation forecast data, as well as the geographic coordinates of each wind turbine within the target wind farm; Based on the geographical coordinates of each wind turbine, determine the four adjacent grid points around each wind turbine in the grid of the grid wind speed data; The second wind speed value at the location of each wind turbine is calculated using the first wind speed value at the four grid points. The second wind speed values at the locations of each wind turbine are aggregated, and the aggregated data is used as the interpolated wind speed data.
3. The method for predicting wind turbine output under icing conditions as described in claim 2, characterized in that, The steps for matching the wind power data with the WRF simulation forecast data according to the time dimension include: The time resolution of the wind power data is standardized to X minutes; The time resolution of the WRF simulation forecast data is standardized to X minutes; Based on the unified time resolution, wind power data and WRF simulation forecast data at the same time are matched accordingly; The matched wind power data and WRF simulation forecast data are combined to form the training dataset.
4. The method for predicting wind turbine output under icing conditions as described in claim 3, characterized in that, The steps to obtain the wind speed optimization model include: The training dataset is divided into a training subset and a test subset according to a preset ratio; The training subset is trained using N sets of algorithms to obtain multiple candidate wind speed optimization models; The performance of each candidate wind speed optimization model is evaluated using the test subset to obtain the evaluation index value corresponding to each candidate wind speed optimization model. Based on the evaluation index value, the best-performing model is selected from multiple candidate wind speed optimization models, and the selected candidate wind speed optimization model is used as the wind speed optimization model.
5. The method for predicting wind turbine output under icing conditions as described in claim 4, characterized in that, The step of selecting the best-performing model from multiple candidate wind speed optimization models based on the evaluation index value includes: Calculate the average relative error of wind power, wind power deviation, root mean square error of wind power, and average percentage error of wind power symmetry for each candidate wind speed optimization model. Weights are added to the calculated evaluation index values, and then the final evaluation value is calculated. The candidate wind speed optimization model corresponding to the smallest final evaluation value is taken as the wind speed optimization model.
6. The method for predicting wind turbine output under icing conditions as described in claim 5, characterized in that, The wind power calculation method is expressed as follows: ； In the formula, n is the number of records within the specified time period. air density; Let the cube of the i-th wind speed record be... This is wind power data; The formula for calculating air density is as follows: ； in, Atmospheric pressure, This is the dry air specific gas constant. The temperature is in Kelvin.
7. The method for predicting wind turbine output under icing conditions as described in claim 6, characterized in that, The formula for calculating wind power is: ； in, This refers to the output power of the fan. The wind power coefficient, air density, This represents the cross-sectional area of the wind turbine hub. The wind speed value is the wind speed value in the optimized wind speed data.
8. A wind turbine output prediction system under icing conditions, using the method described in any one of claims 1-7, characterized in that, include: The wind turbine data acquisition module is used to acquire historical operating data of wind turbines in the target wind farm during the icing period, and to process the historical operating data of wind turbines through wind power calculation methods to obtain wind power data. The WRF simulation module is used to simulate the target wind farm during the icing period to obtain WRF simulation forecast data, and to perform spatial interpolation processing on the WRF simulation forecast data to obtain interpolated wind speed data. The data matching module is used to match the wind power data with the WRF simulation forecast data according to the time dimension to form a training dataset; The model training module is used to train the training dataset using machine learning algorithms to obtain a wind speed optimization model; The wind speed optimization module is used to optimize the interpolated wind speed data using the wind speed optimization model to obtain optimized wind speed data. The power output prediction module is used to substitute the optimized wind speed data and the temperature and air pressure in the WRF simulation forecast data into the wind power calculation formula to obtain the wind turbine power output prediction result and output it.
9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the wind turbine output prediction method under icing conditions as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the wind turbine output prediction method under icing conditions as described in any one of claims 1 to 7.