Wind power prediction method, device, equipment and medium

Abstract

The application provides a wind power prediction method, device, equipment and medium, wherein the method comprises: acquiring real-time wind power trend, real-time wind power and future wind power trend; the future wind power trend is obtained by analyzing future wind speed based on meteorological data; the real-time wind power trend, the real-time wind power and the future wind power trend are input into a pre-trained model to obtain predicted wind power of a to-be-tested wind farm; and the pre-trained model is an optimized neural network model. By predicting future wind speed based on meteorological data and analyzing the future wind speed, the method can more accurately capture the trend change of future wind power, thereby improving the prediction accuracy.

Description

Technical Field

[0001] This invention relates to the field of wind power prediction technology, specifically providing a wind power prediction method, device, equipment, and medium. Background Technology

[0002] Wind energy is an important renewable energy resource, crucial for energy and environmental security. Electricity is a vital secondary energy source, critical for energy conservation, emission reduction, and social progress. Therefore, maximizing the conversion of wind and solar energy into electricity and delivering high-quality power smoothly and without obstruction to users has become a core aspect of the energy revolution. However, the volatility and intermittent nature of wind power introduces significant uncertainty, and its impact on power reliability, power quality, economics, and social welfare becomes increasingly prominent with rising penetration rates. Power system stability reflects the system's ability to withstand system collapse, while power system adequacy reflects its ability to meet user demands for power and electricity. Its risk analysis and control involve uncertainty, bifurcation, and chaos theory, and the large-scale integration of wind power significantly increases the impact of uncertainties. Wind power is gradually becoming the third largest power source after thermal and hydropower. The randomness, volatility, uncertainty, and weak controllability of wind power output pose significant challenges to the safe and stable operation of the power system. High-precision wind power forecasting technology has become an essential operational technique.

[0003] When using meteorological data to predict wind power, the randomness, volatility, and uncertainty of this data pose significant challenges to prediction models. Neural networks, as a powerful tool, excel at fitting nonlinear data and are therefore widely used in wind power prediction. However, the ratio of effective to ineffective information varies among different meteorological data sets. This can lead to ineffective information masking effective information, thus affecting the fitting calculations of the neural network and reducing prediction accuracy.

[0004] Therefore, a new wind power prediction scheme is needed in this field to solve the above problems. Summary of the Invention

[0005] To overcome the above-mentioned shortcomings, this invention is proposed to provide a solution to the problem of low prediction accuracy.

[0006] In a first aspect, the present invention provides a wind power prediction method, comprising: acquiring real-time wind power trend, real-time wind power, and future wind power trend; wherein the future wind power trend is obtained by analyzing the future wind speed based on meteorological data; inputting the real-time wind power trend, real-time wind power, and future wind power trend into a pre-trained model to obtain the predicted wind power of the wind farm to be measured; wherein the pre-trained model is an optimized neural network model, and the optimization process includes at least: training the neural network model based on a training set, verifying the trained model, and determining whether the verification result meets preset conditions; if the preset conditions are met, continuing to optimize the hyperparameters of the neural network model; otherwise, outputting the optimal parameters.

[0007] In one technical solution of the above-mentioned wind power prediction method, before obtaining the future wind power trend, the method further includes: obtaining the wind direction within a certain time interval; and dividing the wind direction within the time interval into different wind direction modules.

[0008] In one technical solution of the above-mentioned wind power prediction method, the meteorological data includes at least: air pressure and air density. Obtaining the future wind power trend includes at least: weighting the actual wind speeds corresponding to different directional modules to obtain a weighted result; inputting the obtained air pressure, future air density, and weighted result of the wind farm to be measured into a trained machine learning model or deep learning model to obtain the predicted future wind speed; and subtracting the obtained real-time wind speed from the predicted future wind speed to obtain the predicted future wind speed trend.

[0009] In one technical solution of the above-mentioned wind power prediction method, obtaining the future wind power trend further includes: inputting the weighted result, the future wind speed trend, and the obtained real-time wind speed trend into a trained machine learning model or deep learning model to obtain the predicted wind power of the wind farm to be tested; and subtracting the obtained real-time wind power from the predicted wind power to obtain the future wind power trend of the wind farm to be tested.

[0010] In one technical solution of the above-mentioned wind power prediction method, the preset conditions are obtained based on the training error of the acquired neural network model and the preset target error.

[0011] In one technical solution of the above-mentioned wind power prediction method, the process of obtaining the training error of the neural network model includes: inputting the training set into the neural network model for training, inputting the validation set into the trained model to obtain the first prediction data; and comparing the first prediction data with the test data to obtain the training error.

[0012] In one technical solution of the above-mentioned wind power prediction method, the process of obtaining the future air density includes: obtaining the temperature and humidity of the wind farm to be measured within the time interval; inputting the temperature and humidity into a trained model to obtain the future air density; the model is at least one of the following: Northern Eagle Algorithm Optimized Deep Extreme Learning Machine Algorithm, Time Series Model, Machine Learning Model, and Deep Learning Model.

[0013] In a second aspect, the present invention provides a wind power prediction device, comprising: an acquisition module for acquiring real-time wind power trend, real-time wind power, and future wind power trend; wherein the future wind power trend is obtained by analyzing the future wind speed based on meteorological data; and a prediction module for inputting the real-time wind power trend, real-time wind power, and future wind power trend into a pre-trained model to obtain the predicted wind power of the wind farm to be measured; wherein the pre-trained model is an optimized neural network model, and the optimization process includes at least: training the neural network model based on a training set, verifying the trained model, and determining whether the verification result meets preset conditions; if the preset conditions are met, continuing to optimize the hyperparameters of the neural network model; otherwise, outputting the optimal parameters.

[0014] In a third aspect, an electronic device includes a processor and a storage device, the storage device being adapted to store a plurality of program codes, the program codes being adapted to be loaded and run by the processor to execute the wind power prediction method of any of the above-described technical solutions of the wind power prediction method.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium storing a plurality of program codes adapted to be loaded and run by a processor to perform the wind power prediction method of any of the above-described technical solutions.

[0016] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects:

[0017] In implementing the technical solution of this invention, the present invention provides a wind power prediction method, comprising: acquiring real-time wind power trend, real-time wind power, and future wind power trend; the future wind power trend is obtained by analyzing the future wind speed based on meteorological data; inputting the real-time wind power trend, real-time wind power, and future wind power trend into a pre-trained model to obtain the predicted wind power of the wind farm to be measured; the pre-trained model is an optimized neural network model, and the optimization process includes at least: training the neural network model based on a training set, verifying the trained model, and determining whether the verification result meets preset conditions; if the preset conditions are met, continuing to optimize the hyperparameters of the neural network model; otherwise, outputting the optimal parameters. Compared with the prior art, the beneficial effects of the wind power prediction method provided by this invention are:

[0018] By predicting future wind speeds based on meteorological data and analyzing these future wind speeds, we can more accurately capture the trend changes in future wind power, thereby improving prediction accuracy.

[0019] Furthermore, by fitting temperature, humidity, air pressure, and weighted wind speed, the future wind speed is obtained. The difference between the future wind speed and the current wind speed is used to obtain the relative trend. Based on the obtained relative wind speed trend, further fitting is performed to predict the future wind power. The above method takes into account the full utilization of data information and extracts mutual information from meteorological data that is weakly correlated with wind power data in another dimension, thereby improving the accuracy of wind speed prediction data. This ensures that the meteorological training dataset is fully utilized and improves the efficiency of the meteorological data prediction model.

[0020] Furthermore, by incorporating wind direction into the weighted calculation of actual wind speed, the temporal characteristics of wind speed are enhanced. This allows for a more accurate capture of wind speed changes and periodicity, thereby improving the accuracy of the prediction model. Additionally, wind speed exhibits a degree of volatility and randomness; enhancing the temporal characteristics helps smooth these fluctuations, reducing abrupt changes and fluctuations in prediction results, thus improving prediction stability. Attached Figure Description

[0021] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein:

[0022] Figure 1 This is a schematic flowchart of the main steps of a wind power prediction method according to an embodiment of the present invention;

[0023] Figure 2 This is a structural diagram of KOA-LSTM according to an embodiment of the present invention;

[0024] Figure 3 This is a diagram of an LSTM neural network structure according to an embodiment of the present invention;

[0025] Figure 4 This is a comparison chart of the prediction results and actual analysis data effects of an LSTM model without introducing trend values ​​according to an embodiment of the present invention.

[0026] Figure 5 This is a graph showing the relative error results of an LSTM model without introducing trend values ​​according to an embodiment of the present invention;

[0027] Figure 6 This is a comparison chart of the prediction results of an LSTM model incorporating trend values ​​and the actual analysis data according to an embodiment of the present invention;

[0028] Figure 7 This is a graph illustrating the relative error results of an LSTM model incorporating trend values ​​according to an embodiment of the present invention.

[0029] Figure 8 This is a main structural block diagram of a wind power prediction device according to an embodiment of the present invention. Detailed Implementation

[0030] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0031] Example 1

[0032] See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the main steps of a wind power prediction method according to an embodiment of the present invention. Figure 1 As shown, the wind power prediction method in this embodiment of the invention mainly includes the following steps S1-S2.

[0033] Step S1: Obtain real-time wind power trend, real-time wind power, and future wind power trend; the future wind power trend is obtained by analyzing the future wind speed based on meteorological data to predict the future wind speed.

[0034] In this embodiment, the real-time wind power trend refers to the changing trend of the power generated by the wind farm at the current moment. By monitoring changes in real-time power data, the power generation status of the wind farm and its potential future power generation trends can be understood. For example, the process of obtaining the real-time wind power trend is: the difference between the wind power at the current moment and the wind power at the previous moment. Real-time wind power represents the actual power generated by the wind farm at the current moment. This is an important indicator of the real-time operating status of the wind farm and can reflect the impact of factors such as wind speed on power generation. Future wind power trend is a prediction of the wind farm's power generation trend over a future period based on meteorological data. By analyzing and predicting future meteorological factors such as wind speed, the future power generation status of the wind farm can be predicted.

[0035] Specifically, real-time wind power trends and real-time wind power output can be achieved through the following methods: monitoring equipment for wind turbines, including anemometers, power meters, and data acquisition systems. The data acquisition system acquires meteorological data such as wind speed, direction, and force in real time, and calculates the real-time wind power output based on the characteristics of the wind turbine. The acquired data is then transmitted to a data center or operations center via a monitoring system or data interface for storage and analysis, ultimately generating real-time wind power trend charts and real-time power values.

[0036] like Figure 2 As shown, in step S2, the real-time wind power trend, real-time wind power, and future wind power trend are input into the pre-trained model to obtain the predicted wind power of the wind farm to be tested.

[0037] The pre-trained model is an optimized neural network model. The optimization process includes at least: training the neural network model based on the training set, testing the trained model, and determining whether the test result meets the preset conditions. If the preset conditions are met, the neural network model is further optimized for hyperparameters; otherwise, the optimal parameters are output.

[0038] In this embodiment, historical data and actual wind power values ​​can be used as the training set. A neural network algorithm learns the patterns and relationships between the training data. The trained neural network model is then tested using validation set data to examine its performance on unseen data. The test results are then evaluated to determine if they meet preset conditions, such as whether the error range is within acceptable limits. If the test results do not meet the preset conditions, the hyperparameters of the neural network model, such as the learning rate, network structure, and activation function, need further optimization. By adjusting the hyperparameters, the model's performance is continuously optimized to obtain more accurate prediction results.

[0039] For example, neural network models can be used to optimize the Kepler algorithm for the Long Short-Term Memory Neural Network Model (KOA-LSTM), as well as for autoregressive moving average models, support vector machine regression, random forest regression, and convolutional neural networks.

[0040] In one embodiment, the preset conditions are obtained based on the training error of the acquired neural network model and the preset target error.

[0041] The process of obtaining the training error of a neural network model includes: inputting the training set into the neural network model for training, inputting the validation set into the trained model to obtain the first prediction data, and comparing the first prediction data with the test data to obtain the training error.

[0042] The following is an example of optimizing a Long Short-Term Memory Neural Network (KOA-LSTM) model using the Kepler algorithm: First, set the evaluation metric for the LSTM model: for example, select the root mean square error (RMSE) to evaluate the performance of the LSTM model, which represents the average difference between the model's predicted value and the actual value. The smaller the RMSE, the higher the prediction accuracy of the model. Then, merge real-time wind power data, wind power trend data, and future wind power trend data with historical datasets, and perform preprocessing steps such as data cleaning and feature engineering. Using the prepared dataset, divide the dataset into training and test sets, and then input them into the LSTM model for training and prediction. Compare the data predicted by the LSTM model with the test data, calculate the root mean square error (RMSE), and use it as the objective function of the Kepler algorithm (KOA). If the calculated root mean square error exceeds the pre-set objective function error threshold, use the Kepler algorithm (KOA) to optimize the initial parameters of the LSTM, update the parameters, and continue training and prediction. Repeat the above steps until the root mean square error of the LSTM is less than the set objective function error threshold. The neural network structure diagram of LSTM is as follows: Figure 3 As shown, Figure 3 The meanings of each quantity in the table are as follows:

[0043] Input gate: i t =σ(W i [x t ,h t -1]+b i )

[0044] Forgotten Gate: f t =σ(W f [x t ,h t -1]+b f )

[0045] Output gate: o t =σ(Wo [x t ,h t -1]+b o )

[0046] Candidate cell state: c t ′=tanh(W c [x t ,h t -1]+b c )

[0047] Cell state: c t =f t *c t-1 +i t *c t ′

[0048] Hidden state: h t =o t *tanh(c t )

[0049] In one embodiment, before obtaining the future wind power trend, the method further includes: obtaining the wind direction within a certain time interval; and dividing the wind direction within the time interval into different wind direction modules.

[0050] In this embodiment, a given interval window is provided, which moves with the time series. The wind direction within this interval is projected onto a polar coordinate three-dimensional system with wind speed as the ordinate. This process correlates wind direction and wind speed, forming a two-dimensional vector of wind direction-wind speed, while also considering the magnitude of the wind speed. Then, based on these wind direction-wind speed vectors, cluster analysis can be performed to divide the wind direction into different modules. Commonly used methods in cluster analysis include K-means clustering, hierarchical clustering, and density clustering. Specifically, the wind direction trend is detected in real time. If a new linear change in wind direction is detected, the new wind direction is determined to be the mainstream wind direction. Based on the continuously changing interval corresponding to the new change, the window extends until a non-linear change occurs, dividing the wind direction into the same module. This process continues to identify different wind direction modules, thus completing the cluster analysis of wind direction.

[0051] Assuming wind direction changes at a certain rate over time, this change can be approximated by a straight line. For example, if the wind direction gradually changes from North (N) to Northeast (NE) and then to East (E), and there are no drastic fluctuations during this change, it can be approximated as a linear change. Within a given time window, a linear fitting method can be used to detect the trend of wind direction change. If the fitting results show that the trend of wind direction change can be described well by a straight line, then the wind direction change during that period can be considered linear. Alternatively, the rate and direction of wind direction change can also be monitored. If the rate and direction of wind direction change remain relatively stable over a certain period (i.e., the rate of change is small and the direction change is minimal), then a linear change can be considered to have occurred.

[0052] In one embodiment, the meteorological data includes at least: air pressure and air density, and obtaining the future wind power trend in step S1 includes at least:

[0053] S11. Weight the actual wind speeds corresponding to modules in different directions to obtain the weighted result;

[0054] In this embodiment, weighting the actual wind speeds corresponding to modules in different directions is performed to more accurately account for the impact of different wind directions on the wind farm. In wind power generation, wind direction significantly affects wind speed because the turbine blades are affected to varying degrees depending on the wind direction. Therefore, wind speeds in different directions have different impacts on the turbine's power generation capacity. Using the above method, the actual wind speed situation of the entire wind farm can be better reflected, thereby improving the accuracy and reliability of wind power prediction.

[0055] S12. Based on the obtained air pressure, future air density, and weighted results of the wind farm to be measured, input them into the trained machine learning model or deep learning model to obtain the predicted future wind speed.

[0056] In this embodiment, future air density refers to the air density at a certain point in time or within a certain time period in the future. The air pressure of the wind farm to be measured, the future air density, and the weighted result can be input into a gradient booster (LightGBM) to predict the future wind speed of the wind farm within that interval. Alternatively, other machine learning models can be used, such as random forests, support vector regression, or one or more of the following deep learning models: gated recurrent units, convolutional neural networks, and ensemble learning, as long as the predicted future wind speed can be obtained.

[0057] S13. The difference between the acquired real-time wind speed and the predicted future wind speed is calculated to obtain the predicted future wind speed trend.

[0058] In this embodiment, the wind speed at the current moment can be obtained by measuring the real-time wind speed using a wind tower.

[0059] Furthermore, the predicted future wind speed is the wind speed after applying a step lag operation to the predicted wind speed. By appropriately lag-processing the predicted wind speed, the inertia and delay effects of wind speed changes can be taken into account, thus better reflecting the actual future wind speed. This operation helps improve the accuracy of wind power forecasting, making the forecast results more valuable and reliable.

[0060] In one embodiment, obtaining the future wind power trend in step S1 further includes:

[0061] S14. Input the weighted result, the future wind speed trend and the obtained real-time wind speed trend into the trained machine learning model or deep learning model to obtain the predicted wind power of the wind farm to be tested.

[0062] S15. The difference between the acquired real-time wind power and the predicted wind power is calculated to obtain the future wind power trend of the wind farm to be tested.

[0063] In this embodiment, the machine learning model or deep learning model can use a long short-term memory (LSTM) network. Alternatively, it can use one or more of the following deep learning models: support vector regression, random forest regression, gradient boosting regression tree, or other deep learning models, as long as they can obtain the future wind power trend.

[0064] Furthermore, the predicted future wind power is obtained by applying a step lag operation to the predicted wind power. This process captures the trend of power variation. This helps to better understand historical patterns and trends in power, thereby improving the accuracy of future wind power predictions.

[0065] In one embodiment, the process of obtaining the future air density in step S12 includes:

[0066] Step S121: Obtain the temperature and humidity of the wind farm to be tested within the time interval;

[0067] Step S122: Input the temperature and humidity into the trained model to obtain the future air density; the model is at least one of the following: Northern Eagle Algorithm Optimized Deep Extreme Learning Machine Algorithm, Time Series Model, Machine Learning Model, and Deep Learning Model.

[0068] In this embodiment, temperature and humidity have a significant impact on the power generation efficiency of wind farms: temperature and humidity are important factors affecting air density, which in turn directly affects the performance of wind turbines. By acquiring and analyzing temperature and humidity data of the wind farm under test within a given time period, future changes in air density can be predicted more accurately, helping to optimize the operation and power generation efficiency of the wind farm.

[0069] Historical temperature, humidity, air pressure, hub height wind direction, hub height wind speed, and power generation data from a wind farm in Xinjiang from 00:00 on January 1, 2019 to 24:00 on December 31, 2019 were used as raw data. This prediction task was executed on the Matlab2023a software platform, and the experimental hardware configuration was Intel(R) Core(TM) i5-8550U CPU / 8GB RAM / NVIDIA GeForce RTX3050.

[0070] This invention selects the temperature and humidity data of a wind farm in Xinjiang for the entire year of 2019, inputs it into the Northern Eagle Algorithm Optimized Deep Extreme Learning Machine Algorithm (NGO-DELM) to predict the air density data of the wind farm in 2019, and then inputs the air density data, air pressure data, and feature-weighted wind speed data into the Gradient Boosting Machine (LightGBM) to predict the wind speed data of the wind farm in 2019.

[0071] After performing a step lag operation on the future wind speed data, the difference between the future wind speed data and the hub height wind speed data is processed to obtain the wind speed trend value from the previous moment to the current moment and the wind speed trend value from the current moment to the next moment. These values, along with the hub height wind speed at the current moment, are then input into the LSTM network to predict the wind power data of the wind farm in 2019.

[0072] The difference between the acquired real-time wind speed and the predicted future wind speed is calculated to obtain the predicted future wind speed trend. The difference between the acquired real-time wind power and the predicted wind power is calculated to obtain the future wind power trend of the wind farm under test. The obtained real-time wind power trend, real-time wind power, and future wind power trend are input into the Kepler optimization algorithm to optimize the Long Short-Term Memory Network (KOA-LSTM) to obtain the final predicted wind power value of the wind farm.

[0073] This invention selects wind power prediction experiments with and without wind speed trend values. Both experiments use an LSTM model for prediction, yielding the wind power prediction image and the relative error image between the two models. The wind power prediction values ​​are taken from the 5000th to the 5200th data point. Figures 4 to 7 The image shows a line graph of the wind power prediction values ​​from the 5000th to the 5200th data point over time. The LSTM model relative error image is the difference between the predicted and actual values ​​of all data samples, presented in the form of a bar chart. As can be seen from the image, the predicted values ​​with the trend value introduced are closer to the actual values ​​than those without the trend value introduced, and the error is also smaller.

[0074] Example 2

[0075] See appendix Figure 8 , Figure 8 This is a main structural block diagram of a wind power prediction device according to an embodiment of the present invention. Figure 8 As shown, the wind power prediction device in this embodiment of the invention mainly includes an acquisition module 11 and a prediction module 12. In some embodiments, one or more of the acquisition module 11 and the prediction module 12 can be combined into a single module. In some embodiments, the acquisition module 11 can be configured to execute the program of step S1. The prediction module 12 can be configured to execute the program of step S2. In one embodiment, a description of the specific functions can be found in steps S1-S2.

[0076] The aforementioned wind power prediction device is used for performing Figure 1 The wind power prediction method embodiments shown are similar in technical principle, the technical problems solved and the technical effects produced. Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the wind power prediction device can be referred to the content described in the embodiments of the wind power prediction method, and will not be repeated here.

[0077] Those skilled in the art will understand that all or part of the processes in the method of the above-described embodiment of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium can include any entity or device capable of carrying computer program code, media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0078] Example 3

[0079] The present invention also provides an electronic device. In one embodiment of the invention, the device includes a processor and a storage device. The storage device can be configured to store a program for executing the wind power prediction method of the above-described method embodiments. The processor can be configured to execute the program in the storage device, which includes, but is not limited to, a program for executing the wind power prediction method of the above-described method embodiments. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The control device can be a control device device comprising various electronic devices.

[0080] Example 4

[0081] The present invention also provides a computer-readable storage medium. In one embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium can be configured to store a program for executing the wind power prediction method of the above-described method embodiments. This program can be loaded and run by a processor to implement the above-described wind power prediction method. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The computer-readable storage medium can be a storage device comprising various electronic devices. Optionally, in the embodiments of the present invention, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0082] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the original technical features, and the technical solutions resulting from these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for predicting wind power output, characterized in that, include: The system acquires real-time wind power trends, real-time wind power output, and future wind power trends. The future wind power trend is obtained by predicting future wind speeds based on meteorological data and analyzing the future wind speeds. The real-time wind power trend, the real-time wind power, and the future wind power trend are input into a pre-trained model to obtain the predicted wind power of the wind farm to be tested. The pre-trained model is an optimized neural network model. The optimization process includes at least the following steps: training the neural network model based on the training set, verifying the trained model, and determining whether the verification result meets preset conditions. If the preset conditions are met, the hyperparameters of the neural network model are further optimized; otherwise, the optimal parameters are output. Real-time wind power data, wind power trend data, future wind power trend data, and historical datasets are merged and preprocessed using data cleaning and feature engineering. The preprocessed dataset is then divided into training and testing sets and input into an LSTM model for training and prediction. The data predicted by the LSTM model is compared with the test data to calculate the root mean square error (RMSE), which is used as the objective function of the Kepler algorithm (KOA). If the calculated RMSE exceeds a pre-set objective function error threshold, the initial parameters of the LSTM model are optimized using the Kepler algorithm (KOA), and training and prediction continue after updating the parameters. This process of training and prediction is repeated until the RMSE of the LSTM model is less than the objective function error threshold. Prior to obtaining the future wind power trend, the following steps are also included: Obtain wind direction over a given period of time; The wind direction within this time interval is divided into different wind direction modules, specifically including: Set an interval window, which moves as the time series progresses; project the wind direction within the interval onto a polar coordinate three-dimensional system with wind speed as the ordinate, forming a two-dimensional vector of wind direction-wind speed; perform cluster analysis based on the two-dimensional vector of wind direction-wind speed to divide different wind direction modules. The cluster analysis process involves real-time detection of wind direction trends. If a new linear change in wind direction is detected, the new wind direction is determined to be the mainstream wind direction. Based on the continuously changing intervals corresponding to the new wind direction, the interval window extends until a non-linear change occurs, and these intervals are divided into the same wind direction module. Different wind direction modules are then identified to complete the cluster analysis of wind directions.

2. The method according to claim 1, characterized in that, The meteorological data includes at least: air pressure and air density; the future wind power trend is obtained by at least: The actual wind speeds corresponding to modules in different directions are weighted to obtain a weighted result. Based on the obtained air pressure, future air density, and weighted results of the wind farm to be measured, the predicted future wind speed is obtained by inputting them into a trained machine learning model or deep learning model. The difference between the acquired real-time wind speed and the predicted future wind speed is calculated to obtain the predicted future wind speed trend.

3. The method according to claim 2, characterized in that, Obtaining the aforementioned future wind power trend also includes: The weighted result, the future wind speed trend, and the acquired real-time wind speed trend are input into a trained machine learning model or deep learning model to obtain the predicted wind power of the wind farm to be tested. The difference between the acquired real-time wind power and the predicted wind power is calculated to obtain the future wind power trend of the wind farm under test.

4. The method according to claim 1, characterized in that, The preset conditions are obtained based on the training error of the acquired neural network model and the preset target error.

5. The method according to claim 4, characterized in that, The process of obtaining the training error of a neural network model includes: The training set is input into the neural network model for training, and the validation set is input into the trained model to obtain the first prediction data. The training error is obtained by comparing the first predicted data with the test data.

6. The method according to claim 2, characterized in that, The process of obtaining the future air density includes: The temperature and humidity of the wind farm under test are obtained within the time interval. The temperature and humidity are input into the trained model to obtain the future air density; the model is one of the following: Northern Eagle Algorithm, Optimized Deep Extreme Learning Machine Algorithm, Time Series Model, Machine Learning Model, and Deep Learning Model.

7. A wind power prediction device, characterized in that, include: The acquisition module is used to acquire real-time wind power trend, real-time wind power, and future wind power trend; the future wind power trend is obtained by predicting future wind speed based on meteorological data and analyzing the future wind speed. The prediction module is used to input the real-time wind power trend, real-time wind power, and future wind power trend into a pre-trained model to obtain the predicted wind power of the wind farm to be tested. The pre-trained model is an optimized neural network model. The optimization process includes at least the following steps: training the neural network model based on the training set, verifying the trained model, and determining whether the verification result meets preset conditions. If the preset conditions are met, the hyperparameters of the neural network model are further optimized; otherwise, the optimal parameters are output. Real-time wind power data, wind power trend data, future wind power trend data, and historical datasets are merged and preprocessed using data cleaning and feature engineering. The preprocessed dataset is then divided into training and testing sets and input into an LSTM model for training and prediction. The data predicted by the LSTM model is compared with the test data to calculate the root mean square error (RMSE), which is used as the objective function of the Kepler algorithm (KOA). If the calculated RMSE exceeds a pre-set objective function error threshold, the initial parameters of the LSTM model are optimized using the Kepler algorithm (KOA), and training and prediction continue after updating the parameters. This process of training and prediction is repeated until the RMSE of the LSTM model is less than the objective function error threshold. Prior to obtaining the future wind power trend, the following steps are also included: Obtain wind direction over a given period of time; The wind direction within this time interval is divided into different wind direction modules, specifically including: Set an interval window, which moves as the time series progresses; project the wind direction within the interval onto a polar coordinate three-dimensional system with wind speed as the ordinate, forming a two-dimensional vector of wind direction-wind speed; perform cluster analysis based on the two-dimensional vector of wind direction-wind speed to divide different wind direction modules. The cluster analysis process involves real-time detection of wind direction trends. If a new linear change in wind direction is detected, the new wind direction is determined to be the mainstream wind direction. Based on the continuously changing intervals corresponding to the new wind direction, the interval window extends until a non-linear change occurs, and these intervals are divided into the same wind direction module. Different wind direction modules are then identified to complete the cluster analysis of wind directions.

8. An electronic device comprising a processor and a storage device, the storage device being adapted to store multiple lines of program code, characterized in that, The program code is adapted to be loaded and run by a processor to perform the wind power prediction method of any one of claims 1 to 6.

9. A computer-readable storage medium storing a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by a processor to perform the wind power prediction method of any one of claims 1 to 6.

Images