Power prediction method and device of wind driven generator, storage medium and electronic device

Abstract

The invention discloses a power prediction method for a wind driven generator, and the method comprises the steps: obtaining the historical environment data of a wind power station; screening the historical environment data to obtain target historical environment data, extracting wind power data from the target historical environment data, and determining a target wind power change rate based on the wind power data; inputting the historical environment data and the target wind power change rate into a neural network model to obtain historical predicted power; training the neural network model based on the historical prediction power to obtain a power prediction model; wherein the power prediction model is used for predicting the real-time prediction power of the wind driven generator. And furthermore, through accurately predicting the real-time power of the wind driven generator, the operation hidden danger problem of the wind driven generator can be found in time.

Description

Technical Field

[0001] The present application relates to the technical field of wind turbines, and in particular to a method and device for predicting power of a wind turbine, a storage medium, and an electronic device. Background Art

[0002] As a clean and renewable energy source, wind power generation is becoming increasingly important. As the core component of the wind power system, the operating status of wind turbines directly affects the power generation efficiency and economic benefits of the wind farm. In order to ensure the efficient operation of wind turbines and reduce power generation losses caused by equipment failures, real-time and accurate power prediction of wind turbines and then reflect the actual performance of wind turbines based on the power prediction results have become an indispensable part of wind power operation and maintenance.

[0003] At present, the working status of wind turbines is usually evaluated by comparing actual operating data with the static standard power curve provided by the manufacturer. However, this method has significant limitations. First, the standard power curve is usually based on simulation calculations under ideal conditions and fails to fully consider the complex and changeable actual environment of wind farms. Secondly, the static curve is difficult to capture the dynamic characteristics of the equipment during operation and cannot accurately reflect the actual performance of wind turbines under various working conditions, resulting in a significant reduction in the accuracy and timeliness of detection.

[0004] Therefore, how to accurately predict the power of wind turbines and promptly discover hidden dangers in the operation of wind turbines has become a problem that needs to be solved urgently. Summary of the invention

[0005] The embodiments of the present application provide a method and device for predicting power of a wind turbine, a storage medium, and an electronic device, so as to at least solve the problem of low accuracy of power prediction of a wind turbine in the related art.

[0006] According to one embodiment of the embodiments of the present application, a power prediction method for a wind turbine is provided, comprising: acquiring historical environmental data of a wind farm station; screening the historical environmental data to obtain target historical environmental data, and extracting wind data from the target historical environmental data, and determining a target wind force change rate based on the wind data; inputting the historical environmental data and the target wind force change rate into a neural network model to obtain historical predicted power; training the neural network model based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

[0007] In an exemplary embodiment, the screening of the historical environmental data to obtain target historical environmental data includes: based on the station inventory data of the wind farm, performing an initial screening of the historical environmental data to obtain historical environmental data after initial screening; based on the operating parameters of the wind turbines in the wind farm, performing a re-screening of the historical environmental data after initial screening to obtain historical environmental data after re-screening; and removing abnormal data in the historical environmental data after re-screening to obtain target historical environmental data.

[0008] In an exemplary embodiment, the target historical environmental data includes at least wind data within a preset historical time period; extracting wind data from the target historical environmental data and determining the target wind change rate based on the wind data include: extracting wind data of multiple historical moments from the wind data within the preset historical time period; for any one of the multiple historical moments, determining the historical moment as the current historical moment; determining the current wind change rate of the current historical moment based on the wind data of the current historical moment and the wind data of the previous moment of the current historical moment; and determining the target wind change rate based on the current wind change rate of the current historical moment.

[0009] In an exemplary embodiment, the wind data includes wind speed data and wind direction data; determining the current wind force change rate of the current historical moment based on the wind force data of the current historical moment and the wind force data of the previous moment of the current historical moment includes: performing differential processing based on the wind direction data of the current historical moment and the wind direction data of the previous moment of the current moment to obtain the wind direction change rate of the current historical moment, and performing differential processing based on the wind speed data of the current historical moment and the wind speed data of the previous moment of the current moment to obtain the wind speed change rate of the current historical moment; determining the wind direction change rate of the current historical moment and the wind direction change rate as the current wind force change rate of the current historical moment.

[0010] In an exemplary embodiment, before inputting the historical environmental data and the target wind speed change rate into the neural network model, the steps include: determining a hyperparameter range of the neural network model; determining a hyperparameter from the hyperparameter range, and processing the neural network model based on the hyperparameter to determine performance parameters of the neural network model; reselecting new hyperparameters from the hyperparameter range based on CFOA technology and the performance parameters, and returning to the step of processing the neural network model based on the hyperparameters to determine performance parameters of the neural network model based on the new hyperparameters, and stopping when a first preset condition is met to obtain the target hyperparameters; and setting the neural network model based on the target hyperparameters.

[0011] In an exemplary embodiment, the neural network model is trained based on the historical predicted power to obtain a power prediction model, including: obtaining the historical operating power of the wind turbines in the wind farm; determining the power loss based on the historical operating power and the historical predicted power; and performing gradient feedback on the power loss to update the model parameters of the neural network model, until a second preset condition is met, to obtain a power prediction model.

[0012] In an exemplary embodiment, the method also includes: determining the real-time predicted power of the wind turbine at multiple preset moments predicted by the power prediction model; comparing the real-time predicted power at the multiple preset moments with their respective true powers to determine the power errors corresponding to the multiple preset moments; and determining that the wind turbine has a fault when the power errors corresponding to the multiple preset moments meet the set abnormal conditions.

[0013] According to another embodiment of the embodiment of the present application, a power prediction device for a wind turbine is also provided, including: an acquisition module for acquiring historical environmental data of a wind farm; a screening module for screening the historical environmental data to obtain target historical environmental data, and extracting wind data from the target historical environmental data, and determining a target wind force change rate based on the wind data; an input module for inputting the historical environmental data and the target wind force change rate into a neural network model to obtain historical predicted power; a training module for training the neural network model based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

[0014] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above method when running.

[0015] According to another aspect of the embodiments of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the method through the computer program.

[0016] According to another aspect of the embodiments of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0017] In an embodiment of the present application, the historical environmental data obtained from the wind farm station is screened to obtain the target historical environmental data, thereby avoiding the negative impact of abnormal data on model training, thereby improving the prediction accuracy of the model. On the basis of obtaining the target historical environmental data, the wind data is further extracted, and its target wind change rate is calculated. The introduction of this feature takes into account the dynamic change characteristics of the wind data of the wind turbine, so that the model can capture the correlation in the time series and enhance the sensitivity and adaptability of the model to real-time environmental changes. Further, by taking the screened target historical environmental data and the target wind change rate as input to train the neural network model, the power prediction model obtained can make real-time predictions according to the dynamic changes of the real environment of the wind farm. This dynamic model training based on the target historical environmental data and the target wind change rate enables the model to better adapt to the real-time environment of the wind farm and improve the timeliness and accuracy of subsequent real-time power detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] Figure 1 It is a hardware structure block diagram of a computer terminal of a wind turbine power prediction method according to an embodiment of the present application;

[0020] Figure 2 is a flow chart of a method for generating a power prediction method of a wind turbine generator according to an embodiment of the present application;

[0021] Figure 3 is a flowchart of a power prediction method for a wind turbine according to an embodiment of the present application;

[0022] Figure 4 It is a structural block diagram of a power prediction device for a wind turbine according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices; "plurality" means two or more.

[0025] The method embodiments provided in the embodiments of the present application can be executed in a computer terminal or similar computing device or cloud platform or independent physical server or software platform, wherein the above software platform is run through one or more servers. Taking running on a computer terminal as an example, Figure 1 1 is a hardware structure diagram of a power prediction method for a wind turbine according to an embodiment of the present application. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown in the figure) processor 102 and memory 104 for storing data. In an exemplary embodiment, the computer terminal may also include a transmission device 106 for communication functions and an input and output device 108, wherein the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA. A person skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Equivalent functions or comparisons shown Figure 1 A different configuration with more features is shown.

[0026] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as computer programs corresponding to the methods in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, that is, the above-mentioned methods are implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0027] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0028] In this embodiment, a power prediction method for a wind turbine is provided, which is applied to the above-mentioned computer device. Figure 2 : is a flow chart of a method for predicting power of a wind turbine according to an embodiment of the present application, the flow chart comprising the following steps:

[0029] Step S202, obtaining historical environmental data of the wind farm;

[0030] It should be noted that historical environmental data mainly comes from various sensors and monitoring equipment in wind farms. These devices usually include anemometers, wind vanes, temperature sensors, and humidity sensors. The data covers key environmental parameters such as wind speed, wind direction, temperature, and humidity in wind farms.

[0031] Historical environmental data is usually collected in real time in the form of digital signals and stored in the database system of the wind farm station. The data record can contain a timestamp to record the collection time of each data point, which is crucial for analyzing the dynamic changes of environmental parameters.

[0032] In some embodiments, in order to capture the rapid changes in environmental parameters such as wind speed and wind direction, the data collection frequency is usually high, such as once per minute or more. The collection cycle covers a variety of operating conditions of wind turbines, including but not limited to low wind speed, high wind speed, different wind directions, and seasonal changes, to ensure that the model can fully learn and understand the operating characteristics of the equipment under various environmental conditions.

[0033] Step S204, screening the historical environmental data to obtain target historical environmental data, extracting wind data from the target historical environmental data, and determining a target wind change rate based on the wind data;

[0034] Step S206, inputting the historical environmental data and the target wind speed change rate into a neural network model to obtain historical predicted power.

[0035] Step S208, training the neural network model based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

[0036] In the above steps S202-S208, the target historical environmental data is obtained from the historical environmental data obtained from the wind farm station by screening the historical environmental data, thereby avoiding the negative impact of abnormal data on model training, thereby improving the prediction accuracy of the model. On the basis of obtaining the target historical environmental data, the wind data is further extracted, and its target wind change rate is calculated. The introduction of this feature takes into account the dynamic change characteristics of the wind data of the wind turbine, so that the model can capture the correlation in the time series and enhance the sensitivity and adaptability of the model to real-time environmental changes. Further, by taking the screened target historical environmental data and the target wind change rate as input to train the neural network model, the power prediction model obtained can make real-time predictions according to the dynamic changes of the real environment of the wind farm. This dynamic model training based on the target historical environmental data and the target wind change rate enables the model to better adapt to the real-time environment of the wind farm and improve the timeliness and accuracy of subsequent real-time power detection.

[0037] In an exemplary embodiment, the screening of the historical environmental data to obtain target historical environmental data includes: based on the station inventory data of the wind farm, performing an initial screening of the historical environmental data to obtain historical environmental data after initial screening; based on the operating parameters of the wind turbines in the wind farm, performing a re-screening of the historical environmental data after initial screening to obtain historical environmental data after re-screening; and removing abnormal data in the historical environmental data after re-screening to obtain target historical environmental data.

[0038] It is understandable that based on the initial screening of the station ledger data, data records that are obviously not in normal operating status (such as shutdown, power outage and other abnormal operating status) are eliminated to ensure that the data set used for model training reflects the environmental data of wind turbines under normal operating conditions. The station ledger data contains the operation log of the wind farm, including equipment maintenance records, fault records, downtime, etc. By analyzing this information, it is possible to identify the time intervals when wind turbines are in abnormal operating conditions, such as downtime caused by maintenance, failures or power supply problems. For time periods marked as abnormal operating conditions in the station ledger, the corresponding historical environmental data should be excluded from the data set to avoid these abnormal conditions from misleading model training.

[0039] Further, according to the technical specifications provided by the manufacturer of the wind turbine, the operating parameters of the wind turbine are determined, such as the cut-in wind speed (i.e., the minimum wind speed for starting power generation) and the cut-out wind speed (i.e., the maximum wind speed for stopping power generation). For most wind turbines, the cut-in wind speed is usually around 3m / s, while the cut-out wind speed is around 25m / s. Based on the operating parameters of the wind turbines in the wind farm station, it is ensured that the environmental data in the data set meets the normal operating conditions of the wind turbines, and invalid data caused by too low or too high wind speeds are eliminated. That is, all data records with wind speeds lower than the cut-in wind speed or higher than the cut-out wind speed are eliminated, because these records are either the equipment is not started or shut down due to excessive wind speed, and its power output is not of analytical value.

[0040] Finally, for the data points marked as outliers, delete or correct them to ensure the purity of the data used for training. Removing the outliers in the historical environmental data after the re-screening to obtain the target historical environmental data may include: removing the outliers that may exist in the historical environmental data after the re-screening to ensure the purity of the data set and avoid interference during the model training process. Outliers usually refer to data points that are obviously deviated from the normal range or distribution. Statistical methods, such as the quartile method or the Z score method, can be used to identify and mark outliers. And check whether there are missing values ​​in the data set and handle them according to the specific situation. For time series data, such as wind speed, temperature, etc., the average value of the previous and next time steps can be used to fill in to maintain the continuity of the data. If there are too many missing values, consider deleting the data of the entire time window. And remove constant values, which means that the data sent back by the sensor remains unchanged over a period of time, which may indicate that the sensor is faulty or in a static state. These constant value data should be eliminated to ensure that the model can learn the dynamic environmental change characteristics. For the data points marked as outliers, delete or correct them to ensure the purity of the data used for training.

[0041] In the above embodiment, high-quality target historical environmental data can be obtained through initial screening based on station ledger data, re-screening based on wind turbine operating parameters, and removal of abnormal data. This data set not only excludes data under abnormal operating conditions and invalid wind speed conditions, but also further improves the purity and integrity of the data through outlier detection and data cleaning. The final target historical environmental data will provide a solid data foundation for building a power prediction model and ensure the accuracy and reliability of model training.

[0042] In an exemplary embodiment, the target historical environmental data includes at least wind data within a preset historical time period; extracting wind data from the target historical environmental data and determining the target wind change rate based on the wind data include: extracting wind data of multiple historical moments from the wind data within the preset historical time period; for any one of the multiple historical moments, determining the historical moment as the current historical moment; determining the current wind change rate of the current historical moment based on the wind data of the current historical moment and the wind data of the previous moment of the current historical moment; and determining the target wind change rate based on the current wind change rate of the current historical moment.

[0043] The preset historical time period is a historical time window selected from the target historical environmental data that contains enough wind change information for model learning and prediction. Based on the prediction requirements and data characteristics of the model, a historical time period (for example, the last 24 hours, a week or longer) can be preset to ensure that the time period covers the operation data of wind turbines under different wind speeds, wind directions and environmental conditions. All wind data records within the preset time period are selected from the target historical environmental data set, including key indicators such as wind speed and wind direction.

[0044] Specifically, the wind data at each specific time point is extracted from the preset historical time period in order to calculate the wind change rate. The preset historical time period is divided into multiple historical moments with fixed time intervals (such as every minute). For each historical moment, the wind speed and wind direction data at that moment are obtained from the data set to construct a time series wind data set. The changes in wind data in the time series are analyzed, and the wind change rate is calculated to reflect the dynamic changes of wind speed and direction over time.

[0045] In some embodiments, any point in the time series is selected as the current historical moment, and the current historical moment usually used to calculate the rate of change will traverse the entire time series. Change rate calculation: For the selected current historical moment, the difference between its wind data and the wind data at the previous moment is used to calculate the wind change. The rate of change is further calculated, that is, the ratio of the wind change to the wind data at the previous moment, to reflect the instantaneous change in wind force. According to the wind change rate of all historical moments, the data points that have a significant impact on the model training are screened out as the target wind change rate for subsequent model training.

[0046] In some embodiments, the wind speed change rate at each historical moment is statistically analyzed to identify time points with significant changes. These time points usually correspond to rapid changes in wind speed or wind direction, which have a significant impact on the power output of wind turbines. Change rate selection: select data points whose absolute value of the change rate is greater than a preset threshold or whose change rate changes continuously within a certain period as target change rate data for training the model to enhance the model's sensitivity and predictive ability to wind changes.

[0047] In the above embodiment, by extracting wind data within a preset historical time period, calculating the wind change rate at each historical moment, and selecting wind data points with significant changes as target wind change rates, the model can more accurately capture the dynamic changes of wind over time and the impact of such changes on the power output of wind turbines. This process not only provides the dynamic features required for model training, but also helps the model better understand the laws of wind change, providing more accurate data support for power anomaly detection.

[0048] In an exemplary embodiment, the wind data includes wind speed data and wind direction data; determining the current wind force change rate of the current historical moment based on the wind force data of the current historical moment and the wind force data of the previous moment of the current historical moment includes: performing differential processing based on the wind direction data of the current historical moment and the wind direction data of the previous moment of the current moment to obtain the wind direction change rate of the current historical moment, and performing differential processing based on the wind speed data of the current historical moment and the wind speed data of the previous moment of the current moment to obtain the wind speed change rate of the current historical moment; determining the wind direction change rate of the current historical moment and the wind direction change rate as the current wind force change rate of the current historical moment.

[0049] It should be noted that wind data mainly includes wind speed data and wind direction data. These two data reflect the dynamic characteristics of the wind farm environment and have a direct impact on the operating status and power output of wind turbines. Wind speed data indicates the speed of the wind, usually in meters per second (m / s), while wind direction data indicates the direction of the wind, usually expressed in azimuth. Analyze the changes in wind direction data in time series and calculate the wind direction change rate to reflect the immediate changes in wind direction.

[0050] Specifically, the current historical moment t and the previous historical moment t-1 are two continuous sampling points in the time series. For wind direction data, the wind direction data of the current historical moment t and the wind direction data of the previous historical moment t-1 are obtained from the target historical environment data respectively. The difference between the wind direction data at the two moments is processed to obtain the wind direction change rate. It can be understood that the wind direction data is cyclic data, that is, 360 degrees and 0 degrees represent the same direction. Therefore, it is necessary to calculate the minimum difference angle to avoid misleading the calculation of the wind direction change rate.

[0051] In some embodiments, the change of wind speed data in time series is analyzed and the wind speed change rate is calculated to reflect the instant change of wind speed. Similarly, the wind speed data at the current historical moment t and the wind speed data at the previous historical moment t-1 are obtained from the target historical environment data. The difference between the wind speed data at the two moments is processed to obtain the wind speed change rate.

[0052] In a specific application, the wind direction change rate and wind speed change rate are combined to form a complete description of the wind speed change rate for model training and analysis. The calculated wind direction change rate and wind speed change rate at the current historical moment are combined to form a vector representing the wind speed change at that moment. This change rate vector can be used for further feature engineering, such as time series analysis, or directly as one of the input features of the model to help the model more comprehensively understand the operating status of wind turbines under different wind conditions.

[0053] In the above embodiment, by performing differential processing on the wind direction data and wind speed data at the current historical moment, respectively calculating the wind direction change rate and wind speed change rate, and combining the two into a current wind force change rate, we can more accurately capture the dynamic change characteristics of the wind farm in the time series. This process not only provides the dynamic environmental information required for model training, but also helps the model better understand the impact of wind force changes on the power output of wind turbines, providing a more accurate data basis for power anomaly detection. When processing the wind direction change rate, ensuring the correct calculation of the minimum difference angle and avoiding calculation errors caused by the particularity of the data nature is one of the key steps to ensure the accuracy of model prediction.

[0054] In an exemplary embodiment, before inputting the historical environmental data and the target wind speed change rate into the neural network model, the steps include: determining a hyperparameter range of the neural network model; determining a hyperparameter from the hyperparameter range, and processing the neural network model based on the hyperparameter to determine performance parameters of the neural network model; reselecting new hyperparameters from the hyperparameter range based on CFOA technology and the performance parameters, and returning to the step of processing the neural network model based on the hyperparameters to determine performance parameters of the neural network model based on the new hyperparameters, and stopping when a first preset condition is met to obtain target hyperparameters; and setting the neural network model based on the target hyperparameters.

[0055] Among them, CFOA technology refers to CFOA (Carnivore Flocking Optimization Algorithm).

[0056] It is understandable that determining the hyperparameter range of the neural network model can set a reasonable search space for the hyperparameters to be optimized, ensuring the feasibility and efficiency of the optimization process. For example, for each hyperparameter, a reasonable search range is set based on general experience and domain knowledge of model training. For example, the learning rate may be between 0.001 and 0.1, the batch size may be between 32 and 256, and so on. The hyperparameters that need to be optimized in the neural network model may include learning rate, batch size, number of hidden layers, number of neurons in each layer, and so on.

[0057] Specifically, a set of hyperparameter values ​​is randomly selected from the set hyperparameter range as the initial configuration of the neural network model. The current historical environmental data and wind speed change rate are used as input to train the neural network model based on the selected hyperparameters. After the training is completed, the performance of the model is evaluated using the validation set data. Common performance indicators include mean squared error (MSE), mean absolute error (MAE) or R-squared to measure the accuracy of model prediction. In the CFOA algorithm, the search agent represents multiple candidate solutions in the hyperparameter space. At initialization, a set of random hyperparameter values ​​is assigned to each search agent. Fitness evaluation: The model under each search agent configuration is trained and evaluated to obtain its fitness value, that is, the model performance indicator. Search and update: According to the search mechanism of CFOA, the search agent will update its own hyperparameter value based on the current fitness value and the location information of the global optimal solution and the worst solution. This process includes two phases in the CFOA algorithm: exploration and exploitation. The exploration phase focuses on global search, while the exploitation phase focuses on refining the local optimal solution. The CFOA algorithm will repeat the above search and update process until the preset number of iterations is met or the performance improvement threshold is reached. This process aims to find the hyperparameter combination that optimizes the model performance.

[0058] When the CFOA algorithm meets the first preset condition, such as reaching the maximum number of iterations or the performance index is no longer significantly improved, the current global optimal solution is selected as the target hyperparameter combination. Model configuration and training: Finally, the neural network model is configured using the target hyperparameters, and the final model training is performed to obtain a high-performance power anomaly detection model.

[0059] In the above embodiment, the hyperparameters of the neural network model are optimized by CFOA technology, which can automatically search and determine a set of optimal hyperparameter combinations, significantly improving the performance of the model in the task of detecting abnormal power of wind turbines. This process not only takes into account the global exploration of the hyperparameter space, but also focuses on the refinement of the local optimal solution, which can effectively avoid the problem of overfitting or underfitting of the model caused by improper hyperparameter configuration, and lays a solid foundation for subsequent model training and application.

[0060] In an exemplary embodiment, the neural network model is trained based on the historical predicted power to obtain a power prediction model, including: obtaining the historical operating power of the wind turbines in the wind farm; determining the power loss based on the historical operating power and the historical predicted power; and performing gradient feedback on the power loss to update the model parameters of the neural network model, until a second preset condition is met, to obtain a power prediction model.

[0061] Among them, the historical operating power is the real historical operating power data of wind turbines in the wind farm collected, which is used as the "true value" standard in the training process. The actual operating power data of wind turbines under the corresponding historical environmental conditions are extracted from the historical records of the wind farm station. By ensuring that the historical operating power data and the historical environmental data correspond one-to-one in time, accurate model training and performance evaluation can be carried out. The neural network model is used to predict the historical predicted power of wind turbines based on historical environmental data and wind speed change rate, and compared with the actual historical operating power to calculate the power loss.

[0062] Specifically, the processed and optimized historical environmental data (including wind speed, wind direction, temperature, etc.) and the target wind speed change rate are used as inputs, and forward propagation is performed through the neural network model to obtain the predicted historical predicted power. For each training sample, the difference between its historical predicted power and the actual historical operating power is calculated, and the power loss of all samples is summarized using the loss function to obtain the overall power loss value. According to the calculated power loss, the parameters of the neural network model are adjusted using the gradient descent method to reduce the prediction error.

[0063] In some embodiments, the back propagation algorithm can be used to calculate the gradient of the power loss function with respect to the model parameters (i.e., the partial derivative of the loss function with respect to the parameters). According to the calculated gradient, the optimization algorithm (such as stochastic gradient descent SGD, Adam, etc.) is used to update the model parameters and adjust them in the direction of reducing power loss. Repeat the process of model prediction, power loss calculation, and parameter update until the model converges or reaches a second preset condition such as the set number of training times. When the power loss reduction caused by the model parameter update is less than a certain threshold, or the preset number of training rounds is reached, the model training is stopped. Furthermore, the trained power prediction model can also be evaluated using a test data set to check the generalization ability of the model, that is, the prediction performance for unseen data. Evaluation indicators may include mean square error, mean absolute error, R square value, etc., to comprehensively measure the prediction accuracy of the model.

[0064] In the above embodiment, based on the historical predicted power and historical operating power, the neural network model is trained through gradient backpropagation and parameter iterative update to obtain a high-precision power prediction model. This process continuously adjusts the weights and biases of the model to minimize the power loss between the prediction and the actual until the model reaches the preset training termination condition. The obtained power prediction model can accurately predict the operating power of the wind turbine under specific environmental conditions, providing a key reference for power anomaly detection.

[0065] In an exemplary embodiment, the method also includes: determining the real-time predicted power of the wind turbine at multiple preset moments predicted by the power prediction model; comparing the real-time predicted power at the multiple preset moments with their respective true powers to determine the power errors corresponding to the multiple preset moments; and determining that the wind turbine has a fault when the power errors corresponding to the multiple preset moments meet the set abnormal conditions.

[0066] Among them, environmental data including but not limited to wind speed, wind direction and temperature are obtained from the real-time sensor data of the wind turbine. Feature extraction and preprocessing: Preprocess the real-time collected environmental data, including but not limited to outlier detection, feature normalization, etc., to ensure the quality of input data. The preprocessed real-time environmental data is input into the trained power prediction model to obtain the real-time predicted power at multiple preset times.

[0067] Further ensure that the real-time predicted power is aligned with the actual operating power data of the wind turbine in time for effective comparison. Power error calculation: For each preset moment, calculate the difference between the real-time predicted power and the actual operating power, that is, the power error. According to the operating characteristics and historical data of the equipment, set an abnormal condition, such as the power error exceeds a certain threshold for a certain period of time, or the statistical indicators such as the mean and standard deviation of the power error exceed the preset range. When it is detected that the power errors corresponding to multiple preset moments meet the abnormal conditions, it is judged that there is a potential fault in the wind turbine and corresponding measures are taken.

[0068] In some embodiments, the power error at each preset moment is compared with the set abnormal condition to determine whether there is an abnormality. If the power error at multiple consecutive moments exceeds the threshold of the abnormal condition, or the statistical characteristics of the power error are obviously abnormal, then it can be considered that the wind turbine may be in a faulty state. When it is determined that there is a fault in the wind turbine, the system should immediately issue an early warning signal. The early warning information should include the preset time when the fault occurs, the comparison result of the predicted power and the actual power, possible fault type suggestions, etc.

[0069] In the above embodiment, the operating environment data of the wind turbine is collected in real time, and the trained power prediction model is used for prediction, and compared with the actual operating power to detect whether the power error meets the abnormal conditions. We can timely discover the fault state of the wind turbine and improve the stability and efficiency of the equipment operation. This process combines data-driven prediction technology and real-time monitoring mechanism, providing strong technical support for the intelligent operation and maintenance of wind power systems.

[0070] In order to better understand the process of the power prediction method of the above-mentioned wind turbine, the power prediction method of the above-mentioned wind turbine is described below in combination with an optional embodiment, but it is not used to limit the technical solution of the embodiment of the present application.

[0071] refer to Figure 3 As shown, it is a flowchart of the power prediction method of the wind turbine generator of the present application: first, a prediction model is generated, that is, a process of a power prediction model, including:

[0072] Wind turbine historical operation data collection: Obtain the external environmental factors that affect wind turbine power from the historical operation data of wind farms, such as wind speed, temperature, air humidity, etc., as well as wind turbine operation data related characteristics, such as equipment operating status, power, etc. To ensure the model effect, select data with a sampling interval of 1 minute.

[0073] For the establishment of machine learning models, high-quality data is the guarantee of model performance, so it is necessary to preprocess the data and construct features: perform preliminary screening based on the station ledger data, exclude data in abnormal normal operation states such as shutdown and power outage, and further screen based on the cut-in wind speed and cut-out wind speed of the equipment. For example, when the cut-in wind speed of the fan is about 3m / s and the cut-out wind speed is about 25m / s, the data outside the wind speed range can be excluded. The specific cut-in and cut-out wind speeds need to be set according to the equipment model and rated parameters. Since the data collection and transmission process is easily affected by factors such as electromagnetic interference and acquisition equipment failure, there may be some abnormal data in the available data, which requires further processing. Common abnormal data can be divided into missing values, constant values ​​and outliers. Since variables such as wind speed and temperature are usually continuous variables that change over time, the missing values ​​are filled according to the average value of the previous and next time steps. If there are a large number of missing values ​​in a continuous time, the data during this period is deleted. Constant value means that the data sent back by the sensor remains constant over a period of time. In order to avoid affecting model training, this part of the data also needs to be eliminated. Outliers refer to values ​​that clearly exceed the range of characteristic data, as well as outliers, which can be cleaned up using the quartile method.

[0074] Considering that the power of the wind turbine at the current moment is not only related to the wind speed at the current moment, but also may be related to the wind speed at the previous moment due to the influence of factors such as inertia and control system, new features need to be constructed. After data preprocessing, the wind speed at the current moment and the wind speed at the previous moment are differentially calculated. Similarly, the same operation is performed on the wind direction angle feature to obtain the wind speed change rate and wind direction change rate per unit time as part of the input feature.

[0075] Furthermore, an MLP (Multilayer Perceptron) time series prediction model is built, that is, a wind turbine power prediction model based on MLP is established. MLP is a feedforward neural network model that can establish a mapping between a set of input vectors and a set of output vectors. It is very effective in solving more complex regression problems. MLP is a network with multiple layer structures, which can be divided into input layer, hidden layer and output layer according to the function of the layer. The input layer is responsible for receiving input features, the hidden layer is responsible for extracting features from the input, and the output layer is responsible for integrating the features of the hidden layer and generating the final prediction value. Each level of MLP contains several neuron nodes. These neuron nodes are the most basic units of the model and are responsible for calculating features. Each neuron node consists of a linear combination part and a nonlinear activation part. For a neuron node, given an input feature vector x=[x1,x2,…,x n ], the output of the neuron can be expressed as:

[0076]

[0077] Among them, w i represents the weight of the neuron, b represents the bias term, and f represents the activation function. The present invention uses the commonly used linear rectifier unit (RELU) as the activation function, and the formula is expressed as f(x)=max(0,x). This function has the advantages of simple calculation and fast speed.

[0078] Before model training, it is necessary to optimize hyperparameters based on CFOA. CFOA is a meta-heuristic optimization algorithm that divides the optimization of parameter selection into initialization, exploration, and development stages. It explores the global situation by continuously updating the position of the search agent and ultimately finds the best solution. In the initialization stage, a d-dimensional matrix consisting of N search agent position information is defined. The initialization formula is as follows:

[0079] Fisher i,j =(ub j -lb j )×r+lb j

[0080] Among them, Fisher i,j represents the position of the i-th search agent in the j-th dimension, ub j and lb j is the upper and lower bounds of the jth dimension, and r is a random number between (0, 1). Based on the current position information of each search agent, the fitness value evaluation function is used to calculate the fitness value matrix.

[0081] In the exploration phase, the capture rate parameter α is defined as:

[0082]

[0083] Where EFs represents the number of current evaluations, and MaxEFs represents the maximum number of evaluations. Define a random number p between (0, 1). When p < α, the search agent position matrix is ​​updated according to the following formula:

[0084]

[0085] Among them, fit max and fit min Respectively represent the worst fitness value and the best fitness value after T position updates, Dis represents the Euclidean distance between the i-th individual and the reference, T represents the number of iterations, s is a random unit vector of dimension d, representing the direction and distance of the search agent's movement, and R represents the exploration range. When p ≥ α, the search agent position matrix is ​​updated as follows:

[0086]

[0087]

[0088] Among them, c represents a group of 3-4 individuals whose positions have not been updated, Centre c is the target point surrounded by the c group, r2 represents the speed at which the search agent approaches the center point, and r3 represents the offset of the movement.

[0089] In the development phase, the population variance is defined as:

[0090]

[0091] The update formula is as follows:

[0092]

[0093] Among them, GD is a Gaussian distribution function with a mean of 0 and a variance of σ, and Gbest represents the global optimal position.

[0094] The hyperparameter optimization process is as follows: 1. Initialize the search agent. 2. Calculate the fitness value and update the best position and fitness. 3. Determine whether it is currently in the exploration phase or the development phase. And update the parameters according to the corresponding formula. 4. If the termination condition is not met, repeat 2-3. 5. Output the best position and fitness value.

[0095] The hyperparameters are selected through the CFOA algorithm, and the model is trained and saved. After the model converges or reaches the maximum number of training times, the final wind turbine power prediction model is obtained.

[0096] During the training process, normal equipment operation data that has been manually screened can be used for model training. First, the data set needs to be split into a training set and a validation set in a ratio of 8:2, and then the model is trained. The calculation process of the model includes two stages: forward propagation and back propagation. Forward propagation refers to the process of transferring data from the input layer to the output layer layer by layer. For a network with L hidden layers, the linear combination output of the lth layer can be expressed as z (l) =W (l) a (l-1) +b (l) , where W (l) represents the weight matrix of the lth layer, b (l) is the bias vector of the lth layer, a (l-1) represents the output from the l-1th layer. After the activation function is processed, the output of the lth layer can be expressed as a (l) =f(z (l) ).

[0097] The back propagation process calculates the model loss based on the predicted value and the true value obtained by the forward propagation, and updates the model weights and biases based on the loss value. For regression tasks, the mean square error loss function can be used to obtain the model loss under the current parameters. For a data set containing N training samples, the mean square error loss can be expressed as where y i and They represent the true value and predicted value corresponding to the input features respectively.

[0098] After obtaining the loss function, calculate the gradient of the loss for each layer weight and update the weight parameters by gradient descent. (l) The gradient of can be expressed as:

[0099]

[0100] Among them, δ (l) is the error term of the lth layer and can be calculated recursively using the chain rule:

[0101]

[0102] For the output layer, It can be expressed as:

[0103]

[0104] For the hidden layer, It can be calculated by the error term of the next layer:

[0105]

[0106] After continuous iterations of forward propagation and back propagation, the model can learn the mapping relationship between input features and final output, and make predictions for new inputs.

[0107] In the power anomaly detection part, it includes the acquisition of the operating data of the unit to be detected, that is, the acquisition of data such as wind speed, wind direction, temperature, etc. collected on site, and further development of wind turbine power anomaly detection, that is, the acquired data is input into the power prediction model to obtain the power prediction value under normal operating conditions. The residual between it and the actual power obtained by the sensor is calculated. When the residual between the predicted value and the true value exceeds the threshold within a period of time, it is considered that there is a power anomaly in the wind turbine equipment, and an early warning message is output. The residual threshold and the equipment anomaly duration threshold are set according to the model's prediction performance on normal equipment.

[0108] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0109] The present application also provides a structural block diagram of a power prediction device for a wind turbine generator. Figure 4 is a structural block diagram of a power prediction device for a wind turbine according to an embodiment of the present application; Figure 4 As shown, including:

[0110] An acquisition module 42 is used to acquire historical environmental data of a wind farm;

[0111] A screening module 44 is used to screen the historical environmental data to obtain target historical environmental data, extract wind data from the target historical environmental data, and determine a target wind change rate based on the wind data;

[0112] An input module 46, used for inputting the historical environmental data and the target wind speed change rate into a neural network model to obtain a historical predicted power;

[0113] The training module 48 is used to train the neural network model based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

[0114] Through the above-mentioned device, the historical environmental data obtained from the wind farm station is screened to obtain the target historical environmental data, thereby avoiding the negative impact of abnormal data on model training, thereby improving the prediction accuracy of the model. On the basis of obtaining the target historical environmental data, the wind data is further extracted, and its target wind change rate is calculated. The introduction of this feature takes into account the dynamic change characteristics of the wind data of the wind turbine, so that the model can capture the correlation in the time series and enhance the sensitivity and adaptability of the model to real-time environmental changes. Further, by taking the screened target historical environmental data and the target wind change rate as input to train the neural network model, the power prediction model obtained can make real-time predictions according to the dynamic changes of the real environment of the wind farm. This dynamic model training based on the target historical environmental data and the target wind change rate enables the model to better adapt to the real-time environment of the wind farm and improve the timeliness and accuracy of subsequent real-time power detection.

[0115] In an exemplary embodiment, the screening module 44 is also used to perform a preliminary screening of the historical environmental data based on the station inventory data of the wind farm station to obtain the historical environmental data after the preliminary screening; based on the operating parameters of the wind turbines in the wind farm station, perform a re-screening of the historical environmental data after the preliminary screening to obtain the historical environmental data after the re-screening; remove abnormal data in the historical environmental data after the re-screening to obtain the target historical environmental data.

[0116] In an exemplary embodiment, the target historical environmental data includes at least wind data within a preset historical time period; the screening module 44 is further used to extract wind data of multiple historical moments from the wind data within the preset historical time period; for any one of the multiple historical moments, the historical moment is determined as the current historical moment; based on the wind data of the current historical moment and the wind data of the previous moment of the current historical moment, the current wind change rate of the current historical moment is determined; and according to the current wind change rate of the current historical moment, the target wind change rate is determined.

[0117] In an exemplary embodiment, the wind force data includes wind speed data and wind direction data; the screening module 44 is also used to perform differential processing based on the wind direction data at the current historical moment and the wind direction data at the previous moment of the current moment to obtain the wind direction change rate at the current historical moment, and to perform differential processing based on the wind speed data at the current historical moment and the wind speed data at the previous moment of the current moment to obtain the wind speed change rate at the current historical moment; the wind direction change rate at the current historical moment and the wind direction change rate are determined as the current wind force change rate at the current historical moment.

[0118] In an exemplary embodiment, the apparatus further comprises a setting module; the setting module is configured to determine a hyperparameter range of the neural network model;

[0119] Determine hyperparameters from the hyperparameter range, and process the neural network model based on the hyperparameters to determine performance parameters of the neural network model; re-select new hyperparameters from the hyperparameter range based on CFOA technology and the performance parameters, and return to the step of processing the neural network model based on the hyperparameters to determine the performance parameters of the neural network model based on the new hyperparameters, and stop when a first preset condition is met to obtain target hyperparameters; set the neural network model based on the target hyperparameters.

[0120] In an exemplary embodiment, the training module 48 is also used to obtain the historical operating power of the wind turbines in the wind farm; determine the power loss based on the historical operating power and the historical predicted power; perform gradient feedback on the power loss to update the model parameters of the neural network model until a second preset condition is met, thereby obtaining a power prediction model.

[0121] In an exemplary embodiment, the device also includes a detection module; the detection module is used to determine the real-time predicted power of the wind turbine at multiple preset moments predicted by the power prediction model; compare the real-time predicted power at the multiple preset moments with their respective actual powers to determine the power errors corresponding to the multiple preset moments; when the power errors corresponding to the multiple preset moments meet the set abnormal conditions, it is determined that the wind turbine has a fault.

[0122] An embodiment of the present application further provides a storage medium, which includes a stored program, wherein the program executes any of the above methods when it is run.

[0123] Optionally, in this embodiment, the storage medium may be configured to store program codes for executing the following steps:

[0124] S1, obtain historical environmental data of wind farms;

[0125] S2, screening the historical environmental data to obtain target historical environmental data, extracting wind data from the target historical environmental data, and determining a target wind change rate based on the wind data;

[0126] S3, inputting the historical environmental data and the target wind speed change rate into a neural network model to obtain historical predicted power;

[0127] S4, training the neural network model based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

[0128] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0129] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0130] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0131] S1, obtain historical environmental data of wind farms;

[0132] S2, screening the historical environmental data to obtain target historical environmental data, extracting wind data from the target historical environmental data, and determining a target wind change rate based on the wind data;

[0133] S3, inputting the historical environmental data and the target wind speed change rate into a neural network model to obtain historical predicted power;

[0134] S4, training the neural network model based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

[0135] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.

[0136] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0137] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0138] An embodiment of the present application also provides a computer program, which includes computer instructions stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in any one of the above method embodiments.

[0139] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0140] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0141] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting power of a wind turbine, characterized in that: include: Obtain historical environmental data of wind farms; Screening the historical environmental data to obtain target historical environmental data, extracting wind data from the target historical environmental data, and determining a target wind speed change rate based on the wind data; Inputting the historical environmental data and the target wind speed change rate into a neural network model to obtain historical predicted power; The neural network model is trained based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

2. The method according to claim 1, characterized in that The screening of the historical environment data to obtain target historical environment data includes: Based on the station account data of the wind farm, the historical environmental data is preliminarily screened to obtain the preliminarily screened historical environmental data; Based on the operating parameters of the wind turbines in the wind farm, the historical environmental data after the initial screening is screened again to obtain the historical environmental data after the secondary screening; Abnormal data in the historical environmental data after the re-screening is removed to obtain target historical environmental data.

3. The method according to claim 1, characterized in that The target historical environmental data at least includes wind data within a preset historical time period; The step of extracting wind data from the target historical environmental data and determining a target wind change rate based on the wind data includes: Extracting wind data at multiple historical moments from the wind data within the preset historical time period; For any one of the multiple historical moments, determining the historical moment as the current historical moment; Determine the current wind force change rate at the current historical moment based on the wind force data at the current historical moment and the wind force data at a moment before the current historical moment; The target wind force change rate is determined according to the current wind force change rate at the current historical moment.

4. The method according to claim 3, characterized in that The wind data includes wind speed data and wind direction data; The determining the current wind force change rate at the current historical moment based on the wind force data at the current historical moment and the wind force data at the previous moment before the current historical moment comprises: Performing differential processing based on the wind direction data at the current historical moment and the wind direction data at the moment before the current moment to obtain the wind direction change rate at the current historical moment, and performing differential processing based on the wind speed data at the current historical moment and the wind speed data at the moment before the current moment to obtain the wind speed change rate at the current historical moment; The wind direction change rate and the wind direction change rate at the current historical moment are determined as the current wind force change rate at the current historical moment.

5. The method according to claim 1, characterized in that Before inputting the historical environmental data and the target wind speed change rate into the neural network model, the method includes: Determining hyperparameter ranges of the neural network model; Determining a hyperparameter from the hyperparameter range, and processing the neural network model based on the hyperparameter to determine a performance parameter of the neural network model; Reselecting new hyperparameters from the hyperparameter range based on the CFOA technology and the performance parameters, and returning to process the neural network model based on the hyperparameters based on the new hyperparameters to determine the performance parameters of the neural network model, and stopping when a first preset condition is met to obtain target hyperparameters; The neural network model is set based on the target hyperparameters.

6. The method according to claim 1, characterized in that The step of training the neural network model based on the historical predicted power to obtain a power prediction model includes: Obtaining historical operating power of wind turbines in the wind farm; determining power loss based on the historical operating power and the historical predicted power; The power loss is gradient-backed to update the model parameters of the neural network model until a second preset condition is met, thereby obtaining a power prediction model.

7. The method according to claim 1, characterized in that The method further comprises: Determining the real-time predicted power of the wind turbine at a plurality of preset moments predicted by the power prediction model; Compare the real-time predicted powers at the plurality of preset moments with their respective actual powers to determine power errors corresponding to the plurality of preset moments; When the power errors respectively corresponding to the plurality of preset moments meet the set abnormal conditions, it is determined that the wind turbine generator has a fault.

8. A power prediction device for a wind turbine, characterized in that: include: An acquisition module is used to obtain historical environmental data of wind farms; A screening module, used for screening the historical environmental data to obtain target historical environmental data, extracting wind data from the target historical environmental data, and determining a target wind change rate based on the wind data; An input module, used for inputting the historical environmental data and the target wind speed change rate into a neural network model to obtain historical predicted power; A training module is used to train the neural network model based on the historical predicted power to obtain a power prediction model; wherein the power prediction model is used to predict the real-time predicted power of the wind turbine.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method described in any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images