Deep learning-based fan EFC evaluation method and system
Through the deep learning-based fan EFC evaluation method, the existing technology of the problem of low prediction accuracy and poor adaptability of wind power is solved, and higher prediction accuracy and system stability are achieved, and intelligent operation and maintenance decision-making is supported.
Patent Information
- Application Number
- CN202411327262.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-23
- Publication Date
- 2025-02-18
AI Technical Summary
The existing fan EFC evaluation methods mainly rely on empirical formulas and manual judgments, with low prediction accuracy and poor adaptability, making it difficult to cope with complex and changeable operating environments.
Using the deep learning-based fan EFC evaluation method, by collecting wind power related data of wind farms, missing value processing and normalization processing, extracting key features and clustering, building a Seq2Seq neural network model for wind power evaluation, and according to The evaluation results provide corresponding operation and maintenance measures.
It improves the accuracy of wind power power prediction, enhances the stability and efficiency of the system, realizes the intelligence of operation and maintenance decisions, and has good scalability and adaptability.
Smart Images
Figure CN119474940A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wind power energy management and optimization, and in particular to a wind turbine EFC evaluation method and system based on deep learning. Background Art
[0002] As the global demand for renewable energy continues to grow, wind energy, as a clean and renewable energy form, is becoming increasingly important in the energy structure. As the main form of wind energy utilization, the efficient and stable operation of wind farms is of great significance to ensuring energy supply and reducing environmental pollution. However, wind farms face many challenges in actual operation, such as wind speed fluctuations, wind direction changes, climate impacts, etc. These factors lead to significant uncertainty and volatility in wind power output. The wind turbine EFC (Energy Flow Control) system is a key component in a wind farm, responsible for monitoring and controlling the operating status of the wind turbine, optimizing energy flow, and improving the overall efficiency and stability of the wind power system.
[0003] Traditional wind turbine EFC evaluation methods mainly rely on empirical formulas and manual judgment. These methods often have problems such as low prediction accuracy, poor adaptability, and difficulty in coping with complex and changing operating environments. Summary of the invention
[0004] The purpose of the present invention is to provide a wind turbine EFC evaluation method and system based on deep learning, aiming to solve the problem that the existing wind turbine EFC evaluation method mainly relies on empirical formulas and manual judgment, has low prediction accuracy and poor adaptability, and is difficult to cope with complex and changeable operating environments.
[0005] The present invention is achieved through the following technical solutions:
[0006] A wind turbine EFC evaluation method based on deep learning includes the following steps:
[0007] Collect wind power related data of wind farms and obtain initial data sets;
[0008] The initial data set is processed with missing values and normalized to obtain the preprocessed data set;
[0009] Extract key features from the preprocessed data set, and cluster the extracted key features to obtain a key feature data set;
[0010] Based on the key feature data set, a wind turbine EFC evaluation model is constructed through a neural network;
[0011] The wind power is evaluated through the wind turbine EFC evaluation model, and corresponding operation and maintenance measures are given according to the evaluation results.
[0012] Optionally, the wind power related data of the wind farm includes: wind power real-time monitoring data, wind power historical data and numerical weather forecast data.
[0013] Optionally, the specific process of performing missing value processing on the initial data set is:
[0014] The weighted moving average method is used to process missing values in the initial data set. The expression is shown in the following formula (1):
[0015]
[0016] Among them, M i is the weighted moving average of time t, ω i for y t-i+1 The weight, Y t-i+1 is the observed value at time t-i+1, and N is the number of moving average items.
[0017] Optionally, the specific process of normalizing the initial data set is:
[0018] The maximum and minimum normalization method is used to normalize the initial data set, and the expression is shown in the following formula (2):
[0019]
[0020] Among them, W ′ is the normalized value, W is the original data in the initial data set, and W min and W max are the minimum and maximum values in the original data, respectively.
[0021] Optionally, the specific process of extracting key features from the preprocessed data set and clustering the extracted key features to obtain a key feature data set is:
[0022] Extract key features from the preprocessed data set, including wind speed, wind direction, temperature, air pressure, humidity, and fan operating status;
[0023] Perform cluster analysis on key features and construct several cluster feature vectors based on the cluster analysis results. Each cluster feature vector represents the wind farm status and wind turbine operation status at a time point.
[0024] The K-means clustering algorithm is used to cluster several cluster feature vectors and obtain the key feature data set.
[0025] Optionally, the specific process of constructing the wind turbine EFC evaluation model through a neural network based on the key feature data set is:
[0026] The initial wind turbine EFC evaluation model was constructed using the Seq2Seq neural network model;
[0027] Based on the key feature data set, the initial wind turbine EFC evaluation model is trained and the wind turbine EFC evaluation model is obtained;
[0028] In the process of training the initial wind turbine EFC evaluation model, the parameters of the initial wind turbine EFC evaluation model are updated through the Adam optimizer to minimize the mean square error loss function.
[0029] Optionally, the wind power is evaluated by using a wind turbine EFC evaluation model, the mean absolute error and the root mean square error are calculated according to the evaluation results, and the mean absolute error and the root mean square error are used as evaluation indicators of the wind turbine EFC performance.
[0030] Optionally, the expressions of the mean absolute error and the root mean square error are respectively shown in the following formula (3) and formula (4):
[0031]
[0032] Where n is the number of samples, y i is the true value, is the predicted value, M E is the mean absolute error, R E is the root mean square error.
[0033] Optionally, the specific process of providing corresponding operation and maintenance measures according to the evaluation results is:
[0034] Evaluate the stability and efficiency of the current wind power system based on the prediction results output by the wind turbine EFC evaluation model;
[0035] If the prediction error exceeds the preset threshold, the wind turbine EFC performance is judged to be poor and the following operation and maintenance measures need to be taken:
[0036] Adjust fan parameters: According to the actual value of the prediction error, adjust the key parameters of the fan to optimize energy flow and reduce the prediction error;
[0037] Optimize wind farm layout: If the prediction errors of at least two wind turbines exceed the preset threshold, the layout of the wind farm is optimized;
[0038] Early warning and maintenance: Based on the prediction results, early warning of potential operating failures is provided, and targeted maintenance and repair work is arranged;
[0039] Data analysis and feedback: Feedback the prediction results and operation and maintenance measures data to the deep learning model to continuously optimize and update the model.
[0040] The present invention also provides a wind turbine EFC evaluation system based on deep learning, which is used to implement the wind turbine EFC evaluation method based on deep learning, comprising:
[0041] Data acquisition module: used to collect wind power related data;
[0042] Data preprocessing module: used to preprocess the initial data set;
[0043] Feature extraction and clustering module: used to extract key features from the preprocessed data set, cluster the features through clustering algorithms, and construct a key feature data set;
[0044] Model building module: used to build the wind turbine EFC evaluation model and optimize the model parameters;
[0045] Evaluation and decision-making module: used to evaluate the EFC of wind turbines and provide corresponding operation and maintenance measures based on the evaluation results;
[0046] Data storage and management module: used to store and manage various types of data generated by the system, and supports data query, export and backup;
[0047] User interaction module: used for users to view prediction results, performance indicators and operation and maintenance suggestions, and perform corresponding interactive operations.
[0048] The technical solution of the present invention has at least the following advantages and beneficial effects:
[0049] Improve the accuracy of wind power forecasting: Through the application of deep learning models, the time dependence and nonlinear relationships in wind power data can be captured more accurately, thereby improving the accuracy of wind power forecasting and helping wind farms to better manage and dispatch energy.
[0050] Enhance system stability and efficiency: Through real-time evaluation and prediction of wind power, potential problems of the wind power system, such as performance degradation and efficiency reduction, can be discovered in a timely manner, so that corresponding operation and maintenance measures can be taken; these measures include adjusting wind turbine parameters, optimizing wind farm layout, etc., which help to improve the overall stability and operating efficiency of the wind power system.
[0051] Intelligent operation and maintenance decision-making: The present invention realizes intelligent operation and maintenance decision-making through data analysis and feedback mechanism; the system can automatically evaluate the performance of the wind power system based on the prediction results, and give specific operation and maintenance suggestions, such as adjusting wind turbine parameters, early warning maintenance, etc.; it not only improves the operation and maintenance efficiency, but also reduces the risk of human error.
[0052] Strong data processing capabilities: The system uses a variety of data processing technologies, including missing value processing, normalization processing, feature extraction and clustering, which effectively improves the quality and availability of data; these technologies can deal with common problems such as missing values and outliers in wind power data, ensuring the data quality of model training.
[0053] Scalability and adaptability: The wind turbine EFC evaluation model constructed by the present invention has good scalability and adaptability; with the continuous development of wind power technology and the increase in data volume, the system can easily expand the model scale and improve the prediction accuracy; at the same time, the system can also continuously optimize and update the model according to new data feedback to adapt to changes in wind farms.
[0054] User-friendliness: The system provides a user interaction module that allows users to easily view prediction results, performance indicators and operation and maintenance suggestions, and perform corresponding interactive operations; it improves the system's ease of use and operability, and helps wind farm managers better use the system for decision-making and management. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] Figure 1 Schematic diagram of the flow of the wind turbine EFC evaluation method based on deep learning in Example 1 of the present invention;
[0056] Figure 2 Schematic diagram of the structure of the wind turbine EFC evaluation system based on deep learning in Example 2 of the present invention. DETAILED DESCRIPTION
[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0058] Example 1
[0059] Reference Figure 1 , a wind turbine EFC evaluation method based on deep learning, characterized in that it includes the steps of:
[0060] Step 1: Collect wind power related data of the wind farm and obtain an initial data set;
[0061] In this embodiment, the wind power related data of the wind farm includes: wind power real-time monitoring data, wind power historical data and numerical weather forecast data.
[0062] In this embodiment, the data acquisition system (SCADA system) in the wind farm can be used to obtain in real time the environmental parameters such as the power generation power, wind speed, wind direction, temperature, humidity, air pressure, and the operating status of each wind turbine, such as the operating status, fault information, etc.; through the programming interface (API) or database query, the latest wind power real-time monitoring data can be pulled from the SCADA system at regular intervals.
[0063] In this embodiment, historical wind power data for a period of time in the past, such as historical data for the past year or several years, can be extracted from the data warehouse or database of the wind farm. These data should include the same parameters as the real-time monitoring data. The historical wind power data can be extracted in batches from the data warehouse using SQL query statements or data extraction tools (such as ETL tools).
[0064] In this embodiment, it is possible to cooperate with the meteorological department to obtain numerical weather forecast data for the area where the wind farm is located, including predicted values of wind speed, wind direction, temperature, humidity, air pressure, etc. for a period of time in the future; and obtain numerical weather forecast data for the required time period through the API or data file download method provided by the meteorological department.
[0065] Integrate real-time monitoring data, historical data and weather forecast data according to key fields such as timestamp and wind turbine number to form a unified data format; perform basic verification on the data, such as checking data integrity (whether there are missing values) and rationality (such as whether the corresponding relationship between wind speed and power generation is reasonable), to ensure the accuracy and reliability of the data; save the integrated and verified data as the initial data set. This data set will serve as the starting point for subsequent data processing and analysis, and contains information in multiple dimensions such as real-time monitoring data of wind power, historical data and numerical weather forecast data.
[0066] Step 2: Process missing values and normalize the initial data set to obtain a preprocessed data set;
[0067] In this embodiment, the specific process of processing missing values for the initial data set is:
[0068] According to the characteristics of the data and business needs, different weights ω are assigned to observations at different time points i These weights can be determined based on factors such as how far the observation is from the missing point, the quality of the observation (such as credibility), etc.
[0069] The weighted moving average method is used to process missing values in the initial data set. The expression is shown in the following formula (1):
[0070]
[0071] Among them, M t is the weighted moving average of time t, ωi Y t-i+1 The weight, Y t-i+1 is the observed value at time t-i+1, and N is the number of moving average items;
[0072] According to formula (1), the weighted moving average M of each missing time point is calculated. t That is, for each missing value, take the observed values of the N time points before and after it (including the missing point itself, if available), and perform weighted average according to the assigned weights; the calculated weighted moving average M t As a replacement for missing values, fill in the corresponding positions in the initial data set.
[0073] In this embodiment, the specific process of normalizing the initial data set is as follows:
[0074] Traverse each feature in the initial data set (such as wind speed, wind direction, temperature, etc.) and find the minimum value W of the feature in all samples. min and the maximum value W max ;
[0075] The maximum and minimum normalization method is used to normalize the initial data set, and the expression is shown in the following formula (2):
[0076]
[0077] Among them, W ′ is the normalized value, W is the original data in the initial data set, and W min and W max are the minimum and maximum values in the original data respectively; according to formula (2), each eigenvalue of each sample is normalized. That is, the original data W is subtracted from the minimum value W min , and then divided by the maximum value W max With the minimum value W min The difference between ′ ; Replace the corresponding values in the original data set with the normalized values to complete the normalization of the entire data set.
[0078] Step 3: extract key features from the preprocessed data set, and cluster the extracted key features to obtain a key feature data set;
[0079] In this embodiment, the specific process of extracting key features from the preprocessed data set and clustering the extracted key features to obtain a key feature data set is as follows:
[0080] Select features that have a significant impact on the wind turbine EFC evaluation from the preprocessed data set; extract key features from the preprocessed data set, including wind speed, wind direction, temperature, air pressure, humidity and wind turbine operating status, wind turbine operating status features such as speed, pitch angle and other key features;
[0081] Perform cluster analysis on key features, and construct several cluster feature vectors based on the cluster analysis results. Each cluster feature vector represents the wind farm state and wind turbine operation state at a time point. For each time point, the extracted key feature values are combined into a cluster feature vector, which contains the wind farm environmental state (such as wind speed, wind direction, temperature, air pressure, humidity) and wind turbine operation state (such as speed, pitch angle, etc.) at that time point.
[0082] The K-means clustering algorithm is used to cluster several cluster feature vectors and obtain the key feature data set. K-means is a widely used clustering algorithm that can divide data points into K clusters, so that the similarity of data points in the same cluster is high, while the similarity of data points between different clusters is low. In practical applications, the selection of the number of clusters K needs to be determined according to the characteristics of the data and business needs. The optimal K value can be selected by trying different K values and evaluating the clustering effect (such as silhouette coefficient, Cali nski-Harabasz index, etc.). Use the selected K value and K-means The s algorithm clusters the cluster feature vectors. The algorithm iteratively assigns data points to the nearest cluster center and updates the position of the cluster center until the stopping condition is met (such as reaching the maximum number of iterations, the change of the cluster center is less than a certain threshold, etc.); after clustering, each cluster feature vector will be assigned to a cluster, and these clusters represent different combinations of wind farm status and wind turbine operating status; the clustering results are organized into a key feature data set, which contains the cluster feature vector of each cluster and the corresponding cluster label (i.e., the cluster number or name). The key feature data set will be used as the input for the subsequent construction of the wind turbine EFC evaluation model.
[0083] Step 4: Based on the key feature data set, a wind turbine EFC evaluation model is constructed through a neural network;
[0084] In this embodiment, the specific process of constructing the wind turbine EFC evaluation model through a neural network based on the key feature data set is as follows:
[0085] The initial wind turbine EFC evaluation model is constructed through the Seq2Seq neural network model; the Seq2Seq model is a neural network model that is particularly suitable for processing sequence-to-sequence conversion tasks. It consists of two parts: an encoder and a decoder. The encoder is responsible for encoding the key feature data set into a vector of a fixed length, while the decoder is responsible for decoding this vector into predicted wind power; the Seq2Seq model can process input and output sequences of different lengths, and can better focus on the important parts of the input sequence through the attention mechanism, thereby improving the accuracy of the prediction.
[0086] Based on the key feature data set, the initial wind turbine EFC evaluation model is trained and the wind turbine EFC evaluation model is obtained; the input layer of the model receives the key feature data set, and each data point contains features such as wind speed, wind direction, temperature, air pressure, humidity and wind turbine operating status; the encoder is composed of multiple LSTM (long short-term memory network) or GRU (gated recurrent unit) layers, which are used to extract time dependencies and hidden features in the key feature data set; the decoder is also composed of multiple LSTM or GRU layers, which receives the output of the encoder as the initial hidden state and generates a predicted wind power sequence; the output layer is usually a fully connected layer, which is used to convert the output of the decoder into the final wind power prediction value.
[0087] In the process of training the initial wind turbine EFC evaluation model, the parameters of the initial wind turbine EFC evaluation model are updated through the Adam optimizer to minimize the mean square error loss function.
[0088] In this embodiment, the key feature data set is divided into a training set, a validation set and a test set; the training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the model performance; the mean square error (MSE) is selected as the loss function because it can directly reflect the square difference between the predicted value and the true value, thereby guiding the model optimization; the Adam optimizer is used to update the parameters of the model; the Adam optimizer combines the advantages of the momentum method and the RMSprop optimization algorithm, and can adaptively adjust the learning rate to accelerate the convergence process of the model; during the training process, the data of the training set is iteratively input into the model, and the value of the loss function is calculated. Then, the Adam optimizer is used to update the parameters of the model to minimize the value of the loss function. At the same time, the performance of the model is monitored by the validation set to avoid the occurrence of overfitting; when the performance of the model is no longer significantly improved on the validation set, or when the preset training rounds are reached, the training process is stopped. The test set is used to evaluate the performance of the trained wind turbine EFC evaluation model. Calculate indicators such as the mean absolute error and root mean square error of the model to evaluate the prediction accuracy and stability of the model; if the performance of the model does not meet the requirements, the model can be adjusted according to the evaluation results, such as increasing the number of network layers, adjusting the learning rate, using regularization methods, etc., and then re-training and re-evaluating.
[0089] Step 5: Evaluate wind power through the wind turbine EFC evaluation model and provide corresponding operation and maintenance measures based on the evaluation results.
[0090] The wind power is evaluated by the wind turbine EFC evaluation model. The mean absolute error and root mean square error are calculated according to the evaluation results. The mean absolute error and root mean square error are used as evaluation indicators of the wind turbine EFC performance.
[0091] In this embodiment, the expressions of mean absolute error and root mean square error are shown in the following formula (3) and formula (4) respectively:
[0092]
[0093] Where n is the number of samples, y i is the true value, is the predicted value, M E is the mean absolute error, R E is the root mean square error.
[0094] In this embodiment, the new wind farm status data (including key features such as wind speed, wind direction, temperature, air pressure, humidity and wind turbine operating status) is input into the trained wind turbine EFC evaluation model; the model outputs the wind power prediction value at the corresponding time point according to the input key feature data set; the mean absolute error M is calculated based on the prediction value output by the model and the actually observed wind power value. E and the root mean square error RE , the calculated mean absolute error M E and the root mean square error R E As an indicator for evaluating the performance of the model and the wind turbine EFC system.
[0095] In this embodiment, the specific process of providing corresponding operation and maintenance measures according to the evaluation results is as follows:
[0096] The stability and efficiency of the current wind power system are evaluated according to the prediction results output by the wind turbine EFC evaluation model; the prediction results include evaluation indicators of the wind turbine EFC performance, such as the mean absolute error and the root mean square error in this embodiment.
[0097] If the prediction error exceeds the preset threshold, the wind turbine EFC performance is judged to be poor and the following operation and maintenance measures need to be taken:
[0098] Adjust wind turbine parameters: According to the actual value of the prediction error, adjust the wind turbine's speed, pitch angle and other key parameters to optimize energy flow and reduce the prediction error;
[0099] Optimize wind farm layout: If the prediction errors of at least two wind turbines exceed the preset threshold, the layout of the wind farm is optimized, such as adjusting the spacing and direction of wind turbines, to improve the utilization efficiency of wind resources and the operating stability of wind turbines;
[0100] Early warning and maintenance: Based on the prediction results, early warning of potential operational failures is provided, and targeted maintenance and repair work is arranged to prevent major failures from occurring;
[0101] Data analysis and feedback: Feedback the prediction results and operation and maintenance measures data to the deep learning model, continuously optimize and update the model, and improve the prediction accuracy and adaptability of the model.
[0102] In this embodiment, when a single fan has poor performance:
[0103] If the prediction error of a wind turbine exceeds the preset threshold, but other wind turbines perform normally, the focus will be on optimizing the operation and maintenance of that wind turbine. According to the specific situation of the prediction error, adjust the key parameters of the wind turbine, such as the speed and pitch angle, to optimize its energy capture efficiency and operating stability; conduct detailed fault diagnosis on the wind turbine, and arrange corresponding maintenance and repair work according to the diagnosis results.
[0104] For situations where multiple fans are underperforming:
[0105] If the prediction errors of multiple wind turbines exceed the preset threshold, the layout or operation strategy of the entire wind farm needs to be adjusted. Re-evaluate the layout of the wind farm, including the spacing and direction of wind turbines, to optimize the utilization efficiency of wind resources and reduce mutual interference between wind turbines; adjust the overall operation strategy according to the actual situation of the wind farm, such as optimizing wind power grid-connected scheduling and improving wind power storage solutions.
[0106] Example 2
[0107] Reference Figure 2 Based on Example 1, in this embodiment, a wind turbine EFC evaluation system based on deep learning is provided, which is used to implement the wind turbine EFC evaluation method based on deep learning, including:
[0108] Data acquisition module 101: used to collect wind power related data; collect wind power real-time monitoring data, wind power historical data, numerical weather forecast data and meteorological data (such as wind speed, wind direction, temperature, air pressure, humidity, etc.) from data sources such as wind farms and meteorological stations; data is acquired through API interface, file transfer or database synchronization, and stored in the system's data warehouse.
[0109] Data preprocessing module 102: used to perform preprocessing operations on the initial data set; perform missing value processing and normalization processing on the collected initial data set; the missing value processing adopts the weighted moving average method to estimate the missing value according to the time series characteristics of the historical data; the normalization processing adopts the maximum and minimum value normalization method to scale all the eigenvalues to the same scale to facilitate subsequent model training.
[0110] Feature extraction and clustering module 103: used to extract key features from the preprocessed data set, cluster the features through a clustering algorithm, and construct a key feature data set; extract key features from the preprocessed data set, including wind speed, wind direction, temperature, air pressure, humidity, and wind turbine operating status; perform cluster analysis on the key features and construct cluster feature vectors, each of which represents the status of a wind farm and wind turbine at a time point; use the K-means clustering algorithm to cluster the cluster feature vectors to obtain a key feature data set for model training.
[0111] Model building module 104: used to build a wind turbine EFC evaluation model and optimize model parameters; based on the key feature data set, an initial wind turbine EFC evaluation model is built through a Seq2Seq neural network model; the model is trained using an Adam optimizer to minimize the mean square error loss function and improve the prediction accuracy of the model; during the training process, the generalization ability of the model is evaluated through cross-validation and other techniques to prevent overfitting.
[0112] Evaluation and decision module 105: used to evaluate the EFC of wind turbines and give corresponding operation and maintenance measures according to the evaluation results; use the trained wind turbine EFC evaluation model to predict wind power, and calculate the mean absolute error and root mean square error as performance evaluation indicators; evaluate the stability and efficiency of the wind power system according to the prediction results, and if the prediction error exceeds the preset threshold, the operation and maintenance decision process is triggered; the operation and maintenance decision includes adjusting wind turbine parameters (such as speed, pitch angle), optimizing wind farm layout, early warning and maintenance, and data analysis and feedback; the data analysis and feedback link feeds back the prediction results and operation and maintenance measures to the model for continuous optimization and updating to improve the adaptability and prediction accuracy of the model.
[0113] Data storage and management module 106: used to store and manage various types of data generated by the system, and supports data query, export and backup;
[0114] User interaction module 107: used for users to view prediction results, performance indicators and operation and maintenance suggestions, and perform corresponding interactive operations.
[0115] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A wind turbine EFC evaluation method based on deep learning, characterized in that: Includes steps: Collect wind power related data of wind farms and obtain initial data sets; The initial data set is processed with missing values and normalized to obtain the preprocessed data set; Extract key features from the preprocessed data set, and cluster the extracted key features to obtain a key feature data set; Based on the key feature data set, a wind turbine EFC evaluation model is constructed through a neural network; The wind power is evaluated through the wind turbine EFC evaluation model, and corresponding operation and maintenance measures are given according to the evaluation results.
2. The wind turbine EFC evaluation method based on deep learning according to claim 1, characterized in that: The wind power related data of the wind farm include: wind power real-time monitoring data, wind power historical data and numerical weather forecast data.
3. The wind turbine EFC evaluation method based on deep learning according to claim 1, characterized in that: The specific process of processing missing values for the initial data set is as follows: The weighted moving average method is used to process missing values in the initial data set. The expression is shown in the following formula (1): Among them, M t is the weighted moving average of time t, ω i for y t-i+1 The weight of y t-i+1 is the observed value at time t-i+1, and N is the number of moving average items.
4. The wind turbine EFC evaluation method based on deep learning according to claim 1, characterized in that: The specific process of normalizing the initial data set is as follows: The maximum and minimum normalization method is used to normalize the initial data set, and the expression is shown in the following formula (2): Among them, W ′ is the normalized value, W is the original data in the initial data set, and W min and W max are the minimum and maximum values in the original data, respectively.
5. The wind turbine EFC evaluation method based on deep learning according to claim 1, characterized in that: The specific process of extracting key features from the preprocessed data set and clustering the extracted key features to obtain a key feature data set is as follows: Extract key features from the preprocessed data set, including wind speed, wind direction, temperature, air pressure, humidity, and fan operating status; Perform cluster analysis on key features and construct several cluster feature vectors based on the cluster analysis results. Each cluster feature vector represents the wind farm status and wind turbine operation status at a time point. The K-means clustering algorithm is used to cluster several cluster feature vectors and obtain the key feature data set.
6. The wind turbine EFC evaluation method based on deep learning according to claim 1, characterized in that: The specific process of constructing the wind turbine EFC evaluation model through a neural network based on the key feature data set is as follows: The initial wind turbine EFC evaluation model was constructed using the Seq2Seq neural network model; Based on the key feature data set, the initial wind turbine EFC evaluation model is trained and the wind turbine EFC evaluation model is obtained; In the process of training the initial wind turbine EFC evaluation model, the parameters of the initial wind turbine EFC evaluation model are updated through the Adam optimizer to minimize the mean square error loss function.
7. The wind turbine EFC evaluation method based on deep learning according to claim 1, characterized in that: The wind power is evaluated by the wind turbine EFC evaluation model. The mean absolute error and root mean square error are calculated according to the evaluation results. The mean absolute error and root mean square error are used as evaluation indicators of the wind turbine EFC performance.
8. The wind turbine EFC evaluation method based on deep learning according to claim 7, characterized in that: The expressions of the mean absolute error and the root mean square error are shown in the following formula (3) and formula (4) respectively: Where n is the number of samples, y i is the true value, is the predicted value, M E is the mean absolute error, R E is the root mean square error.
9. The wind turbine EFC evaluation method based on deep learning according to claim 1, characterized in that: The specific process of providing corresponding operation and maintenance measures based on the evaluation results is as follows: Evaluate the stability and efficiency of the current wind power system based on the prediction results output by the wind turbine EFC evaluation model; If the prediction error exceeds the preset threshold, the wind turbine EFC performance is judged to be poor and the following operation and maintenance measures need to be taken: Adjust fan parameters: According to the actual value of the prediction error, adjust the key parameters of the fan to optimize energy flow and reduce the prediction error; Optimize wind farm layout: If the prediction errors of at least two wind turbines exceed the preset threshold, the layout of the wind farm is optimized; Early warning and maintenance: Based on the prediction results, early warning of potential operating failures is provided, and targeted maintenance and repair work is arranged; Data analysis and feedback: Feedback the prediction results and operation and maintenance measures data to the deep learning model to continuously optimize and update the model.
10. A wind turbine EFC evaluation system based on deep learning, used to implement the wind turbine EFC evaluation method based on deep learning as described in any one of claims 1 to 9, characterized in that: include: Data acquisition module: used to collect wind power related data; Data preprocessing module: used to preprocess the initial data set; Feature extraction and clustering module: used to extract key features from the preprocessed data set, cluster the features through clustering algorithms, and construct a key feature data set; Model building module: used to build the wind turbine EFC evaluation model and optimize the model parameters; Evaluation and decision-making module: used to evaluate the EFC of wind turbines and provide corresponding operation and maintenance measures based on the evaluation results; Data storage and management module: used to store and manage various types of data generated by the system, and supports data query, export and backup; User interaction module: used for users to view prediction results, performance indicators and operation and maintenance suggestions, and perform corresponding interactive operations.