Valley fan-based wind speed prediction method

Abstract

The invention provides a wind speed prediction method based on a valley fan, and relates to the field of wind speed monitoring, and the method comprises the steps: data collection: collecting historical meteorological data of a valley, the historical meteorological data comprises air pressure, temperature, humidity, wind direction, wind speed and solar radiation data, and collecting data at the same time interval; the beneficial effects of the invention are that the method is specially designed for the unique wind regime characteristics of the valley region, comprehensively considers the topographic data and various meteorological factors of the valley, such as air pressure, temperature, humidity, wind direction and the like, and employs advanced data processing and feature extraction technologies, thereby achieving the real-time monitoring of the valley region. According to the method, the complex change rule of the wind speed of the valley fan can be more accurately captured through the effective learning of the LSTM on the time sequence features and the CNN on the spatial features in the deep learning model, compared with a traditional prediction method, the prediction error is remarkably reduced, and more reliable data support is provided for the stable operation and efficient management of a wind power plant.

Description

Technical Field

[0001] The invention relates to the field of wind speed monitoring, and in particular to a wind speed prediction method based on a valley wind turbine. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] As the global demand for clean energy continues to grow, wind power generation has been widely developed and applied as an important form of renewable energy. The site selection of wind farms often takes into account a variety of factors. Valley areas often have certain wind energy resource advantages due to their special topographic conditions, and many wind farms are built in valleys.

[0004] However, the wind conditions in the valley are extremely complex. The valley wind effect is significant. Valley wind prevails during the day, blowing from the valley to the hillside; mountain wind dominates at night, blowing from the hillside to the valley. This alternating pattern of wind direction changes during the day and night makes the wind speed and direction in the valley highly nonlinear, non-stationary, and strongly intermittent.

[0005] Traditional wind speed prediction methods, such as the autoregressive moving average model (ARIMA) based on time series analysis and the numerical weather forecast model based on physical principles, face many challenges when applied to the prediction of wind speed of wind turbines in valleys. The ARIMA model is difficult to effectively capture the complex nonlinear relationship of valley winds, and although the numerical weather forecast model can take into account large-scale meteorological factors, it has limited simulation accuracy for the microscopic changes in wind speed under local complex terrain such as valleys, and cannot accurately reflect the special airflow patterns and wind speed mutations in the valleys.

[0006] These problems lead to inaccurate predictions of wind turbine power generation during the operation of valley wind farms, which in turn affects the dispatch and stable operation of the power system. On the one hand, inaccurate wind speed predictions will make the operation and maintenance plans of wind turbines unreasonable, increasing operation and maintenance costs; on the other hand, they will cause deviations in the power forecast in power market transactions, affecting the economic benefits and market competitiveness of wind farms.

[0007] Therefore, there is an urgent need for a wind speed prediction method specifically for valley wind turbines that can fully consider the unique wind conditions in valley areas and accurately predict the wind speed changes at the wind turbine location, thereby improving the operating efficiency of wind farms, reducing costs, and enhancing the stability and reliability of the power system, providing strong technical support for the sustainable development of wind power generation in valley areas. Summary of the invention

[0008] The purpose of the present invention is to provide a wind speed prediction method based on valley wind turbines. It is designed specifically for the unique wind conditions in valley areas, comprehensively considers the terrain data of the valley and various meteorological factors, such as air pressure, temperature, humidity, wind direction, etc., and uses advanced data processing and feature extraction techniques, as well as the effective learning of LSTM for time series features and CNN for spatial features in the deep learning model. It can more accurately capture the complex changes in the wind speed of valley wind turbines, significantly reduce the prediction error compared to traditional prediction methods, and provide more reliable data support for the stable operation and efficient management of wind farms.

[0009] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0010] A wind speed prediction method based on valley wind turbines includes data collection, collecting historical meteorological data of the valley, the historical meteorological data including air pressure, temperature, humidity, wind direction, wind speed and solar radiation data, collecting data at the same time interval, and the collection time is not less than a preset time threshold, and at the same time collecting valley terrain data, the valley terrain data including valley altitude, slope, slope direction and valley direction information.

[0011] It also includes data preprocessing, which processes the collected historical meteorological data and standardizes the data so that the mean of the data is zero and the variance is one. The processed historical meteorological data is integrated with the terrain data to construct a comprehensive data set.

[0012] It also includes feature extraction, extracting wind speed-related features from the comprehensive data set, the related features include pressure differences at different altitudes, temperature gradients, humidity change rates, and wind direction change frequencies, and using principal component analysis methods to perform dimensionality reduction processing on the extracted features, retaining the principal component features whose cumulative contribution rates reach a preset contribution rate threshold as the final wind speed prediction feature vector.

[0013] The invention comprises model construction, constructing a wind speed prediction model based on deep learning, wherein the model adopts an architecture combining LSTM and CNN, the LSTM network is used to learn the time series characteristics of wind speed data, the CNN network is used to extract the spatial characteristics in the feature vector, and the wind speed prediction feature vector after dimensionality reduction processing is input into the model for training.

[0014] During the model training process, an adaptive learning rate adjustment strategy is adopted to dynamically adjust the learning rate according to the changes in the loss function during the training process. When the loss function does not decrease for several consecutive training cycles, the learning rate is reduced, and an early stopping method is introduced to prevent overfitting. When the loss function on the validation set does not improve for several consecutive cycles, the training is stopped and the model parameters at this time are saved.

[0015] It also includes model evaluation, using an independent test data set to evaluate the trained model. The evaluation indicators include root mean square error, mean absolute error and correlation coefficient. If the evaluation indicators do not meet the preset performance requirements, the model is optimized and adjusted. The optimization methods include increasing the amount of training data, adjusting the model structure or hyperparameters.

[0016] The performance of the model meets the requirements, and wind speed prediction is performed. The current meteorological data and terrain data collected in real time are subjected to the same data preprocessing and feature extraction operations, and the obtained feature vector is input into the trained model to obtain the predicted wind speed value of the wind turbine, and the predicted power generation of the wind turbine is calculated based on the predicted wind speed value, providing a basis for the operation and scheduling of the wind farm.

[0017] It also includes model updating, regularly adding newly collected historical data to the training data set, retraining and optimizing the model to adapt to the long-term and seasonal changes in valley wind conditions and maintain the model's predictive accuracy and reliability.

[0018] Compared with the prior art, the beneficial effects of the present invention are:

[0019] Improve wind speed prediction accuracy: By designing specifically for the unique wind conditions in valleys, taking into account the valley's terrain data and a variety of meteorological factors such as air pressure, temperature, humidity, wind direction, etc., and using advanced data processing and feature extraction techniques, as well as the effective learning of LSTM for time series features and CNN for spatial features in deep learning models, it can more accurately capture the complex changes in wind speed of valley wind turbines, significantly reduce prediction errors compared to traditional prediction methods, and provide more reliable data support for the stable operation and efficient management of wind farms.

[0020] Optimize wind farm operation and maintenance arrangements: Accurate wind speed prediction enables the wind farm operation and maintenance team to know in advance the wind speed conditions that the wind turbines may face, and arrange equipment inspection, maintenance and maintenance work reasonably. For example, during periods of high and unstable wind speeds, strengthen inspections of key components to avoid equipment failures and downtime due to sudden strong winds, reduce unplanned downtime, thereby reducing operation and maintenance costs and extending the service life of wind turbines.

[0021] Enhance the stability and reliability of the power system: At the power system dispatching level, accurate wind speed forecasting can help dispatchers plan the allocation and balance of power resources in advance. Wind farms can deliver electricity to the power grid more stably based on the forecast results, reduce the impact of power generation fluctuations caused by wind speed forecast deviations on the power grid, improve the overall stability of the power system, ensure the reliability of power supply, and reduce safety risks caused by imbalances in power supply and demand.

[0022] Improve the economic benefits of wind farms: On the one hand, due to the improvement of prediction accuracy, wind farms can more accurately estimate power generation in power market transactions, formulate reasonable trading strategies, and avoid economic losses or default risks caused by inaccurate power forecasts; on the other hand, by optimizing operation and maintenance to reduce costs and improve power generation efficiency, the overall revenue of wind farms can be increased, the competitiveness of wind farms in the renewable energy market can be improved, and the sustainable development of the wind power industry in valley areas can be promoted.

[0023] Adapt to changes in valley wind conditions: The model has a regular update mechanism, which can continuously learn the long-term and seasonal changes in valley wind conditions based on newly collected data, continuously optimize its own prediction performance, and always maintain high accuracy and adaptability in predicting wind speeds of valley wind turbines, ensuring that it can provide strong technical support for wind farm operations under different time scales and climatic conditions, and effectively respond to the complex and changeable wind energy resource environment in valley areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Attached Figure 1 It is a schematic diagram of the steps of the present invention.

[0025] Attached Figure 2 It is a schematic diagram of the process of the present invention.

[0026] Attached Figure 3 It is a schematic diagram of the valley information matrix in the present invention. DETAILED DESCRIPTION

[0027] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the application equally.

[0028] The present invention is a wind speed prediction method based on a valley wind turbine, the main structure of which includes

[0029] Data collection: 1. Meteorological data collection: Multiple meteorological monitoring stations are reasonably arranged in the valley area where the wind turbine is located to ensure that the meteorological information at different locations and heights in the valley can be fully covered. High-precision meteorological sensors are used to collect data such as air pressure, temperature, humidity, wind direction, wind speed, and solar radiation at set time intervals (for example, every 10 minutes). The collection period is at least 3 years to obtain a sufficiently rich historical meteorological data sample.

[0030] The collected meteorological data is aggregated to the data center server via wireless or wired network transmission for unified storage and management.

[0031] 2. Topographic data collection uses geographic information system (GIS) technology, satellite remote sensing images and field measurements to obtain valley terrain data. To determine the altitude data of the valley, high-precision GPS measuring equipment can be used to measure at multiple key locations in the valley, and calibrated in combination with satellite elevation data; to measure the slope and slope direction of the valley, professional topographic measuring instruments can be used to conduct field measurements in different areas, and the data can be marked on electronic maps; the direction of the valley is analyzed, and its general direction information is determined based on the extension direction of the mountains in the valley and the geographical features, and these terrain data are associated and stored with meteorological data.

[0032] Data preprocessing: 1. Data cleaning: Write a data cleaning program to check the collected historical meteorological data one by one. Set a reasonable data threshold range. For example, data points with wind speeds exceeding the normal range (such as 0-100m / s) are considered abnormal data and deleted; for time periods with continuous missing data exceeding a certain proportion (such as 10%), mark them for subsequent processing.

[0033] Interpolation algorithms based on data characteristics are used to supplement missing data, such as the daily and seasonal variation patterns of historical data and the data correlation of adjacent monitoring stations. For example, if the wind speed data at a certain moment is missing, the average wind speed of the days before and after the same moment of the day and the wind speed data of adjacent stations can be used for interpolation calculation.

[0034] 2. Data standardization, calculate the mean and variance of the processed data, using the standardization formula (where x is the original data, μ is the mean, and σ is the variance) The data is standardized so that the mean of the data is zero and the variance is one, which is convenient for subsequent model training and calculation. The standardized meteorological data and terrain data are integrated according to a specific data format to construct a comprehensive data set. For example, the meteorological data at each time point is combined with the corresponding geographical location (latitude and longitude) and terrain feature data into a data record.

[0035] Feature extraction: 1. Extraction of relevant features, write a feature extraction program to extract key features related to wind speed from the comprehensive data set. Calculate the pressure difference at different altitudes (such as the monitoring data corresponding to the bottom of the valley, the mountainside, the top of the mountain, etc.), which is obtained by subtracting the pressure data of adjacent altitudes; calculate the temperature gradient, and calculate the rate of change of temperature with distance based on the temperature data at different locations in the valley; determine the humidity change rate, and analyze the changes in humidity data in time and space; count the frequency of wind direction changes, and record the number of wind direction changes per unit time.

[0036] 2. Feature dimensionality reduction: Use the principal component analysis (PCA) method to reduce the dimensionality of the extracted features. The extracted multiple features are combined into a feature matrix, the covariance matrix of the feature matrix is ​​calculated, the eigenvalues ​​and eigenvectors of the covariance matrix are solved, the eigenvectors are sorted according to the eigenvalues, and the first few principal component features whose cumulative contribution rate reaches the preset contribution rate threshold (such as 90%) are selected as the final wind speed prediction feature vectors. This not only retains the main information, but also reduces the data dimension and improves the model training efficiency.

[0037] Model construction: 1. Model architecture design, build a wind speed prediction model based on deep learning, using a combination of long short-term memory network (LSTM) and convolutional neural network (CNN) architecture. The LSTM network part sets multiple hidden layers (such as 3-5 layers), and the number of neurons in each hidden layer is reasonably determined according to data characteristics and computing resources (such as 64-128), which is used to learn the time series characteristics of wind speed data and can effectively process the long-term dependency of wind speed data in time series; the CNN network part uses a combination of convolutional layer, pooling layer and fully connected layer. The convolution kernel size and step size are adjusted according to the dimension of the feature vector, which is used to extract the spatial features in the feature vector and explore the impact of spatial factors such as valley terrain on wind speed. The output of the LSTM network is fused with the output of the CNN network, connected to a fully connected layer, and finally the wind speed prediction value is output.

[0038] 2. Model training preparation: Divide the wind speed prediction feature vector after dimensionality reduction into different data sets according to a certain ratio (such as 70% for training set, 20% for validation set, and 10% for test set). Initialize the model, set the initial parameters of the model, such as weights and biases in LSTM and CNN networks, and select appropriate loss functions (such as mean square error function) and optimization algorithms (such as Adam optimization algorithm).

[0039] Model training: In the training process, the training data set is input into the constructed model for training. During the training process, an adaptive learning rate adjustment strategy is adopted. For example, the initial learning rate is set to 0.001. When the loss function does not decrease for 5 consecutive training cycles, the learning rate is multiplied by a decay factor (such as 0.9) to reduce it. At the same time, the early stopping method is introduced. When the loss function on the validation set does not improve for 10 consecutive cycles, the training is stopped and the model parameters at this time are saved to prevent the model from overfitting. During the training process, the loss function value, accuracy and other indicators on the training set and validation set are recorded in real time to observe the training effect and performance changes of the model.

[0040] Model evaluation: 1. Evaluation index calculation, use an independent test data set to evaluate the trained model. Calculate the evaluation indicators root mean square error (RMSE), mean absolute error (MAE) and correlation coefficient (R). The RMSE calculation formula is (where y i is the true wind speed value, is the predicted wind speed value, n is the number of test data samples), which is used to measure the degree of deviation between the predicted value and the true value; the MAE calculation formula is Reflects the average size of the prediction error; correlation coefficient (in is the true mean wind speed, is the predicted wind speed mean), which is used to evaluate the linear correlation between the predicted value and the true value.

[0041] 2. Model optimization decision, if the evaluation index does not meet the preset performance requirements (such as RMSE is greater than the preset threshold or R is less than the preset threshold), the model is optimized and adjusted. Optimization methods include increasing the amount of training data, such as supplementing more historical meteorological data or collecting more field monitoring data; adjusting the model structure, such as increasing or decreasing the number of layers or neurons of the LSTM or CNN network; optimizing hyperparameters, such as adjusting the initial value of the learning rate, the decay factor, the parameters of the early stopping method, etc., and then retraining and evaluating the model until the model performance meets the requirements.

[0042] Wind speed prediction: 1. Real-time data processing: After the model evaluation is completed and the model performance meets the requirements, wind speed prediction is performed. The current meteorological data and terrain data collected in real time are subjected to the same data preprocessing and feature extraction operations. First, the real-time meteorological data is cleaned and standardized to remove abnormal data and make it have the same distribution characteristics as the training data; then the key features related to wind speed are extracted, and feature dimensionality reduction is performed to obtain feature vectors in the same format as during training.

[0043] 2. Wind speed prediction and application: Input the obtained feature vector into the trained model to obtain the predicted wind speed value of the wind turbine. According to the power curve of the wind turbine, the predicted wind speed value is converted into the predicted power value, and the prediction result is transmitted to the monitoring center and wind farm operation management system of the wind farm in real time to provide a basis for the operation and scheduling of the wind farm. For example, according to the predicted wind speed and power, the start and stop sequence of the wind turbine can be reasonably arranged, the power output plan can be optimized, and the overall power generation efficiency and economic benefits of the wind farm can be improved.

[0044] Model update: 1. Data update and model retraining. Newly collected historical data is added to the training data set regularly (such as monthly or quarterly) to retrain and optimize the model. First, the new data is processed according to the steps of data collection, preprocessing, and feature extraction. Then the processed new feature vector is merged with the original training data set. The model is retrained according to the model training method, and the model parameters are adjusted to adapt to the long-term and seasonal changes in valley wind conditions to maintain the prediction accuracy and reliability of the model. During the model update process, model evaluation should also be performed to ensure that the performance of the updated model is not lower than that of the original model.

Claims

1. A wind speed prediction method based on valley wind turbines, characterized in that: It includes data collection, collecting historical meteorological data of the valley, the historical meteorological data including air pressure, temperature, humidity, wind direction, wind speed and solar radiation data, collecting data at the same time interval, and the collection time is not less than the preset time threshold, and collecting valley terrain data at the same time, the valley terrain data including valley altitude, slope, slope direction and valley direction information.

2. A wind speed prediction method based on a valley wind turbine according to claim 1, characterized in that: It also includes data preprocessing, which processes the collected historical meteorological data and standardizes the data so that the mean of the data is zero and the variance is one. The processed historical meteorological data is integrated with the terrain data to construct a comprehensive data set.

3. A wind speed prediction method based on a valley wind turbine according to claim 2, characterized in that: It also includes feature extraction, extracting wind speed-related features from the comprehensive data set, the related features include pressure differences at different altitudes, temperature gradients, humidity change rates, and wind direction change frequencies, and using principal component analysis methods to perform dimensionality reduction processing on the extracted features, retaining the principal component features whose cumulative contribution rates reach a preset contribution rate threshold as the final wind speed prediction feature vector.

4. A wind speed prediction method based on a valley wind turbine according to claim 3, characterized in that: The invention comprises model construction, constructing a wind speed prediction model based on deep learning, wherein the model adopts an architecture combining LSTM and CNN, the LSTM network is used to learn the time series characteristics of wind speed data, the CNN network is used to extract the spatial characteristics in the feature vector, and the wind speed prediction feature vector after dimensionality reduction processing is input into the model for training.

5. A wind speed prediction method based on a valley wind turbine according to claim 4, characterized in that: During the model training process, an adaptive learning rate adjustment strategy is adopted to dynamically adjust the learning rate according to the changes in the loss function during the training process. When the loss function does not decrease for several consecutive training cycles, the learning rate is reduced, and an early stopping method is introduced to prevent overfitting. When the loss function on the validation set does not improve for several consecutive cycles, the training is stopped and the model parameters at this time are saved.

6. A wind speed prediction method based on a valley wind turbine according to claim 5, characterized in that: It also includes model evaluation, using an independent test data set to evaluate the trained model. The evaluation indicators include root mean square error, mean absolute error and correlation coefficient. If the evaluation indicators do not meet the preset performance requirements, the model is optimized and adjusted. The optimization methods include increasing the amount of training data, adjusting the model structure or hyperparameters.

7. A wind speed prediction method based on a valley wind turbine according to claim 6, characterized in that: The performance of the model meets the requirements, and wind speed prediction is performed. The current meteorological data and terrain data collected in real time are subjected to the same data preprocessing and feature extraction operations, and the obtained feature vector is input into the trained model to obtain the predicted wind speed value of the wind turbine, and the predicted power generation of the wind turbine is calculated based on the predicted wind speed value, providing a basis for the operation and scheduling of the wind farm.

8. A wind speed prediction method based on a valley wind turbine according to claim 7, characterized in that: It also includes model updating, regularly adding newly collected historical data to the training data set, retraining and optimizing the model to adapt to the long-term and seasonal changes in valley wind conditions and maintain the model's predictive accuracy and reliability.

Images