Research On Wind Power Prediction Of Wind Farms Based On Spatial And Temporal Correlation

With the development of wind power generation technology,the importance of wind power prediction has become increasingly prominent.Due to the large number of wind turbines and the complexity of wind power data in wind farms,traditional wind power prediction models are unable to fully extract the spatiotemporal characteristics and hidden features from the historical data,resulting in low prediction accuracy.To address these challenges,this paper focuses on short-term wind power prediction,optimizes the input of wind power prediction models,and establishes a spatiotemporal correlation-based wind power prediction model.Furthermore,a spatiotemporal correlation-based wind power short-term prediction system is developed using this model.The specific work of this paper is as follows:(1)The wind power data obtained is analyzed,and wind power data that does not meet the data specifications are removed.Missing data are then filled using the linear interpolation method.The wind power data features that affect the size of the wind power are theoretically and quantitatively analyzed to determine the input of the prediction model.(2)The input of the model is optimized.Using the density-based spatial clustering algorithm(DBSCAN)and the quartile method,abnormal data are removed based on the strong correlation between wind speed and power.The geographic location of wind turbines and wind power data are fused as the wind turbine features.The K-means clustering algorithm is then used to classify the wind turbines.Finally,different wind power scenarios are formed using the K-means clustering of historical data of different classes of wind turbines based on the similarity of wind speed.(3)Based on the optimized model input,a prediction model is established.Firstly,the complete ensemble empirical mode decomposition(CEEMD)is used to decompose the power sequence of different wind power scenarios.The decomposed power sequences are combined with the wind power data feature sequence.Convolutional neural network(CNN)is used to extract spatial features,and gated recurrent unit(GRU)is used to extract temporal features for prediction.The predicted results of each decomposed sequence are combined to obtain the final prediction results.Experimental results show that the proposed prediction model has higher accuracy than traditional single models.(4)The spatiotemporal correlation-based wind power short-term prediction system is developed by integrating the above wind power prediction methods.Based on the requirements of the prediction model,the system’s module functions are determined,and the detailed design of each module is conducted.After testing,all modules of the system performed well and achieved the design goals.Figure 51 Table 11 Reference 82...