Research On The Forecast Method Of Wind Speed In Bohai Bay

The Bohai Bay is a closed shallow sea area with complex geographical environment,frequently affected by extratropical cyclones and cold waves,and frequent windy weather.Gale weather has a great impact on economic activities such as offshore engineering construction and maritime transportation.Therefore,it is of great significance to forecast the wind speed in the Bohai Bay.Machine learning is a method of implementing artificial intelligence.It mainly uses algorithms to parse data,learn deep information in data,and then construct models that provide decisions or predictions for specific problems.In recent years,more and more machine learning methods have been applied to scientific research and industrial processes,and have been effectively developed in many fields and have great potential.Based on the research and analysis of the data of the Bohai Bay A platform meteorological station and the ensemble prediction grid data,this paper proposes a wind speed prediction method based on the echo state network and the integrated learning regression decision tree.The main work is as follows:(1)Feature construction and feature selection.Through the basic data of the Meteorological Station of the Bohai Bay A platform,the characteristics of the combined physical quantities commonly used in meteorological analysis are constructed,and the feature selection of continuous attributes is realized through correlation degree analysis,which effectively reduces the information redundancy between the input physical quantity features.The complexity of the late prediction model is reduced,which lays a foundation for improving the prediction performance of the model and reducing the running time.(2)The wind speed prediction model is constructed based on meteorological physical quantities.The leakage integral neuron and online learning algorithm are introduced to construct the leak integral echo state network(LIESN)time series prediction model,and the key parameters of LIESN are optimized by genetic algorithm,which avoids the blind selection of model parameters.The wind speed forecast based on meteorological physical quantities is realized.(3)The wind speed prediction model is constructed based on the ensemble forecast data.This paper proposes a method of selecting the optimal members by categorical regression tree(CART)pruning,and constructs a regression prediction model by using the gradient learning tree(GBDT)integrated learning algorithm to realize the ensemble prediction wind speed prediction of small samples.The experimental results show that the accuracy of the LIESN prediction model is higher,and the integrated learning method is more effective for small sample data.