| With the development of society,people become more and more dependent on energy.Traditional fossil fuels have been unable to meet the needs of human development,but also harm our environment.In this case,wind energy as a renewable and clean energy is getting more and more attention.Although wind power generation has been developed for many years in China,it has not been used on a large scale.The main reason is that the wind power output has the characteristics of randomness,volatility and uncontrollability.These characteristics make the quality of wind power output is not guaranteed,which brings challenges to large-scale grid-connection.In order to ensure the safety of power grid,the demand for accurate prediction of wind power is more and more urgent.By summarizing the research in related fields in recent years,the accurate prediction methods of wind power are analyzed.The thesis mainly includes the following contents:(1)Firstly,the principle of wind power generation is understood,the factors affecting wind power generation are summarized,and the characteristics of wind power data are analyzed.(2)According to the characteristics of wind power data,LSTM model is used to predict wind power.According to the characteristics of data timing,seven different data sets were constructed,and two model structures based on LSTM were constructed according to different output requirements.The data set and model structure with the best prediction effect were selected through comparative experiment.The influence of Dropout mechanism and activation function on the model was explored.In addition,the advantages and disadvantages of different depth models are compared from the perspectives of prediction accuracy and training duration.(3)An integrated prediction method based on LSTM model is proposed.First,several weak predictors were trained by LSTM,and then weight distribution of data samples and weak predictors was conducted by Adaboost method,and finally a strong predictor was obtained.Simulation results show that the prediction accuracy o f LSTM model enhanced by Adaboost method is 2.6% higher than that of LSTM model.Due to the high demand for timeliness of wind power prediction,dimensionality reduction processing is carried out on the data of the input model to reduce the complexity of the model.Simulation results show that the training time is reduced by one third with the same prediction accuracy.Simulation results show that the LSTM model is better than the traditional model in short-term wind power prediction,and the prediction accuracy of LSTM is 3% higher than that of SVM.The integrated LSTM_Adaboost model is 2.6% better than the LSTM model. |
