Research On Power Prediction Of Wind And Solar Electric Field Based On Fusion Model

The issue of energy is a discuss hot topics as soon as development of social,what’s more,for use energy establish the whole factories and industrial facilities due to the environmental pollution gradually worse.In order to solve this problem,the most important means is enlarged the research and development of new energy production and application.Now,technology and production equipment of Wind and photovoltaic power generation are relatively mature,and they need a low environmental requirements can produce a lot output.However,the problem that people are facing to solve is how to make full use of the energy generated by wind and photovoltaic power generation when the volatility of power generation caused by unstable weather conditions.Because the newly generated energy is not properly use in time,resulting the newly generated energy loss with a low generated energy utilization rate.The prediction of wind and photovoltaic farm output power helps to provide reliable guidance for grid-connection strategy and power dispatching,alleviating grid load pressure and reducing maintenance and operational safety problems caused by grid-connection impact.Therefore,the accuracy of wind farm power prediction is very important in this working mode.In order to deal with the problem of the low power prediction accuracy,here,the thesis analyzed the relationship between the original data and the influence of Numerical weather prediction.Finally,find some deep learning models to deal with the generation power prediction and Compare with the evaluation index.The thesis major research contents as follows:(1)First of all,preprocessing the wind and photovoltaic power data,and the missing data are processed by K-means method,and then the relevance of various meteorological factors to the actual output power is analyzed.(2)Several common deep learning models are used to complete the prediction work.In view of the dependence of power prediction on numerical weather forecast,the time-series historical power data without numerical weather forecast is considered for prediction.Short-term memory method and short-term memory method are used to forecast photovoltaic power generation and wind power generation respectively,and the prediction effect without numerical weather prediction is obtained.At the same time,the results of numerical weather forecast are compared.Based on the high precision of short-term forecast,the numerical weather forecast method is selected to complete the power prediction of wind farm.(3)In recent years,using some deep learning models to deal with the generation power prediction with attention mechanism.By attention mechanism improved the accuracy of features extracted from fusion model.(4)Considering the influence of different season and length from data set,and compare the results.That forecast photovoltaic power generation is consider seasonal factors and forecast wind power generation is consider length factor.(5)If using the MAPE deal with the generation power prediction,the results may deviate in some cases.Hence,this thesis proposes an improved mean maximum absolute percentage error(MMAPE)evaluation index.