Hybrid Model Based On Ensemble Empirical Mode Decomposition For Wind/PV Output Forecasting

In recent years, the developments of wind power and solar power have steppedinto a golden rapid period. However, power grid will suffer great challenges whendistributed power is connected into it because of distributed power’s intermittent anduncertain character. Thus, accurate forecasting of wind and solar power will do greathelp to the optimization of power system’s dispatch, the plan of power system andthe increase of distributed generator’s competitiveness. Wind speed and solarradiation are respectively the most important and direct factors that influence theoutput of wind turbine and PV. Accurate forecasting of wind speed and solarradiation is of great value in practical application.Hybrid wind speed and solar radiation forecasting models based on EnsembleExperience Mode Decomposition (EEMD) are proposed in this paper. The works inthis paper mainly include:1）Hybrid wind speed forecasting model based on EEMD is proposed. Thismodel is applied for very-short time wind speed forecasting. Firstly, original windspeed signal is decomposed into different signals with different frequencies.Secondly, different forecasting model is established for each signal according to theirown characteristics. Lastly, add up all the forecasting results of each signal and theforecasting results of original wind speed are obtained.2）Hourly hybrid solar radiation forecasting model based on EEMD and ELMNeural Network is proposed. In order to increase training effect of Neural Network,hourly similar day’s radiation sequence is established firstly. And then the sequenceis decomposed by EEMD. Establish ELM forecasting model for each signal and addup forecasting results of each signal. Then forecasting results of original solarsequence are obtained. ELM has very fast computation speed. The combination ofELM with EEMD can help save great mount of computation resources and time.Hybrid forecasting model can take advantages of different algorithms and casestudies in this paper show that with the help of hybrid forecasting model, forecastingaccuracy and computation time will be increased greatly.