Abstract
We developed a methodology to unify a collection of time series predictions by a wide variety of time series prediction methods. The proposed ensemble method produces a weighted average of multiple predictions. The method is designed to be highly versatile and adaptable as it can be used for unifying time series predictions with different structures such as the time of delivery, the time resolution, the updating frequency and the prediction lead time. We implemented the ensemble method into a wind energy ramp forecasting system, where multiple forecasting methodologies including meteorological factor analysis, an ensemble forecast method, a statistical method, a machine learning based method and a dynamical systems theory based method are built in. Our method successfully produced the single prediction that results in better performance over other individual predictions.
