Forecasting generation of a PV power plant with a little data-set using information fusion

This paper presents a new information based method of prediction which helps to obtain more information from a little data set and applies it on a data set of photovoltaic power plant of University of Tehran which only contained data of 100 days at the time of research. In this way some conventional models of time series are applied on the data set and output of models are fused by some common fusion algorithms and a new fusion algorithm which is proposed in this paper. Results show that using fusion algorithm a better forecasting could be obtained in the sense of maximum error of forecasting, root mean square of error, similarity in forecasting pattern of power generation in cloudy days and finally statistical characteristics of residuals. Results also express proposed algorithm could be observed as a good fusion algorithm.