PV Power Characteristic Modeling based on Multi-scale Clustering and Its Application in Generation Prediction

The modeling and prediction of stochastic photovoltaic (PV) power curve is a widely concerned issue in power grid operation and planning. Mining and utilization of the big data recorded by the PV stations provide a new method for the modeling of the fluctuating PV power. This paper presents a multi-scale analysis scheme to explore and model both the inherent and the stochastic characteristics of the PV power time sequence. The proposed scheme consists of the multiple clustering analysis at the large time-scale data, the standardized transformations on the small time-scale data, together with the probability distribution modeling. The clustering results are cross validated with the theoretical models from the Meteorology field, and the effectiveness of the proposed methods is verified through an example of day-ahead PV power forecast.