Spatiotemporal dependence modeling of wind speeds via adaptive-selected mixture pair copulas for scenario-based applications

Abstract

The increasing penetration of wind power generation brings significant challenges to the operation and planning of power systems. Appropriate uncertainty modeling of wind speeds is critical to ensure the reliability of optimal decisions, which requires special consideration of spatiotemporal coupled interdependence between wind speeds. However, using only a single type of function cannot fully describe the potential complex dependence structures in historical data, especially in high-dimensional cases, which may lead to serious dimensionality disasters of model. In this paper, a novel spatiotemporal dependence modeling method of wind speeds is presented to flexibly capture the underlying irregular dependency relationships by creatively introducing mixture pair copulas into C-vine structure. The model selection and parameter estimation of mixture pair copulas are carried out adaptively through iterative optimization in expectation maximization (EM) algorithm. Furthermore, a two-step spatiotemporal wind speed scenario generation method is developed based on the constructed model. Experimental results show that the model established by our proposed method can more accurately characterize the spatiotemporal dependence between wind speeds and generate scenarios consistent with the distribution of historical data.