Generating synthetic wind speed scenarios using artificial neural networks for probabilistic analysis of hybrid energy systems

Abstract

Hybrid energy systems (HESs) have been proposed to include and co-optimise multiple energy inputs and multiple energy outputs to enable increasing penetration of clean energy such as wind power. To optimise the system design, extensive datasets of renewable resources for the given location are required, whose availability may be limited. To address this limitation, this paper proposes an innovative methodology to generate synthetic wind speed data. Specifically, artificial neural networks are adopted to characterise historical wind speed data and to generate synthetic scenarios. In addition, Fourier transformation is used to capture the characteristics of the low frequency components in historical data, allowing the synthetic scenarios to preserve seasonal trends. The proposed methodology enables the possibility of Monte Carlo simulation of HES for probabilistic analysis using large volumes of heterogeneous scenarios. Case study of probabilistic analysis is then performed on a particular HES configuration, which includes nuclear power plant, wind farm, battery storage, electric vehicle charging station, and desalination plant. Wind power availability and requirements on component ramping rate are then investigated.