In recent years there have been growing concerns over pollution, energy security, rising energy prices and, in particular, how greenhouse gas (GHG) emissions are contributing to climate change. The world's reliance on fossil fuels for generating electricity is unsustainable.Several studies have shown that it is feasible for Australia and other countries to use renewable energy for 100% of electricity generation. Many of these studies make simplifying assumptions, such as hourly or daily time scales and low spatial resolution. These assumptions are adequate for first approximations, but are not accurate enough for detailed system design.This thesis develops methods for creating renewable energy data with realistic variations at high temporal and spatial resolutions and sufficient accuracy for detailed system design.We address the need for solar irradiance data with high temporal resolution by developing a method for generating synthetic five-minute pairs of global horizontal irradiance (GHI) and direct normal irradiance (DNI), interpolated from hourly mean values. Our method can be applied to locations where only hourly data is available. The results have been published in [3,4].We address the need for short-term probabilistic forecasting of solar irradiation. Renewable energy sources such as wind and solar irradiance are inherently intermittent, and so their integration into an electric power grid requires accurate and reliable estimation of uncertainties. We present a new data-driven method for constructing a full predictive density of solar irradiation, based on a nonparametric