Dissertation
My dissertation was focused on forecasting solar energy. I developed a machine learning model in TensorFlow that used satellite weather imagers to forecast irradiance. We compared this method do a number of ML models that used ground based weather data.
Abstract:
Adoption of renewable energy sources is a key step in curbing the effects climate change. Moving away from traditional energy generation processes presents many challenges. Unlike coal and natural gas many renewable energy sources and specifically solar operate intermittently and have abrupt changes in output. This variability presents a problem both for the generators, trying to meet contractual obligations, but also the grid operators endeavouring to ensure a consistent load. Having a system capable of producing accurate predictions of solar irradiance would be of use to both parties. This project aims to build such a system. Leveraging advancements in both the domains of machine learning and big data processing.
The full paper can be read here.