Close
Thesis title: The smart controller: optimising power grids with machine learning
Primary supervisor: Ilkka Keppo
Secondary supervisor: Aidan O'Sullivan
Start date: September 2017
Completion date: September 2021
The rapid growth of renewable energy sources is a vital means to achieving climate goals set in Paris in December 2015. Nevertheless, exploitation of many renewable resources, notably wind and solar, is hindered by their inherent uncertainty and risk of destabilising the grid. Facilitating the diffusion of renewables into the UK’s electricity mix will require increasingly accurate forecasts, as well as intelligent solutions to unit commitment and economic dispatch.
This research will aim to develop a ‘smart controller’: an agent capable of optimising dispatch of electricity on many time-scales, including the crucial ‘balancing’ of the grid at very short notice. The research will employ state of the art techniques in reinforcement learning to train the smart controller, based on disaggregate historical data for energy generation and load.
Biography
Patrick completed his undergraduate degree in Arts and Sciences at UCL in 2017, taking courses in Mathematics, Physics, French. His undergraduate dissertation aimed to measure the development of topics relating to energy in UK print media using Latent Dirichlet Allocation. Exposure to machine learning methods as applied to energy led Patrick to pursue postgraduate research in this area.
