Machine learning for modelling and control of direct air capture systems
Lead Research Organisation:
University of Bristol
Department Name: Computer Science
Abstract
EPSRC : Stefan Radic Webster : EP/S022937/1
The placement project will investigate the use of machine learning for modelling and control of an offshore wind powered direct air capture (DAC) system. DAC remove carbon dioxide (CO2) from the atmosphere which is injected into deep-sea reservoirs and are a key negative emission technology, which are becoming increasingly vital as a climate change mitigation strategy to achieve the targets of the Paris Agreement. The problem of controlling DAC systems is difficult due to the complex operation and the intermittency of the wind power supply. A traditional approach for controlling DAC systems uses mathematical modelling, which is time consuming, requires specialist expert knowledge and is computationally expensive to execute. The project will use a data-driven approach to modelling and controlling DAC systems by drawing on the latest developments in machine learning such as deep learning and reinforcement leaning. The objective is to show the feasibility of machine learning in this setting and implement an adaptive, learning-based controller that maximises the CO2 capture rate.
The placement project will investigate the use of machine learning for modelling and control of an offshore wind powered direct air capture (DAC) system. DAC remove carbon dioxide (CO2) from the atmosphere which is injected into deep-sea reservoirs and are a key negative emission technology, which are becoming increasingly vital as a climate change mitigation strategy to achieve the targets of the Paris Agreement. The problem of controlling DAC systems is difficult due to the complex operation and the intermittency of the wind power supply. A traditional approach for controlling DAC systems uses mathematical modelling, which is time consuming, requires specialist expert knowledge and is computationally expensive to execute. The project will use a data-driven approach to modelling and controlling DAC systems by drawing on the latest developments in machine learning such as deep learning and reinforcement leaning. The objective is to show the feasibility of machine learning in this setting and implement an adaptive, learning-based controller that maximises the CO2 capture rate.
