I am undertaking PhD research as part of the Water Informatics: Science and Engineering (WISE) EPSRC Centre for Doctoral Training. My research is based at the University of Exeter within the Centre for Water Systems.
Thesis title: Artificial neural networks for the model predictive control of urban drainage networks & wastewater treatment plants.
Rapid urbanisation and changing climates are overwhelming our drainage networks. Treatment plants can alter their operation for high flows, but need advanced warning of flows and pollutant loads. Traditional models are too slow to respond, so we need to develop a fast and accurate surrogate.
Artificial Neural Networks are data-driven machine learning models that, whilst slow to train, are exceptionally fast to deploy. They can be trained on measured or synthetic data, using powerful open source tools. An accurate ANN that keeps training whilst it’s used, and can be run multiple times to provide a range of forecasts, would be the perfect surrogate.
- Train and evaluate a range of ANNs as surrogates
- Develop ensemble methods to ascertain forecast uncertainty
- Deploy the ANNs such that they can keep learning “on-the-job”
- Implement the ANNs within a Model Predictive Control framework
- Utilise the novel MPC strategy within a real world drainage system