
Robert Steven
Academic and research departments
Information and process systems engineering, Sustainable energy and materials, School of Chemistry and Chemical Engineering.About
My research project
Optimisation of Distributed Energy SystemsThis research aims to continue on the same theme as a previous project examining how Optimal Power Flow (OPF) constraints can be added to Distributed Energy Systems (DES) modelling. This project looks to build on the same modelling techniques by adding control and distributed optimisation functionality.
Supervisors
This research aims to continue on the same theme as a previous project examining how Optimal Power Flow (OPF) constraints can be added to Distributed Energy Systems (DES) modelling. This project looks to build on the same modelling techniques by adding control and distributed optimisation functionality.
Publications
Modern power grids have become increasingly complex, with greater uncertaintydue to the widespread integration of renewable energy resources potentiallyleading to higher operating costs. The optimal operation of these networks canbe accomplished using optimal power flow (OPF), a fundamental optimisationtool for power networks with objectives including generation cost minimisation.Whilst the OPF problem itself is not new, quickly solving problems of a practicalscale remains an active research area. Two approaches here are distributedoptimisation and, more recently, machine learning (ML). Distributed optimisationimproves scalability, avoids single points of failure, and enhances userprivacy, whilst ML has the potential to provide solutions significantly fasterthan traditional optimisation methods.The goal of this review is to present approaches which overlap both areas, identifyingcomplementary aspects as well as areas for further exploration. For example,one drawback of the alternating direction method of multipliers (ADMM),a distributed optimisation algorithm, is that it has slow convergence. Several reviewedpapers have mitigated this, using ML to accelerate convergence throughthe prediction of consensus variable values, demonstrating improvements interms of convergence time. Challenges remain, including the generalisation ofresults across different network topologies, something with the potential to beaddressed with additional ML models such as graph neural networks (GNNs).Further areas to explore at the intersection of these two areas are identified, includingaugmented Lagrangian alternating direction inexact Newton (ALADIN)and overlapping Schwarz decomposition optimisation methods and ML modelssuch as GNNs and physics-informed neural networks (PINNs).