Optimal Coordination of District-Scale Multi-Energy Systems using Multi-Agent Control Architecture

  • Michael Taylor

Student thesis: Phd

Abstract

A major challenge in the transition to a net-zero energy system is how to decarbonise energy use for heating, cooling and transport via electrification, whilst simultaneously ensuring the security of a power system with high penetration of renewable energy generation. One possible way to address this challenge is to implement a multi-energy systems approach, in which traditionally separate energy systems for the delivery of electricity, gas, heating and cooling are co-optimised as an integrated whole. A major benefit of this approach is that flexible distributed energy resources in non-electrical systems can be exploited in support of the power grid. This thesis presents a novel multi-energy system optimisation modelling framework, capable of quickly generating large-scale optimisation problems. These problems are readily integrated into model predictive control (MPC) schemes, providing a method for online energy management of a continuously evolving system. Such schemes can optimally manage distributed energy resources at district-scales, ensuring that all energy demands are met, networks are operated within acceptable limits and costs savings are delivered to both customers and network operators. Given the potentially large size of the resulting control problem, a multi-agent control architecture and associated coordination algorithms are also presented. These ensure that near-optimal, feasible control actions can be determined within timescales that are suitable for online energy management. In an exemplary case study, considering a 15 minute sampling interval for a district comprising 84 buildings and multiple energy supply networks, a maximum computation time of around 55 minutes for a single controller is reduced to just over 1 second using the novel multi-agent MPC scheme, demonstrating the substantial benefit of the proposed approach.
Date of Award1 Aug 2023
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorOgnjen Marjanovic (Supervisor) & Alessandra Parisio (Supervisor)

Keywords

  • Decentralised Optimisation
  • Mixed-Integer Programming
  • Multi-Energy Systems
  • Model-Predictive Control

Cite this

'