Decentralised Predictive Control of Multi-Energy Resources in Buildings

This work concerns model predictive control of multi-energy buildings with shared network capacity constraints. More specifically, it is concerned with how to effectively coordinate a large number of locally constrained and coupled buildings without violating indoor comfort and network constraints, whilst maintaining the privacy of building owners. Buildings are regarded as agents and modelled using a hybrid system approach within a multi-energy context. The developed models consider buildings connected to constrained electricity and gas supply networks. Building energy resources can include battery devices, heat-pumps or micro-turbine based combined heat and power units to satisfy both the local electricity and heating demands. The building and network modelling is incorporated into a predictive control scheme to optimally coordinate multiple buildings, with the objective of minimising individual building gas and electricity costs. However, for large systems consisting of many buildings, the computation time required to coordinate and optimise building operations in a centralised manner becomes prohibitive. Hence, in order to devise a scalable control framework, a decentralised approach is employed, where each building agent is required to solve a tractable Mixed Integer Linear Program at each time step. By doing so, multiple buildings are cost-effectively coordinated and operated so as to achieve a common target with significantly reduced computational time. The proposed approach can handle an arbitrarily large number of buildings and is validated through a relevant case study.