Surrogate modelling and optimisation in EnergyPlus environments for smart buildings

The work of this thesis is linked to the design and management of intelligent buildings supporting advanced building simulation features. Smart buildings may utilise various existing technologies and are planned or modified so that future technological improvements may be integrated. The specific goal of this thesis, part of the educational activities of the PRELUDE project, is to add functionality to the PREDYCE software (Python semi-Realtime Energy DYnamics and Climate Evaluation), which is a Python library developed inside the PRELUDE and E-DYCE projects to work as an EnergyPlus simulation platform, allowing automatic editing of IDF files (building models) and KPIs computation on both simulation results and monitored data. These additional functionalities are optimisation methods that use genetic algorithms (e.g., NSGA2) to solve multi-objective problems, as well as "black-box" simulation methods (surrogate models) that use the design and training of artificial neural networks (ANN) to simulate or predict the KPIs of a building, using the building's structural parameters or external climate data as input. These functions have been designed to be helpful during the design phase of a building but also as a tool in the context of intelligent buildings, helpful in terms of saving consumption and comfort management since the implementation of artificial neural networks allows for a large number of simulations to be performed almost in "real-time."