Adaptive-fine tuning of building energy management systems using co-simulation

A model-assisted fine-tuning methodology to adapt and improve performance of energy management systems is presented: to start, a detailed building thermal simulation model acting as surrogate of the real building is required, along with a naïve controller, a “good” initial controller, or even a set of rules with tunable parameters. Given weather and occupancy predictions for a predefined time window - say a day - an algorithm is used to create candidate controller parameters, and the (co-)simulation model is used to evaluate candidate solutions. Controller parameters are updated so that a good-performing controller - in terms of a predetermined cost function - is created. This controller, adapted to the forecasted conditions and the actual building - or at-least a comprehensive representation of the real building as obtained using validated thermal simulation models - is used to operate the building until new forecasts trigger a controller-parameter update. Apart from operational performance benefits, updating controllers over short periods, means that simpler in terms of mathematical structure controllers can be used. Corroborating numerical experiments are presented to illustrate the potential of the proposed methodology.