A bilinear approach to model predictive control for thermal conditioning of adaptive buildings

The high resource and energy consumption of the building sector in both construction and operation is a growing problem worldwide. The largest contributor to operational energy consumption is thermal conditioning of the indoor space. In this context, inefficient control algorithms or parametrizations become a serious problem requiring thermal simulation models of buildings for system sizing and control parameter adjustments. However, the high complexity of the underlying dynamic models makes the design of model-based controllers difficult. Furthermore, typically used control schemes such as PI-control cannot incorporate all types of actuators that an adaptive building may provide. In this work, we derive a bilinear thermal model for adaptive ultra-lightweight buildings from the linearized model output of the Modelica library BuildingSystems by incorporating environmental and internal disturbances as well as a number of possible actuators for an adaptive building into the model as time-varying bilinear inputs. Based on the bilinear model, a model-predictive control algorithm is devised that incorporates disturbance forecasts. Exemplary simulations for a summer day show the efficacy of the control algorithm in employing indirect actuation.