Evaluation of linear and nonlinear system models in hierarchical model predictive control of HVAC systems

Buildings are responsible for one third of the global final energy consumption. Model predictive control (MPC) can reduce their energy consumption and improve thermal comfort. However, designing the required models can be time consuming. Splitting the control problem into smaller subproblems could make the modeling process more modular and therefore cheaper. A hierarchical MPC structure is proposed in this work, where the building model is divided into a lower layer consisting of the producer side and an upper layer consisting of the consumers. Linear and non-linear model equations as well as a cost-based and a control quality-based cost function for a building energy system are developed. In a simulation, the nonlinear controller outperforms the linear controller in both constraint satisfaction and energy costs.