Reducing energy consumption for buildings under system uncertainty through robust MPC with adaptive bound estimator

Abstract

Model Predictive Control (MPC) has emerged as an alternative to traditional control method to reduce building energy consumption. With the presence of model uncertainty, such as mismatch between the plant and control-oriented model, the use of MPC may result in thermal comfort violation or energy waste. The influence of model uncertainty becomes even more significant as the size and complexity of the investigated building increase. Robust MPC (RMPC), which requires knowledge on the system uncertainty, has been investigated for enhancing the stability of MPC. However, the implementation possibility of the RMPC is prevented by increased computational burden and conservativeness of controller performance. This paper deals with the latter issue by presenting a novel adaptive RMPC scheme for temperature regulation in commercial buildings. The novelty comes from the development of a comparison model built based on a nonlinear autoregressive model for worst-case analysis. This comparison model enables us to transform a linear, robust MPC problem into an adaptive one with a time-varying uncertainty bound. The proposed method is tested on a simulation model developed from building data collected from a spacious hall at an airport terminal. By conducting simulation using different MPCs, it is found that the proposed RMPC method is able to behave robustly against uncertainty with the least performance loss. This means the maximum energy saving and the least thermal comfort violation.