Controlling a House’s Air-Conditioning Using Nonlinear Model Predictive Control

As per literature, there is a potential for energy savings of (5% to 20%) using building embedded control systems. This letter aims to control a house air-conditioning system using model predictive control (MPC) with a neural state space prediction (NSSP) Model to maintain interior temperature set-point and reduce energy consumption. Here, we present a high-fidelity model that is validated by an experimental prototype of a house air-conditioning system that is controlled using a nonlinear MPC. The house air-conditioning system models the air-conditioner and the thermal dynamics of the house. The outdoor temperature is modeled by simulated signals and real measurements. The controller problem is to maintain a house temperature within 20 °C to 22 °C and to minimize energy costs. Compared with the generic nonlinear MPC controller, multistage nonlinear MPC provides a more flexible and efficient way to implement MPC with staged costs and constraints.

I. Introduction

According to recent studies, over 40% of all building energy is used for building ventilation, cooling, and heating. According to current related studies, there is a potential for energy savings of between 5% and 20% using building control systems [1]. Because variable renewable energy sources are being used more and more, managing demand and supply in the power system is getting harder. Buildings can better utilize potential energy flexibility from the passive thermal mass by using model predictive control (MPC). The heterogeneous character of the building stock, however, makes it difficult to design computationally tractable control-oriented models that accurately depict the intricate and nonlinear thermal dynamics of particular buildings [2], [3].