Comfortable Achievement in Complex, Multi-zone Environments of A HVAC-VRF System by Multi-objective Function of Data-driven Neural Network Model Predictive Control

Conventional control strategies, such as fixed-setpoint and PID controllers, often fail to achieve an optimal trade-off between energy efficiency and occupant comfort in multi-zone buildings due to the complex, nonlinear thermal dynamics and strong thermal coupling inherent in such environments. To address this problem, this paper presents the design and simulation of a Neural Network Model Predictive Controller (NNMPC) for a multi-zone VariableRefrigerant Flow (VRF) HVAC system of a campus building at Universitas Indonesia. A data-driven predictive model using an Artificial Neural Network (ANN) was developed to capture the thermal-building dynamics and integrated into a Model Predictive Control (MPC) framework. Within thisframework, a multi-objective Genetic Algorithm (GA) optimizes control actions by identifying the knee-point solution on a Pareto-front to concurrentlyminimize energy consumption and thermal discomfort, quantified by the Predicted Mean Vote (PMV) index. The system was implemented in a highfidelity, software-in-the-loop co-simulation of EnergyPlus and Matlab. Whenbenchmarked against a standard 24°C fixed-setpoint controller, the NNMPCachieved energy savings of 10.25% and 6.91% in the energy-prioritized and balanced scenarios, respectively. Conversely, the comfort-tracking scenarioresulted in a 5.94% energy increase to prioritize thermal comfort, showcasing the controller’s flexibility. In all cases, occupant comfort was maintainedwithin the limits defined by the ASHRAE 55 standard, confirming the proposed controller’s capability to effectively manage the energy-comfort tradeoff. Manuscript currently under review/preparation. Full text available upon request for academic or scholarship verification purposes.