Portable model predictive control service for HVAC systems using semantics-aware architecture

Model Predictive Control (MPC) is a promising approach for improving the energy efficiency of building operations. However, its adoption is often hindered by bespoke engineering practices and the substantial manual effort required for configuration and deployment. Recent research has explored the use of semantic data modelling and ontologies to standardise metadata and enable portable control applications across diverse building systems. Despite this progress, most existing applications are still rule-based and do not support more advanced strategies such as MPC. This study introduces a modular and portable MPC service for Heating, Ventilation, and Air Conditioning (HVAC) systems utilising a semantics-aware architecture. The proposed service-oriented approach leverages semantic data to streamline the deployment process by automating the instantiation of pre-defined system models, populating constraints within a standardized optimal control problem, and adapting weight matrices in the objective function from a baseline reference case. This process enhances portability and reduces the need for manual intervention. The framework is validated in an office building equipped with a radiator heating system and a Variable Air Volume ventilation system. Instances of a representative MPC algorithm are automatically configured and deployed across four rooms, each with distinct technical layouts. Both simulation and real-world experiments confirm the effectiveness of the approach, demonstrating consistent, energy-efficient control performance. The study shows that semantic data can be used to automate the instantiation and deployment of an MPC algorithm with a pre-defined structure. By improving both algorithmic and semantic portability, the proposed framework facilitates the reuse of MPC applications across similar types of building energy systems, reduces engineering overhead, and supports the wider implementation of advanced control strategies in practice.