Chiller plants in commercial buildings account for a substantial portion of energy usage, playing a pivotal role in our journey towards carbon neutrality. Hong Kong as a city known for its dense urban environment and high-rise buildings, has immense potential for implementation of energy optimization strategies. Conventional chiller plant control have been heavily reliant on rule-based strategies, potentially resulting in energy wastage and imposing restrictions in flexibility of control. In this project, an analytics-as-a-service (AaaS) model was developed to enable centralized chiller plant control through an Artificial Intelligence (AI) hub located at the Electrical and Mechanical Services Department headquarters. An AI model was pre-trained with historical data of a target chiller plant for analysis of real-time plant and weather data. A hybrid optimization algorithm combining merits of swarm intelligence and evolutionary algorithm was employed. Through a bi-directional data communication network, real-time optimised control recommendations are delivered to the building management system of the target venue for chiller plant control. With an overall annual energy saving of 3-5% achieved in a round-the-clock trial at a local multi-storey clinical laboratory building, the AaaS model is poised for implementation of AI-based, dynamic solutions at scale, unlocking countless opportunities for better building energy performance.
