InnoCatalogue - ChillStream®: AI-based Chiller Plant Optimization System 2024-07-8

Air-conditioning system is the largest energy consumer inside a building, usually more than 50% of total building energy use. A chiller plant is traditionally either manually controlled or using rule-based controllers. However, both are unable to fully optimize the system's performance, resulting in wasting energy.   The best operating point of an air-conditioning system is achievable with a combination of influencing factors, including outdoor climate, dynamic building energy load, performance of individual equipment, etc. that involves over millions of options which is never and impossible to be optimized by human or rule-based controller.  Chiller optimization solutions in market are costly and inefficient, since they are proprietary with limited optimizing options, and they are not readily for use among different chiller plants.

ChillStream® is fully developed by the EMSD, which is a replicable AI solution for real-time optimized control of chiller plant. ChillStream® learns the chiller plant characteristics using historical data with its Artificial Neural Networks.  It then utilizes real-time building electrical power load, other building operating data, together with the forecast weather data to predict the cooling demand, and searches for the best combination of control set-points from over ten-millions of options using our self-developed data-driven AI algorithm, the GA-PSO.  At every 5 minutes, it automatically commends the chiller plant toward its optimal performance sweet spots.

ChillStream® is a long term and scalable solution. Analytics-as-a-service (AaaS) model is adopted for ChillStream® by means of AI resources which are supplied by a centralized AI hub, the Regional Digital Control Centre (RDCC), located in EMSD Headquarters. The nature of full system ownership would make sure the output is traceable for reducing operational risk. This kind of service model would significantly reduce the implementation time and cost for many more public and private venues. The flexible scaling capabilities of AaaS platform allow EMSD to quickly scale up or down the computing resources, storage, and analytical capabilities to match the varying demands.

Public Health Laboratory Centre

ChillStream® underwent pilot operation in a clinical laboratory building (Public Health Laboratory Centre) in Hong Kong with seven chillers. Initial trials demonstrated promising energy-saving potential while maintaining a stable 24-hour AI automated operation.