Smart Water Auditing for Hong Kong

Principal Investigator: Dr. Edith C.H. NGAI (Associate Professor from Department of Electrical and Electronic Engineering)

This project is showcased in the second exhibition – Digitization in Innovation Wing Two

About the scholar

Dr. Edith C.H. NGAI

Research interests:
• Internet-of-Things
• Machine learning
• Data analytics
• Smart cities
• Smart health
Email: chngai@eee.hku.hk
Website: https://www.eee.hku.hk/~iotlab/index.html

Project information

In the past 20 years, many international cities have registered a steady decline in per capita domestic water consumption level. Hong Kong, however, has wandered in the opposite direction.

To find out why Hong Kong has become an outliner in per capita domestic water-use, an inter-disciplinary research team at the University of Hong Kong, through conducting a “Smart Water Auditing for Sustainable Hong Kong” project, is going to figure out the composition of domestic water-use, which, in turn, will help us understand how Hong Kong people use water at home and point to solutions to the problem of over-use.

The research team has developed a Smart Meter Analyser (SMAN, in short). By fastening a SMAN onto a government-issued billing meter, the billing meter’s data can be read automatically and digitized by an edge-AI optical character recognition model.

Household level data, with the help of U-Net, a deep learning algorithm, could then be disaggregated into specific end-use categories, such as showering, basin-tap, kitchen-tap and washing machine.  

Participants can log on to the Project’s mobile website to track their own water usage level and the composition of that usage at anytime, anywhere. This study harnesses the power of IoT technologies, data analytics and advanced household profiling techniques to generate insights to assist policymakers to formulate more effective water conservation measures.

Project video
Project images
The Project Team has developed a tap sensor and installed it at every water outlet in 21 households to collect ground truth data for 6 months.
Unlike conventional classification models, U-Net can extract from a ground truth dataset vital information for the task of water end-use categorisation, such as showering, kitchen tap, and washing machine.
A non-invasive Smart Meter Analyser (SMAN) clamped onto a government-issued billing meter. Total water-use data can be recorded by using an edge-AI OCR model built into a SMAN
A SMAN, together with the U-Net model can produce an accurate, and nuanced understanding of how much water people are using, when they use it, and for what purposes, at home.
Achievement of the Project
Enquiry / Feedback

Please feel free to give your enquiry / feedbacks to the research team by filling the form (https://forms.gle/JV59N47nTj19ndYz6). Thank you!