The adoption of Artificial Intelligence (AI) and Big Data analytics for energy efficiency is not a new phenomenon in the traditional building electrical and mechanical (E&M) industry. But in a global city and financial centre like Hong Kong, the increasing effects of climate change and the urgency to work towards carbon neutrality in the coming years mean that more innovative ways of conserving energy are vital.
The Hong Kong Government, through its Electrical and Mechanical Services Department (EMSD), identified Semantic AI as an important component in developing knowledge-based systems - including the traditional building E&M industry - that can increase efficiency and promote sustainable solutions in smart buildings.
City-level management - Regional Digital Control Centre (RDCC)
Smart buildings today are equipped with advanced building automation systems, consisting of a network of sensors and digital controllers that collect real-time measurements. With a portfolio of more than 8,000 facilities, EMSD established its first RDCC as a centralised data processing and modelling centre. EMSD’s RDCC utilised open data platform and AI solution, collecting up to 600,000 data from a single building each day. The ultimate goal was to enhance EMSD assets’ operational efficiency and environmental performance using Big Data analytics.
Figure 1: Regional Digital Control Centre (RDCC)
The RDCC harnessed big data to build a dashboard with three operation modes: daily, energy and disaster modes. Daily mode focused on E&M systems status, and the alarms of various sites and systems. Energy mode was tasked with the energy performance of the buildings. The system is supported by an analytic engine with AI and machine learning capability, which is used to effectively manage and improve the energy performance of its building portfolio.
Figure 2: Dashboard showing the real-time monitoring in daily mode
Figure 3: Dashboard showing the real-time monitoring in disaster mode
Figure 4: Dashboard showing the real-time monitoring in disaster mode
Disaster mode is used to monitor real-time remote plant room conditions and geographical information directly to RDCC under adverse weather. This livestream of visual images allowed for better labour coordination.
This dashboard is designed by frontline staff, engineers and senior management to monitor various assets at multiple sites. The establishment of EMSD’s RDCC created a system that solved many pain points within smart buildings by using Big Data analytics. It also worked with international and local partners to establish a portable AI service and model deployment, which was validated by implementation in building energy management.
Semantic AI innovation in Hong Kong
With the abundant data on hand, EMSD strives to utilise them in usage analysis and system optimisation. However, the data collected and disseminated can vary greatly from building to building. This creates challenges in the mass deployment of AI applications in multiple buildings - Semantic AI is one tool that can be used to overcome this.
Traditionally, it has been challenging for engineers to handle multiple sets of data from different buildings, with significant labour required to pre-process these data. As EMSD recognised, Semantic AI can be utilised to make data machine-readable, saving time and effort to process, which frees up experts to create applications to improve building efficiency.
Together with experts from international institutes and universities, EMSD has developed a semantic data platform (Figure 5) which consists of a graph database for storing the semantic model and a time-series database for building data. The building data associated with the semantic model (Figure 6) is ingested into a time series database from the big data server. Entity is reachable through Semantic API path query. Data query can then be formed by the semantic data platform to retrieve the time series data that are linked to the entity. This enables different domain experts to perform analytics and diagnostics separately, making the programmes of the AI services portable across buildings within a shorter timeframe.
Figure 5: Architectural overview of semantic data platform
Figure 6: Water-cooled chiller system at a government building (partial view)
Semantic AI addresses the need in the traditional building E&M industry for data that are readily interpretable, meaningful and machine-readable. By fusing knowledge graphs together, data scientists are able to develop Semantic models and deploy advanced AI applications to be used across multiple buildings - no matter the type of the automation system they use. This removes the need to create different AI models for each individual building.
Semantic AI enables subject matter experts to really leverage data without having to process the mathematics or algorithms to understand the logic behind it, allowing them to apply it to create innovative ideas.
In the building E&M services industry, experts are able to quickly create an innovative application, and then port it to other buildings without knowing the details of individual vendors or machines. The logic behind Semantic AI is maximising efficiency, and to be quick and fast in solving industry problems. EMSD has created an award-winning Semantic AI model that improves the engineering facility in buildings, ultimately conserving energy.
Figure 7: Knowledge graph of Semantic AI model
EMSD has continued its digital transformation and commitment to its Smart City Blueprint for Hong Kong 2.0, to further promote energy efficiency and conservation, with a focus on smart buildings. Part of this strategy is to eventually achieve carbon neutrality by 2050, with Big Data analytics and AI models being crucial to this.
Global AI Challenge for Building E&M Facilities event
With the success of Semantic AI development, EMSD forged ahead - together with the Guangdong Provincial Association of Science and Technology - to share its passion and innovation of Semantic AI with the world by organising the Global AI Challenge for Building E&M Facilities - a global event highlighting AI development and applications in the building services industry. It aimed to promote international innovation and technology ideas through exchange and cooperation.
Featuring a diverse mix of events including a global technical conference and workshop, the Challenge’s AI Competition was a direct way to encourage the proposal of innovative AI solutions and inspirational AI technology in the building E&M industry.
Figure 8: Global AI Challenge for Building E&M Facilities – Technical Conference
The Global AI Challenge for Building E&M Facilities – Technical Conference was an international conference focusing on AI technologies in the building E&M industry. The conference gathered innovators from around the world who shared a common interest in practical solutions for smart cities through AI innovation. It explored how data and AI can create a smarter world, with the theme: AI applications revolutionise traditional E&M industry, laying the groundwork for smart cities.
Hosted in-person and also virtually, the conference brought together a diverse group of attendees including university students, professional engineers, young innovators and IT professionals, participating in guest lectures and sharing sessions. There were more than 2,300 online views globally, in addition to more than 200 on-site attendees in Hong Kong. Readers can learn more about the Challenge and also watch replays of the conference, which features industry experts from all over the world, at www.globalaichallenge.com.
The Global AI Challenge for Building E&M Facilities - AI Competition was open to participants from all over the world, where teams were tasked with developing a Semantic AI model to predict the cooling demands of a commercial building. The competition presented a unique opportunity to inspire participants, industry leaders, innovators, and researchers to exchange ideas, and to progress the role of AI technology so that it has a positive impact on the world.
The competition proved to be popular, with more than 120 teams from ten regions all over the world vying for over US$200,000 in prizes.
The Challenge’s organising committee member Professor Linda Xiao, from The Hong Kong Polytechnic University’s Department of Building Environment and Energy Engineering, said the division of the competition into two categories - Academic and Open - enabled a more diverse mix of participants.
“This is a very valuable opportunity to involve different participants in this competition. By dividing it into two streams [Academic and Open], we are able to engage different parties in the industry,” she said.
Figure 9: Global AI Challenge for Building E&M Facilities – Workshop
Set in three parts, the competition began with a workshop outlining EMSD’s expertise in Semantic AI, with teams then completing a written questionnaire about the benefits of Semantic AI in building management. Dang Quang Minh, a student from Vietnam’s Hanoi University of Science and Technology, said this section of the competition was beneficial in harnessing expertise in Semantic AI.
“The programme expanded our knowledge of Semantic AI through the workshop session. Moreover, we were also provided with the Energon tool which helped us to understand Semantic AI a lot better because we could actually use it by ourselves.”
In the second part of the competition, teams were then tasked with developing an AI model to predict the hourly cooling load of a commercial complex, using authentic data from a multi-chiller system of a commercial complex provided by EMSD.
Huang Yixiao, a student in City University of Hong Kong’s team in the Academic category, said the example of actual real-world data was useful.
“This programme gives us an idea of how well-understood and well-structured data can guide the semantically precise model construction to improve daily operational efficiency.”
Lastly, shortlisted teams from Part B entered the final round of the competition, where they submitted pre-recorded presentation videos to present their work, which formed the basis of the judging panel.
Dr Dan Wang, Organising Committee Member and Professor at the Department of Computing of The Hong Kong Polytechnic University, said he was impressed by the calibre of the Semantic AI models submitted.
“The Semantic AI models were very skilful, the quality was really high and they were very professional in using multiple algorithms. Young teams like challenges and they truly exceeded my expectations,” he said.
Chan Ming-chung, a research scientist in Arch & Fire Professional International Limited’s team in the Open Category, welcomed the competition as an introduction into Semantic AI and its benefits.
“Before the programme, I had almost no knowledge about Semantic AI. But afterwards, I deeply understand the value of Semantic AI in streamlining the deployment of a prediction model in different types of buildings,” Chan said. “It also let me know how Semantic AI works conceptually and how to deal with it in a real situation.”
The popularity and broad participation of the event was evident, with more than 80 local and international organisations supporting the global event.
Kin Tsang, Organising Committee Member and Chief Innovation Advisor at Logistics and Supply Chain MultiTech R&D Centre, said the success of the event was seen in the widespread participation from all over the globe.
“The fact that we can have more than 100 teams from all over the world and locally - it is a very successful event indeed. For any innovation to get off the ground, industry participation is the key. If you look at this event, there are so many industry partners and academia joining in. This is one key element that made the event a success,” he said.
Professor Xiao agreed: “It engaged a lot of researchers and companies and it’s very good to promote Semantic AI and AI applications to address carbon neutrality challenges in the industry.”
Another Organising Committee Member, Ir Prof Samson Tai from Hong Kong Baptist University, said he was looking forward to the next event.
“We really would like to make the event annual or biannual - this is a good event in terms of attracting talents from different parts of the world and local universities to participate and generate ideas,” he said.
“I received feedback from students to the effect that they really learned a lot by participating in this AI challenge. It was great that we provided real problems with real data for competition entrants.”
Future of Semantic AI in Hong Kong and the world
Hong Kong was in a unique and favourable position to promote, nourish and accelerate Semantic AI, benefiting the city as a whole.
“We have all these elements and it’s a matter of initiating and putting knowledge together to collaborate with academics around the world to be the centre of this Semantic AI innovative field,” Mr Tsang said.
“I do hope Hong Kong can become a great technical leader in Semantic AI.”
Semantic AI is still a relatively new phenomenon in the industry, with the majority of developments happening outside of Hong Kong – in North America, Singapore and Europe. But with the promotion and support from the Hong Kong Government through events like the Challenge, as well as industry support, there is hope that Hong Kong can be a driving force in Semantic AI in the next few years.