La Trobe University wrangles a billion data points with cloud and AI to achieve carbon neutral goal

If you can’t measure it, you can’t manage it – but what if you’re measuring a billion data points from 200 data sources – how do you manage that? La Trobe University has found a way, using cloud, analytics, automation, and AI, and in the process is helping the institution achieve its goal to be carbon neutral by 2029.

La Trobe has significantly invested to achieve the goal and its five regional campuses are set to be carbon neutral in 2022. One key investment is the La Trobe Energy Analytics Platform (LEAP) which takes the billion data points from 200 different sources and then uses Microsoft Azure and Artificial Intelligence to create actionable insights that allow facilities managers and individuals to take informed action to rein in energy use and emissions.

Designed and developed by La Trobe University’s award-winning AI research team in the Centre for Data Analytics and Cognition (CDAC), in collaboration with Microsoft experts, the platform monitors consumption patterns, renewables generation and building performance and creates easy to understand dashboards that show how a building or campus – or indeed the whole university – is performing (or behaving). They have also developed an AI infused chatbot designed to allow anyone on campus to ask questions of the data to find out how La Trobe is tracking towards its net zero goal.

Andrew Jennings is director of the University’s Carbon Neutral Strategy, Infrastructure and Operations Group, and explains that La Trobe’s approach has been; “Primarily on leveraging assets of the university whether that be human capital or land assets or stakeholders, or students to try and achieve net zero.”

Solar PV has been deployed on all campuses, and LED lighting rolled out, and now LEAP is providing valuable insights about how buildings are performing against expectations. By consolidating all one billion data points (many collected every 15 minutes from 200 different and often proprietary sources) into an Azure Data Lake, La Trobe now has a vast reservoir of information that it can use for actionable energy insights.

Using historical data, CDAC has developed algorithms that predict how individual buildings should be  “behaving”; variations are automatically detected and compared with similar buildings to see how/where optimizations can be made.

Besides taking data sources from meters and rooftop solar, from temperature and humidity gauges, La Trobe’s data lake ingests data from the Bureau of Meteorology and feeds in a real time stream of satellite observations for solar irradiance data. The CDAC AI team then combine these data with overlays of people movement, time of the year, whether it is term time or holidays, and then create AI models that predict energy usage and the ‘micro-seasons’ of a building.

In one recent case LEAP identified a building that was operating outside of its expected limits – when the facilities manager investigated it turned out that a new chiller had been installed, but software that controlled overnight and weekend usage hadn’t been turned on – the chiller was running flat out night and day. A simple software update and the problem was fixed.

Dr Daswin De Silva is associate professor at CDAC, and one of the leads on LEAP. Besides creating an important insights platform for the university that supports its mission to rein in emissions, De Silva explains that the program is a gamechanger for the eight PhD candidates working on the initiative. Through CDAC’s LEAP work they have access to real-life, real-time data, and the opportunity to not only develop new deep learning algorithms and working artificial intelligence models, but also to deploy and operationalize AI.

“LEAP is a flagship AI project – it works as a training ground for new PhDs who can work alongside academics and more senior PhDs to acquire state of the art understanding of what it means to wrangle Big Data, build AI and actually deploy for practical applications. It’s the real deal – we have a billion data points – internal and external data – consumption, generation and optimisation data, and we work together on advancing cutting edge AI, such as transformers, autoencoders, swarm intelligence, and multi-agent reinforcement learning” says De Silva.

Jennings adds; “We are providing a platform for students to actively learn in and gain real world insights – the team that comes out of the group are eminently employable and move onto roles with some of the largest corporates in Australia because they have had access to the real world challenges and complexities of Big Data and AI.”

He explains that the solution has now been deployed and is being used by La Trobe’s facilities management team to identify and help investigate variations in consumption patterns that ultimately will reduce consumption and emissions. LEAP will also allow the team to measure and validate the impact of any changes they make to a building or campus. “The LEAP platform allows you to make that change, mark the day, and then it will compare what consumption is both pre and post to tell you how effective the project has been and puts a value on it in emissions in dollars and in kilowatts,” says Jennings.

That insight can be used to build new business cases for further building innovation.

LEAP is also used in conjunction with Power BI to streamline the University’s emissions reporting requirements that are essential to it meeting its carbon neutral goal.

The chatbot that allows people to interrogate LEAP has been developed using Azure tools and turns the billion data points into 1,000 insights that a user can inquire about. Jennings says it’s an important step in engaging the whole La Trobe community, in the University’s sustainability plans.

“Our goal with net zero is to get people engaged with the entire program and journey – through static reporting to enable them to see progress, or more dynamic items – students, stakeholders, staff – anyone on the University platform is able to ask a question of the chatbot to see how we are progressing to net zero.”

De Silva says that the selection of Microsoft as the digital platform for both chatbot and LEAP was based on;

The technical advancement of the Azure platform, and products and services – coupled with the agility of transforming them to building blocks that generate immediate practical value was crucial. It ticked the boxes in technicality and how quickly we can turn this around and build our platform.

“LEAP uses Azure SQL Server, Azure Data Lake, Azure Blob Storage, Logic Apps, Azure Monitoring and Azure Machine Learning. CDAC was also supported by Microsoft with weekly meetings with Microsoft FastTrack engineers (Azure’s customer-facing team who engage directly with clients) who were able to advise about optimising the infrastructure and troubleshooting, as well as providing reference implementations and code,” says De Silva.

Jennings says that while LEAP is delivering significant benefits to La Trobe already it was always developed with an eye to commercialisation, as the same challenges in terms of wrangling large volumes of data from disparate systems are faced by other Universities, hospitals, or large property fleet managers. “Our goal is to have something that is easy to deploy. That’s where a significant amount of Microsoft’s input has been, in getting the architecture right – and how to get all that data into usable format. We would like to share the information and our project with others in a commercial fashion.”

LEAP is already winning accolades, having secured the innovation award from the Tertiary Education Facilities Management Association (TEFMA), while CDAC and La Trobe have collaborated with energy AI researchers at Universities in Sweden and Finland which are keen to understand how LEAP’s AI and data lake could help them further improve their own sustainability and accelerate the world’s migration to net zero.

• Azure Data Lake
• La Trobe University