»Ê¼Ò»ªÈË

XClose

Advanced Research Computing

Home
Menu

Legion unlocks home truths on health and energy efficiency

Leading-edge computer modelling at »Ê¼Ò»ªÈË is helping to pinpoint how buildings can be made more energy efficient without harming their occupants' health

Different types of housing

20 July 2018

Using the Legion High Performance Computing (HPC) cluster and Research Data Storage (RDS) service, both operated by RITS, a team at the Bartlett is developing tools designed to show in unprecedented detail how insulation and other carbon-cutting measures modify the risk of overheating and exposure to poor indoor air quality.

The ultimate aim is to help identify the best balance between saving energy and safeguarding health in UK buildings - ground-breaking work that will make a crucial contribution to shaping energy efficiency policy and guidance at both national and local level.

Building understanding: a model method

The facts and figures speak for themselves. The UK's 30 million buildings account for nearly a fifth of this country's greenhouse gas emissions. So ensuring that buildings are more energy efficient doesn't just make sense in terms of cutting energy costs and reducing the strain on supplies - it's also a vital weapon in the fight against climate change.

But is it really that simple? Energy efficiency adaptations can increase the risk of overheating in buildings - a particular concern for elderly and other vulnerable people, especially in view of projections of a rise in the number of heatwaves hitting the UK. They can also mean that buildings are less well ventilated, trapping pollutants originating both inside and outside the building.

Whenever the goal is to find out how buildings function and perform, there are two basic options:

  • Monitoring - a lengthy, laborious process whose value is lessened by the limited number of buildings that can be physically measured in a single survey.
  • Modelling - a quicker, cheaper, easier, more flexible approach enabling simulation of multiple building types and an almost unlimited range of parameters.

This second option - and specifically the development of increasingly sophisticated models capable of delivering increasingly detailed insights - is at the heart of the Bartlett's drive to shed much-needed light on the potential health effects of energy efficiency measures. But how could the effectiveness of this modelling work be maximised?

"Researchers here used to run simulation models on personal computers, with long delays waiting for results - not to mention quite a few burnt-out laptops," says Dr Phil Symonds. "To meet our requirements, we needed a step change in processing power that would enable us to explore a huge number of scenarios and cope easily with massive datasets."

Phil's previous work as a PhD student in high-energy physics at Brunel University had given him first-hand experience of the extraordinary capabilities of HPC - a solution that he realised could be applied equally effectively to the buildings sector.

Legion unleashed

One task in particular posed a real challenge in computational terms: to develop a model that expanded the capabilities of open-source building energy simulation software called EnergyPlus, and validate it using data from the English Housing Survey's Energy Follow-Up Survey of over 800 nationally representative dwellings. "I knew HPC would be essential to achieving success in this HPRU (Health Protection Research Unit) funded project. That's why I contacted RITS."

Support from RITS benefitted the project in three key respects:

  • Speedy, Sophisticated Modelling - RITS recommended that Phil register to use the Legion HPC research computing platform, whose heterogeneous architecture makes it possible to support many different types of computational workload with a mixture of general purpose and specialised nodes. To date, Phil has used the platform to conduct around 30,000 individual simulations; runs that would otherwise take months to complete are executed in a matter of days. "Quite simply, it would have been impossible to build our meta-model without Legion," he explains. "With our new model, we can now explore all kinds of climate change and overheating scenarios and assess the potential health impacts of retrofitting energy efficiency measures to the UK's entire building stock."
  • Safe Storage - Utilising the RDS service has provided peace of mind that simulation data is being securely held in a single, central location. "We arranged with RITS for the standard storage allocation of 200 gigabytes to be increased to 1 terabyte," says Phil, who also made use of the Data Safe Haven provided by Information Services Division for storing sensitive data. In future, Myriad, RITS's new platform for data-intensive computing, will offer even easier access to RDS data, as well as enhanced performance and capacity for researchers working with large datasets.
  • Specialist Support - RITS has provided additional input to help optimise the running of the project. Taking advantage of RITS's drop-in sessions and email support, Phil received guidance on working with the programming language Python; on how to run his version of EnergyPlus on Legion; on the best way to distribute data to worker nodes; and on when to opt for parallel processing and when to run calculations in series. "Expert advice from RITS meant that everything was set up as I wanted. All I had to do was press 'go'!"

Better buildings by design

Phil Symonds is in absolutely no doubt about the critical difference that HPC can make to research projects, whatever the field. "Most modelling is now at the very limit of what can be achieved using laptops and PCs," he comments. "We're at the point where using HPC, as well as dependable secure data storage, needs to be part and parcel of standard research practice."

For Phil's work, the pay-offs have been impressive: "The model we've developed using Legion enables us to model buildings and their performance much more precisely than ever before, harnessing much bigger datasets from a variety of sources and considering timeframes out to 30, 50 or 80 years. Our modelling will provide government, the healthcare sector and others with increasingly robust predictions that inform interventions and help ensure that decarbonising the UK's building stock doesn't seriously compromise people's health. HPC delivered through the specialist services of RITS is the perfect way of generating the understanding we need."

Links