AI system predicts emissions rates in under a second

Dr Georgina Cosma and postgraduate student, Kareen Ahmed, of the university’s School of Science, have designed and trained an AI model to predict building emissions rates values with 27 inputs with little accuracy loss.

The proposed AI model was created with the support of Cundall’s head of research and innovation, Edwin Wealand and trained using large-scale data obtained from the UK government energy performance assessments. The system can generate a building emissions rate, or BER, almost instantly.

Dr Cosma said the research was “an important first step towards the use of machine learning tools for energy prediction in the UK” and also shows how data can “improve current processes in the construction industry”.

The Artificial Intelligence Model

The AI system they have revealed in their latest paper is able to generate building emissions rates for non-domestic buildings in less than one second and with as few as 27 variables and little loss in accuracy. It works by using what they call a ‘decision tree-based ensemble’ machine algorithm. They built and validated the system using 81,137 real data records that contain information such as building capacity, location, heating, cooling, lighting and activity.

The team also focused on calculating the rates for non-domestic buildings - such as shops, offices, factories, schools, restaurants, hospitals and cultural institutions - because these are some of the most inefficient buildings in the UK in terms of the amount of energy use.

The findings were presented at the Chartered Institution of Building Services Engineers, or CIBSE) Technical Symposium 2021 and will be published on the CIBSE’s website later this year.

The aim to reach net zero

Dr Cosma said that studies on the applications of machine learning on energy production of buildings already exist, but currently they are limited and only make up eight per cent of all buildings. She added non-domestic buildings account for 20% of the UK’s total CO2 emissions.

And Wealand went on to say that he hoped to build on the techniques developed in the project to predict real operational energy consumption. He said: “By predicting the energy consumption and emissions of non-domestic buildings both quickly and accurately, we can focus on the more important task of reducing energy consumption and trying to reach net zero.”