[object Object]

by Dr Mikkel Lykkegaard

Updated 26 September 2023

autoEPC: A Step Change for Energy Efficiency

Using Machine Learning to Predict EPC Scores
[object Object]

According to the European Commission, household energy consumption accounts for roughly 20% of total CO2 emissions in Europe, and heating losses are responsible for approximately half of this figure. Hence, improving the energy performance of houses, as reflected by the EPC rating of a property, is critical to reaching net-zero goals. From a regulatory perspective, all existing tenancies in the UK must be at least EPC Band C from 31 December 2028. Additionally, landlords will be required to give lettings agents an updated and compliant EPC certificate for the property they are advertising from 2025 and onwards, and they will not be permitted to advertise their property if it does not have a rating of C or above.

It is essential to understand buildings' fabric performance in order to reduce their environmental footprint and incentivise projects that align with net zero goals and government regulations. Existing EPC certificates are often inaccurate since they rely on the assessments of surveyors, which are often only approximate, whilst sensor setups combined with long-term controlled testing aren't cost-effective for scaling across domestic and industrial markets.

autoEPC is an easy-to-use machine learning based solution that uses self-learning algorithms to provide accurate fabric performance with no specialist equipment, installation or monitoring. autoEPC is targeted at 3 broad customer groups, namely homeowners who want to know what they can do to improve the EPC rating of their house without having to hire an expensive surveyor, developers who want to evaluate large-scale planning scenarios quickly, and landlords, such as councils, housing associations and large private landlords, who want to assess which of their properties are in need of improvement.

Machine Learning to the Rescue!

autoEPC is a cloud-based software application, which allows assessing property EPC ratings fast, cost-effectively, and reliably, with minimal input from the user. It is built on digiLab's industry-leading machine learning platform twinLab, providing robust, data-driven predictions with Uncertainty Quantification (UQ). twinLab's unique UQ capabilities provide actionable insights allowing you to make decisions with confidence.

Under the hood of autoEPC, the software collates data from multiple different sources, including Ordnance Survey data, building floor plans, and data from the existing national EPC building register. These data are combined and then compressed, to allow for making accurate EPC score predictions including rigorous UQ. The EPC score predictions also include suggestions of how to improve the EPC score as cost-efficiently as possible, so that you can make sure that your home-improvements are reflected in your energy bill.

The Future of Energy Efficiency

At digiLab, we are committed to making a low-carbon future a reality. This journey begins at the local level: the houses where we live. The current EPC rating process is awkward and expensive, and makes it hard for homeowners and landlords to adapt to the changing climate and energy market. autoEPC allows for fast and accurate evaluations, for single houses or for entire portfolios, taking the pain out of planning future-proof house renovations.

Dr Mikkel Lykkegaard
Principal Scientist, digiLab
Mikkel leads the Data Science team, creating bleeding edge tech to solve real engineering problems. His specialty is finding the right machine learning solutions for a specific engineering problem whether that's control, prediction, risk assessment or inference. When he's not cracking open a new dataset and analysing it, you'll find him camping in the wilds, bouldering and cooking.

Featured Posts

If you found this post helpful, you might enjoy some of these other news updates.