Heroes compares buildings performance to make them more energy efficient

19 January 2022

Heroes compares buildings performance to make them more energy efficient

Knowing that Europe’s buildings are responsible for 36% of the continent’s GHG emissions Heroes is participating in the Energy Lot of AI4Cities with a solution that is using AI, IoT and Cloud technologies to reduce the energy consumption of non-residential buildings. It does so by identifying similar buildings (‘peers’), comparing their energy performance, and taking targetted action to increase the energy effeciency of underperforming buildings. Its aim is to support building owners and other relevant stakeholders in their decision making, giving them an esimation of how much energy will be spent in the next days/weeks/months.

To determine similar buildings, Heroes is using a k-nearest neighbour alghorithm model. Using variables such as Total area, Heated Area, Volume, Age, and Building type, the model looks at the distance between the meta data. The buildings with the shortest distance between them are said to be the most similar. For each building, three pears are selected, followed by an analysis of how its past energy consumption compares to that of its peers.

With this information, Heroes’ solution is able to use AI to predict the energy consumption of buildings and to recommend how to optimise energy usage based on historical and forecasting data. Furthermore it provides information on how much energy consumption was reduced after a certain upgrade. Additionally, the solution also makes anomaly detection possible, defining anomaly as an observation which deviates so much from the other observations that it raises suspicion. Finding such anomalies in energy consumption data can contribute to better control and understanding of the building processes and prevent indirect energy losses.

Finally, Heroes’ solution can also be used to optimise HVAC systems. For this, it is first needed to create models to predict energy usage, temperature and CO2-levels. These models allow for predictions based on on set points and other variables. The setpoints that reduce consumption as much as possible while maintaining comfortable temperatures and CO2-levels are then recommended.

All news