Paper Conference

Proceedings of eSim 2022: 12th Conference of IBPSA-Canada


Fault detection in residential buildings using sub-hourly meter data

Gilbert Larochelle Martin, Benoit Delcroix, Simon Sansregret, Ahmed Daoud
Laboratoire des technologies de l'énergie (LTE), Hydro-Québec, Shawinigan, Québec, Canada

Abstract: The current deployment of smart meters and home energy management systems provides utilities’ residential customers with more customer-centric energy data than ever before. This data can be used to provide residential customers with meaningful insights into their energy consumption and peak power demand. In this paper, machine learning techniques and smart meter data are leveraged to identify personalized periods of over or under consumption which could be indicative of abnormal energy consumption or detrimental occupant behaviour. The lower and upper centiles are modelled using a feed-forward neural network trained using a quantile loss function. The model is then applied to a set of sub-metered residential houses to assess its performance. This model could be used to provide residential building occupants with personalized energy feedback aimed at improving their energy-related behaviour and lower their peak demand and energy consumption.
Keywords: ANN, FDD, residential, quantile regression