Learning Dominant Usage from Anomaly Patterns in Building Energy Traces

Building energy usage is growing at a rapid pace under increasing urbanisation tendencies in both the developing and the developed world, at high environmental and social costs. Decentralised control architectures for local energy grids are seen as a key solution to optimise energy management at the local level. As the infrastructure for data collection, communication and embedded computing becomes more capable, new online algorithms can be deployed for forecasting and anomaly detection of large consumers. Fine grained tendencies and unusual artefacts can be thus exploited to improve local and grid level energy management. Our two-fold approach first leverages the Matrix Profile technique for time series data mining to build a dataset of anomaly patterns from public building energy traces and extract analytics information. Subsequently the labeled dataset is used in a supervised learning classification model to discriminate between various related dominant usage patterns. The case study is carried out on a public dataset of academic buildings. The approach can prove useful for exploiting complementary energy consumption patterns in a decentralised control structure towards grid balancing and economic operation.