Smarter Buildings for the Smart Grid ? Let Them Forecast Their Power Consumption ! ( Demo

An increasing number of buildings are equipped with embedded sensing systems in order to capture what is happening within. These smart buildings process collected sensor data to increase user comfort and safety, cater for ambient assisted living, or help the residents save energy. However, saving energy is not always beneficial to the power grid, especially when renewable sources are present. More specifically, the volatile nature of their primary energy carriers (e.g., fluctuating wind speeds) may lead to situations where significant surplus energy is being generated, which must be consumed in order to keep the power grid stable. Likewise, when unexpected drops in the generation occur, utilities must react and possibly even disconnect loads. At present, grid operators only react to the observed power consumptions, and the efficacy of the measures taken to maintain grid stability is moderate. We demonstrate how the sensor infrastructure present in smart buildings can be leveraged to accurately predict future power consumptions. Our system is based on commercially available device-level measurement units that transmit consumption data to a central building server. The server extracts characteristic fingerprints from historical power consumption data and uses time series pattern matching in order to detect similarities. Our demo shows that long-term predictions of an appliance’s power consumption can be made, even when an appliance has been in operation for less than a minute.