Responding to climate change requires efforts from utility providers for the production of and engagement with more advanced integrated electrical distribution grids. For example, much of the smart grid effort focuses on bridging together renewable energy sources (as well as an increased level of smart monitoring and automation of electrical transmission). Within the smart grid, the smart meter also has the potential to play a role in reducing the level of carbon emissions for all residential customers, as smart meters provide the means for customers to manage and reduce their electricity usage. For example, detailed energy profiles of energy usage patterns can be constructed and reported back to the end-user. As such, this research investigates residential home environments and how the data produced by smart meters can be used to profile energy usage in homes. In particular, this project concerns the design of a system and algorithm to model and predict household behavior patterns from smart meter readings. The aim is to model the behavioral trends in homes to develop an autonomous system that can advise home users on changes that can be adopted to reduce their carbon emissions. The data used for the research were constructed from a digital simulation model of multiple smart home environments. Using a two-class boosted decision tree, the system is able to detect anomalous users with 71.7% AUC.
Read full abstract