A Non-Intrusive Load Monitoring Framework for Robust Real-Time Disaggregation of Smart Meter Data

Reducing household electricity cost is traditionally a difficult task, as users have to use educated guesses to identify energy wastage caused by the inefficient use of appliances. This difficulty is largely a result of the poor feedback offered by traditional electricity bills, given that only the total energy consumption for a given billing cycle is shown, leaving users clueless about the actual cause of high energy consumption. While dedicated sensors can be installed to measure the energy consumption of each appliance in a house, the installation process is typically costly and intrusive. To that end, this thesis investigates and proposes mathematical techniques for Non-intrusive Load Monitoring, whereby measurements of total power consumption (from only a single sensor) are decomposed into their appliance contributions, enabling the relative proportion of energy of each appliance in the house to be obtained inexpensively.