Real-Time Embedded Low-Frequency Load Disaggregation

Understanding how the electrical appliances and devices in a house consume power is an important factor that can allow occupants to make intelligent and informed decisions about conserving energy. This is often not an easy task. These electrical loads can turn ON and OFF either by the actions of occupants, or by automatic sensing and actuation (e.g. thermostat). Even if we could keep track of when loads turned ON and OFF, it is difficult to understand how much a load consumes at any given operational state because of the lack of proper measurement reporting within equipment manuals. Occupants could buy sensors that would help, but this comes at a high financial cost. Power utility companies around the world are now replacing old electro-mechanical meters with digital meters (smart meters) that have enhanced communication capabilities. These smart meters are essentially a free sensor that offer an opportunity to use computation to infer what loads are currently running in a house and to estimate how much each load is consuming, a process often referred to as load disaggregation. This thesis presents a new load disaggregation algorithm (i.e. a disaggregator). This new disaggregator uses a super-state hidden Markov model and a new Viterbi algorithm variant which preserves dependencies between loads and can disaggregate multi-state loads, all while performing computationally efficient exact inference. Our sparse Viterbi algorithm can efficiently compute sparse matrices with a large number of super-states. Our disaggregator is the first of its kind to run in real-time on an inexpensive embedded processor using low sampling rates (e.g. per minute). Other contributions of the research include an analysis of electrical measurements, the release of a publicly available dataset, and a method for comprehensive accuracy testing and reporting.