9 Month Report: Using Hidden Markov Models for Non-intrusive Appliance Load Monitoring

Today I submitted my 9 month report:

Title: Using Hidden Markov Models for Non-intrusive Appliance Load Monitoring

Abstract: With the global problems of diminishing fossil fuel reserves and climate change comes the need to reduce energy consumption. One approach in the domestic sector is to improve consumer understanding of how much energy is consumed by each of their household appliances. This empowers consumers to make educated decisions on how they can change their lifestyle in an effort to reduce their energy consumption. This report investigates how total household energy consumption can be disaggregated into individual appliances using low granularity data collected from a single point of measurement, such as a smart meter. First, this work discusses how this problem can be solved using combinatorial optimisation. Second, a novel approach is introduced using hidden Markov models to model each appliance. We show through empirical evaluation that our implementation of the hidden Markov model approach outperforms the combinatorial optimisation method. Our future work is then outlined to show how we plan to better model domestic appliances and therefore improve the accuracy of our approach.