Exploiting Temporal Difference for Energy Disaggregation via Discriminative Sparse Coding

Abstract: This thesis analyzes one hour based energy disaggregation using Sparse Coding by exploiting temporal differences. Energy disaggregation is the task of taking a whole-home energy signal and separating it into its component appliances. Studies have shown that having device-level energy information can cause users to conserve significant amounts of energy, but current electricity meters only report whole-home data. Thus, developing algorithmic methods for disaggregation presents a key technical challenge in the effort to maximize energy conservation. In Energy Disaggregation or sometimes called Non- Intrusive Load Monitoring (NILM) most approaches are based on high frequent monitored appliances, while households only measure their consumption via smart-meters, which only account for one-hour measurements. This thesis aims at implementing key algorithms from J. Zico Kotler, Siddarth Batra and Andrew Ng paper "Energy Disaggregation via Discriminative Sparse Coding" and try to replicate the results by exploiting temporal differences that occur when dealing with time series data. The implementation was successful, but the results were inconclusive when dealing with large datasets, as the algorithm was too computationally heavy for the resources available. The work was performed at the Swedish company Greenely, who develops visualizations based on gamification for energy bills via a mobile application.