Biblio

Smart meter devices enable a better understanding of the demand at the potential risk of private information leakage. One promising solution to mitigating such risk is to inject noises into the meter data to achieve a certain level of differential privacy. In this paper, we cast one-shot non-intrusive load monitoring (NILM) in the compressive sensing framework, and bridge the gap between theoretical accuracy of NILM inference and differential privacy's parameters. We then derive the valid theoretical bounds to offer insights on how the differential privacy parameters affect the NILM performance. Moreover, we generalize our conclusions by proposing the hierarchical framework to solve the multishot NILM problem. Numerical experiments verify our analytical results and offer better physical insights of differential privacy in various practical scenarios. This also demonstrates the significance of our work for the general privacy preserving mechanism design.

With Electricity as a fundamental part of our life, its production has still large, negative environmental impact. Therefore, one strain of research is to optimize electricity usage by avoiding its unnecessary consumption or time its consumption when green energy is available. The shift towards an Advanced Metering Infrastructure (AMI) allows to optimize energy distribution based on the current load at residence level. However, applications such as Demand Management and Advanced Load Forecasting require information further down at device level, which cannot be provided by standard electricity meters nor existing AMIs. Hence, different approaches for appliance monitoring emerged over the past 30 years which are categorized into Intrusive systems requiring multiple distributed sensors and Non-Intrusive systems requiring a single unobtrusive sensor. Although each category has been individually explored, hybrid approaches have received little attention. Our experiments highlight that variable consumer devices (e.g. PCs) are detrimental to the detection performance of non-intrusive systems. We further show that their influence can be inhibited by using sensor data from additional intrusive sensors. Even fairly straightforward sensor fusion techniques lead to a classification performance (F1) gain from 84.88 % to 93.41 % in our test setup. As this highlights the potential to contribute to the global goal of saving energy, we define further research directions for hybrid load monitoring systems.