Abstract

The processing and communication capabilities of the smart grid provide a solid foundation for enhancing its efficiency and reliability. These capabilities allow utility companies to adjust their offerings in a way that encourages consumers to reduce their peak hour consumption, resulting in a more efficient electrical grid. The smart grid achieves this through the introduction of smart meters; which collect and transmit consumers’ detailed power consumption information in an automated way. Despite their benefits, these readings introduce a major privacy threat to residential consumers as they reveal details that could be used to infer information about the activities of the occupants of a home.

In this thesis, we first propose a method for scheduling a community’s power consumption such that it becomes almost flat. Our methodology utilizes distributed schedulers that allocate time slots to soft loads probabilistically based on pre-calculated and pre-distributed demand forecast information. This approach requires no communication or coordination between scheduling nodes and the computation performed at each scheduling node is minimal. Obtaining a relatively constant consumption makes it possible to have a relatively constant billing rate and eliminates operational inefficiencies. We also analyze the fairness of our proposed approach, the effect of the possible errors in the demand forecast, and the participation incentives for consumers.

In the second part of the thesis, we question the need to disclose high frequency readings produced at the home’s level. Instead, we propose equipping smart meters with sufficient processing power enabling them to provide their corresponding utility company with a set of well-defined services based on these readings. For demand side management, we propose the collection of high frequency readings at a higher level in the distribution network, such as at local step-down transformers, as this readily provides the accumulated demand of all homes within a branch. In addition, we study the effect of the proposed approach on consumers’ privacy, using correlation and relative entropy as measures. We also study the effect of load balancing on consumers’ privacy when using the proposed approach. Finally, we assess the detection of appliances using high frequency readings collected for demand side management purposes.