Optimal privacy-preserving load scheduling in smart grid

With the wide deployment of smart meters in the power grid, it is becoming much easier to gather the detailed power consumption data of residential users, which enables the possibility of smarter and greener power grid. However, the fine-grained load profile of the individual user also introduces the severe concern of privacy leakage as the private information such as personal living habits may be inferred by the malicious third parties for unauthorized use and benefits. Different from most existing privacy-preserving energy management works which are solely based on the control of rechargeable batteries, we further introduce the proactive scheduling of widely used thermostatically controlled devices, including air conditioner, water heater, and laundry drier for effective load hiding. To minimize the weighed sum of financial cost, the deviation from the pre-defined load profile, and the user dissatisfaction, we formulate a novel load scheduling problem which is subject to both the device/battery physical dynamics and the practical user requirements. In order to solve the overall problem effectively under the uncertain price, we decompose the primal problem into a series of subproblems through dual composition, and design a stochastic gradient based two-level iterative distributed algorithm. Extensive simulations under various parameters are employed to demonstrate the effectiveness of our design.