Comparative analysis of load shaping based privacy protection strategies in smart grids

The increase in the amount and the variety of electricity production and consumption globally during recent years necessitated a more modern electricity grid. The concept of smart grid aims to fulfill this necessity by combining the traditional grid with a communications network. Hence, the grid can be monitored in real-time, outages and failures can be responded more quickly, energy theft can be detected, the grid can be managed in a distributed way and new services such as demand response can be available. Advanced Metering Infrastructure (AMI) is one of the core elements of the smart grid architecture. AMI measures, collects and analyzes electricity usage either on-demand or in a periodic basis. Therefore, AMI enables the grid to be more efficient and reliable. Smart meters which are in the center of the AMI systems can measure electricity usage in real-time and can send frequent and accurate readings to the Utility Provider (UP) or to the customer. However, the fine-grained metering data can also be used for other purposes due to the advances in data science. Especially, Nonintrusive Appliance Load Monitoring (NIALM) techniques can be used to disaggregate the metering data into individual appliance usages or end-uses. Therefore, private information (e.g., vacancy of the house, demographics of the residents, daily habits) can be inferred from the smart meter data. Thus, it is of utmost importance to protect SM data from adversaries. Different approaches to ensure consumer privacy have been proposed in the literature. Some of these use cryptographic techniques to hide the data, while others perturb it by adding random noise. Using pseudonyms instead of consumer ID's (anonymization) and decreasing the measurement frequency (downsampling) are also proposed to decrease the privacy threat. Among such methods, load shaping (LS) methods, where the consumer load is scheduled to release as little information as possible, shine out in terms of feasibility and effectiveness. In this thesis, a series of comparative analyses of load shaping-based privacy protection strategies were presented. In the introductory section, the concepts of smart grid, AMI, smart metering, demand response and NIALM were explained. Various privacy protection approaches were reviewed. After the introductory section, there are four main sections. Each of these sections narrates an independent study about load shaping based privacy protection. The first study analyzes and compares 8 different LS-based privacy protection strategies from the literature under a unified stochastic optimization framework in a demand response scheme. The second study extends the first study by considering the UP and its goals, and analyzes how emphasizing the consumer's objectives (including privacy) affects the UP's objectives. The third study focuses on reactive power metering and it shows that in order to ensure privacy, real and reactive power must be shaped simultaneously. The fourth study considers the advancing smart water and natural gas metering in addition to smart electricity metering, and tackles the privacy problem in a holistic sense. Concluding remarks were presented in the concluding chapter. The series of comparative analyses presented in this thesis enables a fair comparison of different LS-based privacy protection strategies as well as their effects on UP's objectives. Furthermore, it sheds light on the concept of holistic privacy as well as real and reactive power based privacy protection, which are virgin lands on the SG literature.