| Thanks to the bi-directional communication capabilities, smart meters are widely deployed in the modern smart grid infrastructure in replacement of the traditional analog electricity meters. Letting alone the convenience and efficiency smart meters bring, they are vulnerable to cyberattacks and energy thefts, which pose serious threats to the integrity of the smart home billing infrastructure and the smart energy Cyber-Physical System (CPS). In this dissertation, three energy theft/cyberattack strategies are analyzed and the corresponding detection mechanisms are proposed. For detection of energy theft in smart homes by narrowing the search zone, a dynamic programming algorithm for inserting the minimum number of Feeder Remote Terminal Units (FRTU) into the distribution network is proposed. For energy theft detection in data centers, the dynamic programming algorithm is extended with an anomaly rate estimation algorithm based on Minimum Covariance Determinant (MCD) algorithm. For combating energy theft and cyberattacks targeting the sponsor incentive pricing scheme, a detection algorithm featuring an integrated short-term and long-term detection framework is proposed. The short-term detector identifies potential energy theft/cyberattack behaviors by analyzing the energy usage patterns of each customer using Binary Logistic Regression (BLR). The long-term detector maintains the probabilistic view of the monitoring history and optimize the timing for on-site inspection.;The experimental results demonstrate that, compared with the previous work, the proposed dynamic programming algorithm inserts less FRTUs without sacrificing the solution quality. More importantly, it is more scalable to handle large test cases. The integrated detection framework can effectively mitigate the impacts of both energy thefts and cyberattacks targeting the sponsor incentive pricing scheme. |
