| Title | Data-driven Physics-based Solution for False Data Injection Diagnosis in Smart Grids |
| Publication Type | Conference Paper |
| Year of Publication | 2019 |
| Authors | Trevizan, Rodrigo D., Ruben, Cody, Nagaraj, Keerthiraj, Ibukun, Layiwola L., Starke, Allen C., Bretas, Arturo S., McNair, Janise, Zare, Alina |
| Conference Name | 2019 IEEE Power Energy Society General Meeting (PESGM) |
| Keywords | anomaly detection, artificial intelligence, bad data detection strategy, chi-squared test, composability, cyber physical systems, cyber-attacks, data-driven machine intelligence, data-driven physics-based solution, False Data Detection, False Data Injection, false data injection diagnosis, gross error analysis, Human Behavior, intelligent FDI attack, machine intelligence, physics based state estimation process, physics-based method, power engineering computing, power grid, power system faults, power system security, power system state estimation, pubcrawl, Reed-Xaoli, resilience, Resiliency, security of data, SG technology, Smart grids, smart power grids, statistical testing, statistical tests, undetected bad data |
| Abstract | This paper presents a data-driven and physics-based method for detection of false data injection (FDI) in Smart Grids (SG). As the power grid transitions to the use of SG technology, it becomes more vulnerable to cyber-attacks like FDI. Current strategies for the detection of bad data in the grid rely on the physics based State Estimation (SE) process and statistical tests. This strategy is naturally vulnerable to undetected bad data as well as false positive scenarios, which means it can be exploited by an intelligent FDI attack. In order to enhance the robustness of bad data detection, the paper proposes the use of data-driven Machine Intelligence (MI) working together with current bad data detection via a combined Chi-squared test. Since MI learns over time and uses past data, it provides a different perspective on the data than the SE, which analyzes only the current data and relies on the physics based model of the system. This combined bad data detection strategy is tested on the IEEE 118 bus system. |
| DOI | 10.1109/PESGM40551.2019.8974027 |
| Citation Key | trevizan_data-driven_2019 |
Data-driven Physics-based Solution for False Data Injection Diagnosis in Smart Grids