Resilient Plant Monitoring Systems: Techniques, Analysis, Design, and Performance Evaluation

Resilient plant monitoring systems are sensor networks that degrade gracefully under malicious attacks on their sensors, causing them to project misleading information. This paper develops techniques to ensure resiliency and illustrates their application using a power plant. Specific techniques developed are: active data quality acquisition, process variable and plant condition assessments, sensor network adaptation, and plant decomposition with knowledge fusion. Based on these techniques, a five-layer resilient monitoring architecture is proposed and analyzed under various cyber-physical attack scenarios. As quantified by Kullback-Leibler divergence, in all scenarios considered, the system offers effective protection against misleading information and identifies the plant conditions normal or anomalous in a reliable manner.