Dynamic Load-Altering Cyberattacks against Direct Load Control of Residential Electric Water Heaters

Digital technologies are necessary to support smart grid’s load-management capabilities, such as direct load control (DLC), and the emergence of new controllable high-wattage loads, such as electric vehicles. Increased penetration of these technologies expand the grid’s exposure to cyber threats including load-altering attacks (LAAs). These attacks target disrupting power grid services by controlling and manipulating customers’ WiFi-connected, high-wattage appliances. In this paper, we consider a dynamic LAA model against peak shaving using DLC of residential, electrical water heater (EWH) load. We assess its potential impact on the grid in the presence of two machine-learning-based detection methods and a cybersecurity-by-design mitigation solution. Experiments show that the considered dynamic LAA is capable of evading detection in several cases leading to significant load surges; however, experiments also show that the risk of this LAA can be significantly attenuated using a combination of a Time-Delay Neural Network (TDNN) detector and hardwired mitigation.