Digital Twin-Based Resilient Model Predictive Control for Industrial Cyber-Physical Systems

Resilience is an essential property for resisting cyber-attacks and ensuring the stability of industrial cyber-physical systems (ICPSs). This article proposes a novel digital twin-based resilient model predictive control (MPC) method for ICPSs with intelligent false data injection (FDI) attacks. First, a general digital twin is developed based on the long short-term memory model to predict the next system state. Specifically, the sensor measurements are utilized to predict the actual system state when the system is running normally; when the system is under attacks, the digital twin turns to a self-iteration mode and starts a roll-back mechanism to predict the system state by using the historical datasets. Then, a resilient Lyapunov-based MPC method with the predictive model established by the digital twin is designed to resist intelligent FDI attacks and the stability of the closed-loop system is ensured. Finally, the experiments with three types of intelligent FDI attacks are conducted on a self-build hardware-in-the-loop platform to test the performance of the proposed control framework for the industrial two-continuous stirred tank reactor-in-series process. The results show that the digital twin can accurately replicate the system behaviors and the proposed control scheme is efficient to resist cyber-attacks and guarantee the system stability.