Encrypted distributed model predictive control of nonlinear processes

In this research, we present an encrypted iterative distributed model predictive controller (DMPC) to enhance the computational efficiency and cybersecurity of large-scale nonlinear processes. In this configuration, a single large process is divided into numerous smaller subsystems, each regulated by a unique Lyapunov-based MPC (LMPC) that utilizes the complete process model and exchanges control inputs with other LMPCs to address the interactions between subsystems. Further, to enhance cybersecurity, all communication links between sensors, actuators, and control input computing units are encrypted. Through a comprehensive stability analysis of the encrypted iterative DMPC, bounds are established on errors arising from encrypted communication links, disturbances, and the sample-and-hold implementation of controllers. Practical aspects such as reducing data encryption time by appropriate key length choices, sampling interval criterion, and quantization parameter selection are discussed. Simulation results of the proposed control scheme, applied to a nonlinear chemical process, showcase its effective closed-loop performance in the presence of sensor noise and process disturbances. Specifically, a non-Gaussian noise distribution is obtained from an industrial data set and added to the state measurements to justify the practical effectiveness of the proposed approach.