Privacy Preserving Bumpless Transfer MPC of Markov Switching Power Systems

Abstract

Abstract

ABSTRACT This paper studies privacy‐preserving bumpless‐transfer control for Markov switching power systems under hard state and input constraints. A single‐machine infinite‐bus system subject to random operating‐condition changes is modeled as a constrained Markov switching system, and its state information is transmitted through a masking‐noise‐based encryption–decryption interface. To handle both stochastic mode transitions and privacy‐induced perturbations, a privacy‐preserving bumpless‐transfer model predictive control (PPBT‐MPC) scheme is developed. The controller employs a mode‐time dependent gain schedule to distribute gain adjustment over a finite transition window, so that abrupt control changes at switching instants can be suppressed even when deterministic dwell times are unavailable. The controller synthesis is cast as a convex semidefinite program, in which the disturbance bound induced by the masking mechanism is incorporated into the Lyapunov conditions, tightened state/input constraints, and bump‐related inequalities. It is shown that the resulting closed‐loop system is recursively feasible, that the decrypted‐state dynamics are mean‐square practically stable, and that the true physical state admits an explicit ultimate bound determined by the privacy‐noise level. Simulations on a Markov switching single‐machine infinite‐bus system show that the proposed method yields smoother control transitions and stronger resistance to state reconstruction by an unauthorized observer than representative baseline controllers.