An enhanced MPC-based strategy for non-disruptive load shedding.

Model predictive control (MPC) methods have a well-earned reputation for providing on-line solutions to optimal feedback control problems, particularly for systems with control and parameter constraints. Previous work has shown the value of MPC in designing non-disruptive load-shedding strategies for power systems. The nonlinear, non-smooth dynamics of power systems make direct application of MPC difficult though. Therefore previous load-shedding applications of MPC have made use of an approximate discrete-time linear dynamic model that describes perturbations to the system's nominal behavior over a finite-time horizon. This approximate model is based on trajectory sensitivities. The article pursues several enhancements of such MPC-based load-shedding strategies. Specifically, at each MPC stage, we propose using a two-step optimization process to determine the optimal input sequence. This helps in combating the possibility of growing error in the discrete-time approximation if large input modifications are needed. We also consider the effects of varying voltage constraints over the MPC optimization horizon. The new two-step MPC strategies are used to design load-shedding controls that prevent voltage collapse in a ten-bus benchmark-system example.