New Algorithm Optimizes Microgrid Power Flow with Generator Constraints

A team led by Jürgen Gutekunst and colleagues has advanced economic nonlinear model predictive control (NMPC) for microgrids by addressing the challenge of including discrete decision variables for generator runtimes and startup costs. Traditional methods lead to large-scale mixed-integer nonlinear programs (MINLPs) that are computationally prohibitive for real-time control. The authors propose replacing nonlinear AC power flow equations with convex quadratic approximations, transforming the problem into a mixed-integer quadratically constrained program (MIQCP). This reduces complexity significantly, enabling off-the-shelf solvers like CPLEX to handle it in reasonable time.

The paper also tackles recursive feasibility—a critical property for MPC stability—by introducing time-coupled constraints instead of simple terminal conditions. They prove that with these constraints, the controller remains feasible over time, and under periodic dissipativity, stability follows. Simulations on a realistic 6-bus microgrid under varying demand scenarios demonstrate the method's effectiveness. This work bridges the gap between theoretical optimal power flow and practical implementation for microgrids with discrete generator constraints.