New control framework cuts power grid costs by 30% in extreme weather

This paper by Furkan Sezer tackles a core problem in multi-area power system coordination: how to align the incentives of independent regional operators during extreme weather events. The author models the problem as a continuous-time stochastic Stackelberg game where a leader (e.g., a central coordinator) cannot directly control followers but can influence them through two channels: a public information structure (Gaussian signals with finite-dimensional belief tracking) and a transfer mechanism (a Groves scheme). Under truthful disclosure, efficient behavior becomes a dominant-strategy best response. The distributionally robust version, which accounts for ambiguity in the jump-diffusion environment via relative entropy, collapses the bilevel problem into a single robust control problem with an exact first-order condition. The value function is characterized as a viscosity solution, and the switching set is shown to be Lebesgue-null, enabling practical numerical solutions.

Calibrated to ERCOT's near-islanded interconnection during the 2021 Winter Storm Uri (a 0.82 GW tie, under 2% of peak), the author solves an Isaacs equation and finds that mutual aid removes about 8% of social cost under existing infrastructure. When interregional transfer capability is raised to FERC/DOE-recommended levels, that savings jumps to roughly 30%. A reserve-scheduling experiment further shows that public disclosure lowers welfare cost by 37% under autarky and 48% under market coupling. Importantly, information design and market coupling are complements under systemic (common) risk, not substitutes. The framework provides a rigorous, scalable method for designing coordination policies that improve grid resilience and reduce economic losses during extreme events.