Operating a power grid means solving hard computational problems on a short clock, and the standard way to make that tractable is to replace the AC power flow equations with a linear approximation. Those approximations are almost always designed and judged by how closely they reproduce the power flow equations themselves, with no reference to what the model will actually be used for. This project argues that the accuracy that matters is the accuracy of the application output, and it closes the design loop around the end use.

We develop certifiably optimal power flow linearizations tailored to three settings, deterministic optimization, stochastic optimization, and dynamic modeling, covering canonical problems such as bulk system dispatch, distributed energy resource management, voltage control, uncertainty management, planning, and time-domain simulation. The technical approach draws on bilevel and stochastic optimization, systems theory, sensitivity analysis, automatic differentiation, system identification, and neural networks, so that input data distributions and application structure enter the linearization process explicitly rather than by convention.

This project improves the strength and flexibility of the electric grid by making better use of grid-enhancing technologies, or GETs: advanced conductors, power flow controllers, phase-shifting transformers, and dynamic line rating. GETs are known for raising transmission capacity, but their broader system-level effects are far less understood, particularly their contribution to grid strength in systems rich in inverter-based resources.

We are developing new metrics, scalable algorithms, and integration strategies that characterize and optimize GETs for modern IBR-rich grids. The goal is to widen the understanding of what GETs can do at the system level, and to make them usable inside the planning and operation tools that utilities already run, including energy management systems, so the grid becomes more resilient and more adaptable.