Distributed Predictive and Real-Time Energy Management in Future Electric Power Systems

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

Control of large power grids today is largely performed by centralized computer programs run at centers like the Independent System Operators (ISO). In order to cope with the growing stress and difficulty of these computations, as more renewable energy and more complicated users of power come on line, this team plans to develop new systems for distributed computation, both for scheduling of power supply and demand, and for balancing decisions which maintain stability of the grid in real time. The research feeds into an active power education effort, including special efforts by the PI in Women in Engineering and in summer engineering programs for girls. This work fits within the growing new effort to take advantage of distributed or parallel computing, after the speed of classical sequential computers has hardly improved at all for many years. The distributed scheduling here will be based on multi-stage DC optimal power flow, though the team plans to move ahead to the nonlinear case of realistic AC flows as soon as they are able. The real-time balancing will try to balance dispatchable and nondispatchable generation and loads based on a consensus-plus-innovation approach