Michael McGuigan

Acting Director, Brookhaven National Laboratory

Smart Grid Analysis using High Performance Architectures and Algorithms

This study applies supercomputing resources to the optimization of "smart grid" technology that will suggest upgrades and expansions of the current United States electric grid to support the rising demand for electricity and enable the integration of renewable energy sources. Using IBM's Blue Gene/L supercomputer available at Brookhaven National Laboratory, this study evaluated the minimization of a cost function by Lagrangian Relaxation to model future energy scenarios (Google Clean Energy 2030, Department of Energy goals, and the Solar Grand Plan), and the scalability of the LSQR and LSMR algorithms for solving large over-determined systems of power flow equations on the millisecond timescale. Input data on energy supply and demand was filtered and compiled from several databases including the Energy Information Administration's Annual Energy Outlook and the Environmental Protection Agency's Emissions and Generation Resource Integrated Database. Running the energy scenario models determined that the Solar Grand Plan has a higher cost than both Google Clean Energy 2030 and Department of Energy goals through 2030, where the cost is highly dependent on predicted demands for each scenario. The parallel implementation of the cost function minimization routine had good scalability on the Blue Gene/L supercomputer up to 512 processors and beyond that communication overhead between processors caused performance to decrease. It was also determined that the LSMR algorithm is more efficient than the LSQR algorithm for solving large ill-conditioned systems of equations on parallel architectures. Uncertainty in future demand and intermittency of renewable power sources necessitates real time solutions to large-scale computational problems based on a complex transmission network. This study supports the application of supercomputing resources to the planning and operation of the smart grid.


Michael McGuigan is a Computational Scientist and Acting Director of the Computational Science Center of Brookhaven National Laboratory. He specializes in the field of scientific visualization, parallel computing and data analysis. His computational research interests include parallel visualization and parallel simulation. He works in the field of nanoscience on the implementation of parallel Monte Carlo Heisenberg models on supercomputers such as BlueGene to study magnetic materials. He is also developing exascale density functional codes that can be applied to nanoscience in particular nanorings, nanotubes and nanoparticles of gold and carbon. Before coming to Brookhaven he was an editor for the Physical Review and postdoctoral fellow at the University of Florida and the Institute for Advanced Study at Princeton.

Workshop Program
updated: 2011-10-19