Problems of large-scale nonlinear optimization are central to the solution of difficult and novel problems that arise from the need to distribute electric power in the most efficient and reliable way. In recognition of the importance of optimization in this area, DOE & apos;s Office of Advanced Scientific Computing (ASCR) has made substantial investments in research that have led to powerful, robust software for very general classes of nonlinear optimization problems. There remains however a substantial need to extend and adapt these efforts to better take advantage of high-performance computing (HPC) and to unlock their value for new users. Power industry optimization is a particularly promising area for such an undertaking, in light of the complex decisions involved and the challenges of changing costs and technologies. We propose to focus on three complementary areas of investigation to meet this need. First, we will re-engineer nonlinear optimization software, previously developed with ASCR support, so as to take advantage of high-performance computing (HPC) concepts that address energy efficiency problems too large for current codes. Second, we will adapt and focus software technology from previous work for ASCR to address the specific needs of power grid applications, in which HPC techniques are needed in order to re-optimize very quickly a large number of problems of similar structure. Lastly we will develop new software techniques for tackling complex, large-scale power applications that model disjunctive conditions. All of these initiatives will strengthen the internals of algorithms to give the software wider applicability in HPC settings and to better deal with greatly increased problem size and complexity due to uncertainty. Commercial Applications and Other
Benefits: This work will be of direct public benefit by enabling scarce and costly resources to be used much more effectively for purposes of power distribution. The first area of investigation will be applied in particular to optimal power flow with contingencies and to short-term stochastic dispatch for systems with a high level of renewable penetration. The second area will be applied in improving grid short-term management, from both technical and economic stand- points, particularly in taking full advantage of new demand side management and generation technologies to reduce costs and environmental impact subject to security constraints. The third area will allow for the solution of more complex power flow optimization models, including for example the ability to switch on/off various power production constraints.
