École Nationale Supérieure de
Techniques Avancées

Abstract

This research projects aims to improve the optimization techniques for large-scale industrial energy planning linear programming (LP) models, using Interior PointMethod (IPM)s. The Hydro Unit Commitment (HUC) is employed as a case study, beginning with problem modeling. While LPs include energy-related problems, the HiGHS IPM solver experiences inefficiencies, particularly with large-scalemodels that contain one or more dense columns in the constraint matrix. This complexity not only heightens computational demands of direct solution techniques but also increases costs in iterative processes. As a solution, Sherman-Morrison-Woodbury (SMW) formula was explored for its potential to reduce computational costs. Alternatively, an indefinite solver that leverages system sparsity might be more cost-effective. By employing metrics like the fill-in factor and factorization time, this study assesses the performance and accuracy of decomposers. The insights gathered are invaluable for the further enhancement and development of a HiGHS direct IPM solver tailored for intricate LP scenarios.