Premium
Parallel matheuristics for the discrete unit commitment problem with min‐stop ramping constraints
Author(s) -
Dupin Nicolas,
Talbi Elghazali
Publication year - 2020
Publication title -
international transactions in operational research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.032
H-Index - 52
eISSN - 1475-3995
pISSN - 0969-6016
DOI - 10.1111/itor.12557
Subject(s) - mathematical optimization , computer science , power system simulation , heuristic , limit (mathematics) , integer programming , scheme (mathematics) , variable (mathematics) , consistency (knowledge bases) , power (physics) , optimization problem , quality (philosophy) , mathematics , electric power system , mathematical analysis , philosophy , physics , epistemology , quantum mechanics , artificial intelligence
Abstract The discrete unit commitment problem with min‐stop ramping constraints optimizes the daily production of thermal power plants, subject to an operational reactivity of thermal units in a 30‐minute delay. Previously, mixed integer programming (MIP) formulations aimed at an exact optimization approach. This paper derives matheuristics to face the short time limit imposed by the operational constraints. Continuous relaxations guide the search for feasible solutions exploiting tailored variable fixing strategies. Parallel matheuristics are derived considering complementary strategies in parallel. Tests were performed on more than 600 real‐life instances. Our parallel matheuristic provides high‐quality solutions and outperforms the MIP approach in the time limits imposed by the industrial application. This paper illustrates a special interest for matheuristics in industrial highly constrained problems: many tailored neighborhood searches can be derived from an MIP formulation, and their combination in a parallel scheme improves the solution quality as well as the consistency of the heuristic.