Cost minimization of large-scale infrastructure for electricity generation and transmission. (October 2020)

• Record Type: Journal Article
• Authors: Märkle-Huß, Joscha; Feuerriegel, Stefan; Neumann, Dirk
• Is Part Of: Omega. Volume 96(2020)
• Publication Date: 2020-10
• Subjects: Energy -- Electricity generation and transmission planning -- Neighborhood heuristics -- Metaheuristics -- Variable neighborhood search -- Real-world application; Management -- Periodicals; 658.4005
• DOI: 10.1016/j.omega.2019.05.007
• Languages: English
• ISSNs: 0305-0483

Abstract

Highlights: We propose a novel model for optimizing electricity generation and transmission. It determines capacity, type & location of power plants and transmission lines. Exact solvers based on branch-and-reduce fail to solve even small instances. We develop different heuristics based on variable neighborhood search and GRASP. These also work for large-scale problems such as the full German infrastructure. Abstract: Electricity infrastructure confronts societies with immense costs as it must ensure the generation of power and its transmission to locations with consumption requirements. We minimize these costs by formulating an electricity generation and transmission problem that facilitates the design of electricity infrastructure on a macro level. Our problem specifies the capacity, type, and location of power plants and, at the same time, determines the appropriate arrangement of high-voltage transmission lines in order to fulfill the demand of individual cities. We specifically incorporate the non-linear nature of cost functions for power generation that are common in practice. This results in a mixed integer non-linear problem, for which the branch-and-reduce solver from GAMS exceeds runtime constraints, even for small instances with 25 locations. As a remedy, we develop heuristics based on the reduced variable neighborhood search and the greedy randomized adaptive search procedure (GRASP). Their performance enables us to address large-scale problems that arise in real-world applications. We demonstrate this with an actual, nationwide example that spans all 4537 municipalities in Germany.