A tight upper bound for quadratic knapsack problems in grid-based wind farm layout optimization. Issue 3 (4th March 2018)

• Record Type: Journal Article
• Authors: Quan, Ning; Kim, Harrison M.
• Is Part Of: Engineering optimization. Volume 50:Issue 3(2018)
• Page Start: 367
• Page End: 381
• Publication Date: 2018-03-04
• Subjects: Wind farm -- layout optimization -- mixed integer programming -- quadratic knapsack problem -- upper bound -- greedy algorithm; Engineering design -- Periodicals; Mathematical optimization -- Periodicals; 620.0042
• DOI: 10.1080/0305215X.2017.1316844
• Languages: English
• ISSNs: 0305-215X

Abstract

ABSTRACT: The 0-1 quadratic knapsack problem (QKP) in wind farm layout optimization models possible turbine locations as nodes, and power loss due to wake effects between pairs of turbines as edges in a complete graph. The goal is to select up to a certain number of turbine locations such that the sum of selected node and edge coefficients is maximized. Finding the optimal solution to the QKP is difficult in general, but it is possible to obtain a tight upper bound on the QKP's optimal value which facilitates the use of heuristics to solve QKPs by giving a good estimate of the optimality gap of any feasible solution. This article applies an upper bound method that is especially well-suited to QKPs in wind farm layout optimization due to certain features of the formulation that reduce the computational complexity of calculating the upper bound. The usefulness of the upper bound was demonstrated by assessing the performance of the greedy algorithm for solving QKPs in wind farm layout optimization. The results show that the greedy algorithm produces good solutions within 4% of the optimal value for small to medium sized problems considered in this article.