A machine learning-based two-stage approach for the location of undesirable facilities in the biomass-to-bioenergy supply chain


Yunusoğlu P., Özsoydan F. B., Bilgen B.

APPLIED ENERGY, vol.362, no.122961, pp.1-30, 2024 (SCI-Expanded)

  • Publication Type: Article / Article
  • Volume: 362 Issue: 122961
  • Publication Date: 2024
  • Doi Number: 10.1016/j.apenergy.2024.122961
  • Journal Name: APPLIED ENERGY
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Biotechnology Research Abstracts, CAB Abstracts, Communication Abstracts, Compendex, Environment Index, INSPEC, Pollution Abstracts, Public Affairs Index, Veterinary Science Database, Civil Engineering Abstracts
  • Page Numbers: pp.1-30
  • Dokuz Eylül University Affiliated: Yes

Abstract

Biomass-to-bioenergy supply chain management is an integral part of the sustainable industrialization of energy conversion through biomass to bioenergy by managing economic, environmental, and social challenges encountered in each supply chain stage. Motivated by a real-world biomass-to-bioenergy supply chain network design (BSCND) problem, this study addresses the location of undesirable facilities for the first time in the BSCND literature. The motivation of this study is to develop a machine learning-based two-stage approach for solving the BSCND problem with undesirable facilities that have a negative impact on surrounding communities. The first stage employs the k-means clustering algorithm to alleviate the complexity of the problem, and the second stage utilizes a novel pre-emptive goal programming (PGP) approach to optimize two distinct objectives hierarchically. The first objective maximizes the sum of the distances between all clients and the open facilities, which is the well-known objective of the obnoxious p-median (OpM) problem. The second objective maximizes the total profit of the entire supply chain. The applicability of the proposed solution approach is shown through the case problem, and performance of the two-stage approach is validated using randomly generated test problems. The computational results indicate the effectiveness of the clustering methodology in reducing the complexity of the problem while the PGP achieves the optimal configuration of the biomass-to-bioenergy supply chain handling the hierarchical objectives. The optimal solution of the case problem was achieved within 25,239.36 s execution time, and the total profit of the supply chain is $6,776,870.22 with 735 km total distance to clients. The average optimality gap for the first phase of the PGP is 4.97%, and the average optimality gap for the second phase of the PGP is 0.01% for the generated test problems.