A comparison of evolutionary algorithms on a Large Scale Many-Objective Problem in Food–Energy–Water Nexus

Abstract

Food, energy, and water resources form a complex system called Food–Energy–Water Nexus (FEWN) that is crucial for the survival of human beings. These resources interact with each other in a way that achieving the objectives of one can conflict with the objectives of the others. Therefore, there is a need to have approaches such as Multi-Objective Optimization Evolutionary Algorithms (MOEAs) to address the conflicting objectives in FEWN. However, MOEAs are inefficient in solving problems with more than 3 objectives and with at least 100 decision variables. In recent years Large-Scale Many-Objectives Optimization Evolutionary Algorithms (LSMaOEAs) have been developed to solve such problems. The performance of these algorithms has been tested on some real-world applications and benchmark problems. However, evaluation of their performance on the Food–Energy–Water Nexus (FEWN) rarely appears in existing literature. In this paper we propose a Large-Scale Many-Objectives Problem in Food–Energy–Water Nexus (LSMaOPFEWN) that consists of 5 objectives and 315 decision variables. This problem is formulated as a Leontief Input–Output model that promotes sustainable consumption and production of resources in FEWN. Thereafter, we report experiments that evaluated the performance of selected LSMaOEAs on LSMaOPFEWN. Experimental results demonstrate that Linear Combination-based Search Algorithm (LCSA) performs better than 4 other selected algorithms due to its enhanced exploration and exploitation capability. This paper demonstrates that an algorithm utilizing dimensionality reduction techniques can be effective in solving a real world LMaOPFEWN.

Keywords

• Multi-Objective Optimization Evolutionary Algorithms
• Large Scale Many-Objectives Optimization Evolutionary Algorithms
• Food–Energy–Water Nexus
• Leontief Input–Output model
• Sustainable consumption and production
• Dimensionality reduction