Comparing Performance of Evolutionary Algorithms for Optimizing Synthetic and Real-world Case Studies
Volume Title: 1
1Chemical Engineering Department, Faculty of Engineering, Ferdowsi University of Mashhad
2Institute for Intelligent Systems Research and Innovation (IISRI), Deakin University, Australia
The application of meta-heuristic algorithms (MAs), for optimizing processes in chemical industries has rapidly grown in recent years. This is mainly driven by the increased cost of raw materials and the economic pressure for reducing the operational costs. It is already known that there is no best method. The performance of MAs is case-specific and may change. Accordingly, selecting a proper MA for a desired optimization has a special importance. For this reason, a comparison must be made to assess the capabilities of different algorithms. In this study, a fair comparison was made on some benchmark functions and a real-world case study (neural network model of a natural gas to liquids (GTL) process) between three MAs including of genetic algorithm (GA), cuckoo search algorithm (CSA) and particle swarm optimization (PSO). In doing so, functions were selected with various features and the algorithms were tuned using parameter meta optimization (PMO) method. Four performance indicators including of solution quality, accuracy, effort and success were determined for assessment. The results showed that CSA has better performance than others in most cases. Lower mean solution quality, standard deviation, required time (or iteration) and higher success parameter in most cases indicated the superiority of CSA.