Investigating the run-time searching behavior of the differential evolution algorithm applied to water distribution system optimization

In recent years, the differential evolution algorithm (DEA) has frequently been used to tackle various water resource problems due to its powerful search ability. However, one challenge of using the DEA is the tedious effort required to fine-tune parameter values due to a lack of theoretical understanding of what governs its searching behavior. This study investigates DEA's search behavior as a function of its parameter values. A range of behavioral metrics are developed to measure run-time statistics about DEA's performance, with primary focus on the search quality, convergence properties and solution generation statistics. Water distribution system design problems are utilized to enable investigation of the behavioral analysis using the developed metrics. Results obtained offer an improved knowledge on how the control parameter values affect DEA's search behavior, thereby providing guidance for parameter-tuning and hence hopefully increasing appropriate take-up of the DEA within the industry in tackling water resource optimization problems.