PGO: A parallel computing platform for global optimization based on genetic algorithm

This paper presents the design, architecture and implementation of a general parallel computing platform, termed PGO, based on the Genetic Algorithm (GA) for global optimization. PGO provides an efficient and easy-to-use framework for parallelizing the global optimization procedure for general scientific modeling and simulation processes. Along with a core optimization kernel built on a GA, PGO also includes a general input generator and an output extractor that can facilitate its easy integration with various scientific computing tasks. In this paper, we demonstrate the efficiency and versatility of PGO with two different applications: (1) the parallelization of a large scale parameter estimation problem associated with modeling water flow in a heterogeneous deep vadose zone; (2) the parallelization of a complex simulation-optimization procedure for searching for an optimal groundwater remediation design. PGO is developed as an open source code, and is independent of the computer operating system. It has been tested in a heterogeneous computing environment consisting of Solaris 9, Fedora Core 2 Linux, and Microsoft Windows machines, and is freely available for download from http://grid.scut.edu.cn/PGO/.