Optimisation of Water Management Systems Using a GPU-Accelerated Differential Evolution

The aim of this paper is to present a tool that could optimise water levels in dams to provide the inhabitants and industry with enough fresh water while keeping a reasonable flood protection and river navigability. Since the water systems may be quite complicated and the outflow plans have to be precise, we developed a GPU-accelerated algorithm based on the differential evolution to solve this task. The experimental results show that a high quality annual outflow plans can be obtained within a minute, almost 18 times faster than on a hex-core CPU.