A Successful Parallel Implementation of NSGA-II on GPU for the Energy Dispatch Problem on Hydroelectric Power Plants

Abstract

Nowadays, hydraulic sources are responsible for most of the Brazil's energy production. Hydroelectric power plants (HPP) operators in Brazil usually distribute equally the total power required among the generator units available in the plant. However, studies show that this configuration does not guarantee that each generator unit operate close to its optimal operation point. The energy dispatch optimization problem consists in determining which generation units need to be on or off and what is their respective power-set, so that both the overall HPP costs is minimized and the power required by the plant is met. This paper presents a GPU-based parallel implementation of NSGA-II, to solve the energy dispatch problem of a HPP complaying with the real time restrictions posed by the operation of a real HPP from the reception of the power demand to the energy dispatch. Our implementation obtains better solutions than the sequential implementation currently available.