Towards a massively-parallel version of the SimSEE

Abstract

The SimSEE is a simulation software used/designed to aid the decision-making in the electric energy generation market. It is based on Stochastic DynamicProgramming technique and allows to simulate the contribution of several energy sources, such as hydro-electric, solar, thermal or wind energy, to a specific electrical network. Uruguay’s electric generation system has considerably grown and diversified in the past decades. This evolution implies potentially more complex scenarios and also motivates a more precise modeling of some electric sources. Therefore, the computational cost of the simulations is also expected to rise and the use of HPC techniques becomes mandatory. In this work we study the performance bottlenecks in the SimSEE tool. Additionally, and considering the previously mentioned results, we design a parallelization strategy that enables its acceleration using massively-parallel devices such as GPUs.