Análisis de factibilidad del uso de unidades de procesamiento gráfico en algoritmos de optimización meta-heurísticos

Abstract

Currently, several real-world optimization problems have been mathematically modeled. The modeling process takes into account as much information as possible so that the model obtained is as close to reality. However, as information increases, complexity also increases. Consequently, larger computational capacity is needed to solve complex and scalable problems. Exact or mathematical optimization methods are considered exhaustive, which for multi-dimensional problems is not practical. As a result, meta-heuristic algorithms have been developed to solve complex optimization problems. These algorithms are commonly used for problems of two or more dimensions, involving vector and matrix operations. Therefore, to solve this type of problem, it is useful to carry out parallel processes that reduce the runtime. Currently, there are multi-core CPU that can easily solve small problems with parallel calculations. However, the GPU offers the potential to improve performance by incorporating a larger number of cores than the CPU, which is very useful for problems with multiple processes in parallel. A classic problem in the electrical power systems research community is the Transmission Expansion Planning (TEP). In the Department of Electrical, Electronics, and Telecommunications Engineering, in the project “Use of the AC model for multi-stage integrated planning of transmission / sub-transmission line expansion and reactive power sources in electric power systems”, It is tried to improve the performance of the TEP problem. The project seeks to reduce the calculation time by improving the meta-heuristic algorithms. In this final d egree p roject, a s a c omplement t o the aforementioned project, the feasibility of reducing the execution time was analyzed, implementing the meta-heuristic algorithms using a GPU. For which, prior to the implementation of the TEP problem, the use of the GPU in parallelizable problems and the implementation of a meta-heuristic algorithm in a two-dimensional problem were analyzed, in which favorable results have been obtained to motivate the use of the GPU