Estratégias de paralelismo com GPGPU para otimização do processamento do cálculo do fluxo de carga em sistemas elétricos de potência

Abstract

The load flow calculation provides fundamental information for an electric power system. However, the load flow can only be carried out in the steady state, in the event of a system suffering any change, by variation in the loads or modifications of the control equipment, this calculation is necessary to be redone. Because of this need, frequently have to perform the load flow, a research has begun to optimize the time needed for this task. A General-Purpose Graphic Processing Unit (GPGPU) as a cost-effective alternative to parallel architecture runs, which has a GPU not only for graphics purposes but also for general purposes. Several works were taken for the use of GPGPU in load flow calculations, there is no consensus on the content of the material, being in charge of each one of the work of its own method, making it difficult to use the architecture for an implementation of calculations, both for academic purposes and for the market. Due to the lack of consensus and differences found in the work, this dissertation aims to analyze the steps of the load flow, identifying which is more suitable to parallelize in GPGPU in order to perform simultaneous load flow calculations and reduces the time required for the processing, an efficient strategy for large scale systems in the market and not academic environment, facilitate the replication for future works using metaheuristics for optimization of power electrical systems.