Abstract:Parameter estimation (PE) is one of the most challenging problems in photovoltaic (PV) system modeling. Owing to the ability to handle nonlinear functions regardless of the derivatives information, meta-heuristics have attracted many researchers. Recently, many implementations of particle swarm optimization (PSO) based PE method have been proposed in the literature. However, these algorithms utilize multiple agents or particles in the search process, and are normally compute intensive. In this paper, we describe our implementation of PSO on graphic processing units (GPUs) using open computing language (OpenCL). The proposed method has been specifically designed and entirely executed on the GPUs to provide a reduction of computational costs. Results show that the GPU-based PE is faster in comparison with its sequential implementation of PSO, and this proves the efficacy of the GPU framework.

GPU-Based Parameter Estimation Method for Photovoltaic Electrical Models