Reconfiguration of Photovoltaic Arrays Based on a GPU-Accelerated Exhaustive Search Algorithm

This paper proposes a parallel exhaustive search algorithm to find the best electrical configuration of PV arrays. This information is required to reconfigure, in real-time, the PV array in order to mitigate the power losses caused by partial shading and other mismatching conditions. The algorithm is based on a parameterized model of the PV system, and it is designed to run in commercial graphic processing units (GPU). This GPU-accelerated approach provides a significant reduction in the processing time compared with the classical CPU-only serial algorithm. Hence, the proposed GPU-based reconfiguration enables to reconfigure, in real-time, PV systems much larger contrasted with the CPU-only approach.