Accurate optical model for design and analysis of solar fields based on heterogeneous multicore systems

This work deals with parallel algorithms for the optical design and analysis of solar fields for Central Receiver Systems. The proposed implementation based on desktop workstations equipped with multiple Graphics Processing Units (GPU) is motivated by the need to have an accurate and fast simulation environment for studying mirror imperfection and non-planar geometries. Since heliostat field layouts can be composed of thousands of mirrors, and for each one a complex 3D geometry problem must be solved, this can take a long time even with highly optimized CPU-based solvers. The GPU-accelerated solver outperforms an optimized OpenMP-based reference, running on 8 CPU cores, up to 52× depending on the complexity of the problem, at the cost of a very modest hardware upgrade of two GPU GTX 590 and one GTX 480 graphics cards. Several applications demonstrate that the parallel algorithm preserves flexibility to model complex mirror geometries and non-idealities.