Efficient parallel Particle Swarm Optimizers on GPU for real-time harmonic minimization in multilevel inverters

In this paper, we present a parallel implementation of the Particle Swarm Optimization (PSO) on graphical processing units (GPU) using CUDA. By fully utilizing the processing power of graphic processors, we show how our parallel CUDA-PSO can be used to minimize harmonics in multi-level inverters. The computing power of the GPU coupled with the parallelism of our algorithm allows for real-time computation of optimal switching angles for multilevel inverters with several DC inputs while minimizing the 50^th first harmonics. Compared to other solutions, our CUDA-PSO offers superior converging rate by running hundreds of independent swarms in parallel without increasing the computation time.