GPU Computing Using CUDA in the Deployment of Smart Grids

This paper underscores the use of CUDA-based GPUs as high performance parallel computers for the purpose of real time analysis in a smart grid setting. In a smart grid, with the influx of new, renewable, distributed generation technologies, the network is more complex and requires more computationally intensive means of simulation and analysis. To show its usefulness, a power flow analysis case study will be programmed in CUDA C++ and its performance benchmarked against a sequential CPU counterpart. The results show that the GPU performs better than single-threaded CPU programs, in terms of execution time. A lack of optimization in GPU programs decreases the potential performance benefits, however, as system size increases, the scalability advantages afforded by the CUDA model are evident. The results also show that performance is GPU-platform dependent, i.e. dependent on GPU architecture and power.