Optimal power flow based on parallel metaheuristics for graphics processing units

The smart-grid brings new challenges to the optimal dispatch of power. Current research aims to develop optimization techniques capable of handling large networks using accurate models and realistic constraints, all in the shortest possible execution time. For this purpose, this paper presents a metaheuristic-based parallel optimal power flow algorithm for graphics processing units (GPUs). Metaheuristics have the advantage of handling discrete variables and being resilient to premature convergence towards local optima. However, they require significant computing power which limits their use in on-line applications. The proposed implementation addresses this limitation and significantly accelerates the calculation by exploiting the massively parallel architecture of GPUs. The developed software uses a particle swarm optimizer and runs a full ac Newton-Raphson power flow analysis to evaluate the candidate solutions. The algorithm is tested on the IEEE 30-bus, 118-bus and 300-bus networks and provides a maximum speedup of 17.2×.