A Leader-Follower Based Parallel Accelerated Particle Swarm Optimization Algorithm for Smart Grid Resource Allocation

Smart grid resource allocation (SGRA) is a large-scale demand response optimization problem to lower the electricity cost of customers by intelligently shifting their schedulable devices via an aggregator. The SGRA problem aims to maximize the aggregator profit while still considering the inconvenience and the availability constraints of each customer's assets. A heuristic approach provides a reliable method to find near-optimal solutions to non-convex, complex, and multi-dimensional problems by reducing the computational time and not relying on gradients. To further improve the computational time of the heuristics, high-performance parallel computing concepts can be used to divide the formulated problem into smaller sub-problems. This research proposes a leader-follower based parallel Accelerated Particle Swarm Optimization (APSO) to find near-optimal solutions of SGRA using the OpenMP framework in MATLAB. The leader-follower topology allows the parallel execution of each particle, resulting in lower execution time. The proposed parallel APSO was both significantly faster (2 - 7×) than existing SGRA literature methods, and achieves similar aggregator profit, customer savings, and peak demand reduction.