Neural network and meta heuristic-based method for backup controller assignment in SDN-Smart grid

Smart grids or the future of power systems, integrate bidirectional electricity and data flows to establish a highly distributed and automated energy distribution framework. Their proper operation relies on time-sensitive services. This paper introduces a Quality of Service-aware (QoS-aware) fault recovery mechanism for Software-Defined Network-based Smart Grids (SDN-SG). Addressing control plane fault recovery in SDN is an NP-hard problem. A key innovation of the proposed approach is the reduction of this exponential computational complexity through using a holonic multi-agent system. The paper provides mathematical proof that the hierarchical organization of holonic systems enhances time complexity by utilizing a divide-and-conquer strategy. The proposed method represents the first predictive control plane fault recovery mechanism, which assigns backup controllers by forecasting future network load conditions using neural networks. Backup controller selection is framed as an Integer Programming problem, with a meta-heuristic strategy employed to identify near-optimal solutions. The accuracy of this strategy is improved by narrowing the solution space through holonic organization. This claim is validated by comparing the results with state-of-the-art meta-heuristic-based approaches. Experimental findings highlight the superiority of the method over existing techniques in terms of control plane load balancing, packet loss rate, recovered packet percentage, and backup controller selection overhead.