Edge-to-Cloud Computations-as-a-Service in Software-Defined Energy Networks for Smart Grids

Abstract

Modern electricity grids run latency-sensitive functions, protection relays, fault isolation, and microgrid control, that demand millisecond analytics at the edge, while energy-hungry analytics workloads sit in distant clouds, causing missed realtime deadlines and wasted power. We formulate the placement of these grid computations as a service-management problem over a Software-Defined Energy Network (SDEN) that cooptimizes energy and latency across an edge-fog-cloud continuum under Ultra-Reliable Low-Latency Communications (URLLC) constraints. Three contributions follow. First, we develop a queueaware joint offloading model in which node latency increases with the assigned load. The placement then couples all tasks through per-node congestion, and the resulting program is binary nonlinear; we derive its analytical (Lagrangian) structure. Second, since exact solutions do not scale, we present a lightweight greedy heuristic with linear time complexity with respect to the number of tasks and candidate nodes and carries a proven feasibilitypreservation property under load-dependent latency and sharedbandwidth constraints. Third, a tiered federated Graph Neural Network provides reactive detection at the edge, prediction in the fog, and coordination in the cloud. In a discrete-event simulation, the SDEN reduced energy versus a cloud-only baseline by a median of 30.2% (up to 69.65%), lowered transmission delay and jitter under URLLC, and, under a common-cause reliability model, cut effective annual downtime from about an hour to minutes. All the results are simulation-based and demonstrate a service-management path toward grid-scale Computations-as-a-Service.