Semantics-Oriented Cloud-Edge-End Collaborative Power Allocation in Smart Grids

When semantic communication meets the ever-growing complex smart grids, it has great potential in many smart grids-related applications. However, the power consumption from end devices would inevitably impact the performance of the whole smart grids. To address above issue, this article investigates a semantics-oriented cloud-edge-collaborative power allocation scheme. Specifically, we define the special goals of cloud layer, edge layer and end layer respectively. In end layer, we divide all devices into separate groups based on the required semantic tasks, then the edge layer calculates the semantic importance scores of each group using attention mechanism based on their semantic performance and channel environment. Finally, the cloud layer decides the power allocation based on the attention scores and the importance of each task to the specific groups of devices that achieve the maximum system performance. Simulation results demonstrate the effectiveness of our proposed power allocation scheme.