Path Recommendation with Graph Neural Network in Power Communication Management Systems

With the applications of artificial intelligence (AI) technology, graph neural networks (GNNs) have emerged and employed as a framework to handle tasks with graph-structured data. On the other hand, enhancing the efficiency of communication management system is an urgent demand to the power grid enterprises. In this paper, we propose a path recommendation framework with GNN for power communication management systems. First, historical data are collected as datasets. Second, the historical data are converted to structured form data, which are further used to construct graphs. Finally, with the proposed GNN, paths are recommended by the softmax layer, in which the one with the highest probability is the result. This scheme is verified with real data in a communication management system, where one optimal path and several candidate paths are reported in practice.