“…Graph Neural Networks (GNN) have been applied to address many problems in power systems [27]. For example, the application of GNN for learning to branch in B&B algorithms has been demonstrated to effectively reduce the running time of MIP [28][29][30]. DeepMind [31] develops a GNN model proposed in [28] which is built from the MIP model and applies it to neural diving and neural branching for solving problems including electric grid optimization [32].…”