Topology-Aware Neural Networks for Fast Contingency Analysis of Power Systems

Training Neural Networks able to capture the topology changes of the power grid is one of the significant challenges towards the adoption of machine learning techniques for N-k security computations and a wide range of other operations that involve grid reconfiguration. As the number of N-k scenarios increases exponentially with increasing system size this renders such problems extremely time-consuming to solve with traditional solvers. In this paper, we combine Physics-Informed Neural Networks with both a Guided-Dropout (GD) Neural Network (which associates dedicated neurons with specific line connections/disconnections) and an edge-varrying Graph Neural Neural Network (GNN) architecture to learn the setpoints for a grid that considers all probable single-line reconfigurations (all critical N-1 scenarios) and subsequently apply the trained models to N-k scenarios.We demonstrate how incorporating the underlying physical equations for the network equations within the training procedure of the GD and the GNN architectures, performs with N-1, N-2, and N-3 case studies. Using the AC Power Flow as a guiding application, we test our methods on the 14-bus, 30-bus, 57-bus, and 118-bus systems. We find that these topology-aware NNs not only achieve the task of contingency screening with satisfactory accuracy but do this at up to 1000 times faster than the Newton Raphson power flow solver. Moreover, our results provide a comparison of the GD and GNN models in terms of accuracy and computational speed and provide recommendations on their adoption for contingency analysis of power systems.