Efficient Embedding of Neural Network-Based Stability Constraints Into Power System Dispatch

Neural networks have shown great potential to learn complex stability constraints for power system operation with high renewable penetration. However, explicitly embedding neural network-based stability constraints into power system dispatch is computationally intensive for online applications. This letter presents an efficient method to embed neural network-based stability constraints into power system dispatch. The neural network-based stability constraints are embedded into the optimization problem in linear form iteratively. Case studies on NPCC 140-bus system and a realistic power system demonstrate the effectiveness and efficiency of the proposed method.