Real-time proactive control of cascading failures in integrated electricity–gas systems based on a privacy-preserving physics informed deep operator surrogate model

As the coupling between the power system and the gas network increases, the risk of fault propagation between the two systems also escalates, jeopardizing the safe operation of integrated energy systems. However, the computational inefficiency of dynamic energy flow analysis using traditional numerical methods makes it challenging to meet the requirements of real-time emergency control. Additionally, direct model and data sharing between these systems remain impractical. To address these challenges, this paper presents fast proactive control for cascading failures in integrated electricity and gas systems (IEGS), leveraging physics informed gas network surrogate model to significantly expedite the security analysis process. The proposed framework integrates physics informed Deep Operator Neural Network (PI-DeepONet) for fast energy flow computation under fault conditions, coupled with an autoencoder for data compression and encryption. The proposed method is further combined with a real-time application algorithm for proactive control. Numerical case studies demonstrate that the method effectively predicts the dynamics of the gas network, while ensuring the privacy of operational data and models. Besides, the proactive control signals calculated by the proposed method provide the power system with available escape time to respond to the faults in the gas network, thereby reducing potential losses.