Resilient Infrastructure Systems: Robustness And Distributed Coordination

Abstract

The increasing interconnectivity of current critical infrastructure systems, including power grids and transportation infrastructure, is transforming them into a system-of-systems that is susceptible to catastrophic cascading failures triggered by natural disasters, cybersecurity breaches, and operational disruptions. Current literature lacks an effective approach towards capturing the interdependencies among these systems, along with the dynamic nature of their recovery process. Moreover, it fails to offer an appropriate decision-making mechanism to manage their resilience. To overcome these constraints, this research offers a resilient infrastructure system solution based on the coupling of power grid and transportation networks, employing a probabilistic interdependence model in conjunction with reinforcement learning optimization via Dynamic Namib Beetle Optimized Soft Actor-Critic (DNBO-SAC). This method incorporates a distributed coordination scheme to facilitate decentralized yet globally coherent decisions on managing infrastructure systems. Research employs a Kaggle dataset based on transportation network resilience, which includes traffic patterns, environmental disruptions, and infrastructure recovery characteristics. For data pre-processing purposes, the Min-Max normalization technique is employed to scale all features, and Principal Component Analysis (PCA) is adopted to select dominant spatiotemporal attributes. The suggested DNBO-SAC model provides better results of least Mean Absolute Percentage Error (MAPE) of 9.74% and Mean Absolute Error (MAE) of 3.28. The proposed approach implemented by utilizing the Python language. The model incorporating DNBO-SAC offers a great way to enhance the interdependent power grid and transportation system resilience through effective recovery strategies and minimizing cascading failures.