Learn to decompose multiobjective optimization models for large‐scale networks

Abstract Infrastructures can be modeled as large‐scale networks consisting of nodes and arcs, making network optimization a popular modeling option for arising problems. In specific, providing timely restoration plans interdependent infrastructures facing disruptions has been challenge decision makers. this study, we focus on geospatial (co‐location) functional interdependencies to capture the impact cascading failures infrastructure systems. The dynamics real are more complicated captured by one objective function. Therefore, define three functions in pillars sustainability: (a) economic, (b) social, (c) environmental. To solve multiobjective model, develop learn‐to‐decompose framework, evolutionary algorithm based decomposition module Gaussian process regression (GPR) periodically learn from obtained Pareto front guide search direction. We also included heuristic address two significant challenges restoring infrastructures: island scenario co‐location interdependencies. applied proposed framework benchmark problems water transportation City Tampa, FL. carried out sensitivity analyses monitor performance GPR different kernel functions. provided insights makers finding trade‐off between fortification (proactive) (reactive) costs. result demonstrates is feasible applicable networks.