Multi-criteria optimization of budget allocation for resilience improvement of interdependent critical infrastructures exposed to natural hazards

The resilience of interdependent Critical Infrastructures (CIs) can be strengthened through targeted investments to mitigate the economic and social consequences of disruptive natural hazards. The optimal allocation of limited resources is challenged by the dynamic and conflicting preferences of decision-makers, as well as by the uncertainty associated with Climate Change (CC). In this work, the budget allocation problem is formulated as a Sequential Decision Problem (SDP), which is solved with Deep Reinforcement Learning (DRL) to identify the investment policy that minimizes the expected losses caused by disruptive events and the investments cost during the CIs lifetime. The novelty of the approach lies in the integration of: i) multi-criteria decision making to account for evolving stakeholder preferences within the SDP, ii) the probabilistic modelling of investments effectiveness according to CIs response to CC-induced natural hazards. The framework is applied to interdependent power and gas networks subject to coastal flood-induced disruptions. Results show that the method adapts to changes in stakeholder preferences as well as changes in natural hazards intensity, providing a flexible approach for resilienceoriented investment planning under uncertainty.

Keywords: Critical Infrastructures (CIs), Resilience, Sequential Decision Problem (SDP), Climate Change (CC), Deep Reinforcement Learning (DRL).