Optimal stochastic dynamic scheduling for managing community recovery from natural hazards

Abstract

Following the occurrence of an extreme natural or man-made event, community recovery management should aim at providing optimal restoration policies for a community over a planning horizon. Calculating such optimal restoration polices in the presence of uncertainty poses significant challenges for community leaders. Stochastic scheduling for several interdependent infrastructure systems is a difficult control problem with huge decision spaces. The Markov decision process (MDP)-based optimization approach proposed in this study incorporates different sources of uncertainties to compute the restoration policies. The computation of optimal scheduling schemes using our method employs the rollout algorithm, which provides an effective computational tool for optimization problems dealing with real-world large-scale networks and communities. We apply the proposed methodology to a realistic community recovery problem, where different decision-making objectives are considered. Our approach accommodates current restoration strategies employed in recovery management. Our computational results indicate that the restoration policies calculated using our techniques significantly outperform the current recovery strategies. Finally, we study the applicability of our method to address different risk attitudes of policymakers, which include risk-neutral and risk-averse attitudes in the community recovery management. become complicated. The most important characteristics of a rational decision-making approach include:i. The agent must balance the reluctance for low immediate reward with the desire of high future rewards (also referred as "non-myopic agent" or look-ahead property); ii.The agent must consider different sources of uncertainties; iii. The agent must make decisions periodically to not only take advantage of information that becomes available when recovery actions are in progress but also to adapt to disturbances over the recovery process; iv.The agent must be able to handle a large decision-making space, which is typical for the problems at the community level. This decision-making space can cause an agent to suffer from decision fatigue; no matter how rational and high-minded an agent tries to be, one cannot make decision after decision without paying a cost [1]. v.The agent must consider different types of dependencies and interdependencies among networks, because a single decision can trigger cascading effects in multiple networks at the community level. vi.The agent must be able to handle multi-objective tasks, which are common in real-world domains. The interconnectedness among networks and probable conflicts among competing objectives complicate the decision-making procedure. vii.The agent must consider different constraints, such as time constraints, limited budget and repair crew, and current regional entities' policies. viii.External factors, like the available resources and the type of community and hazard, shape the risk attitude of the agent. The different risk behaviors must be considered.Community-level de...

Impact of Demand Response on Reliability Enhancement in Distribution Networks

“…Also, severe operational occurrences and the effects of demand response programs on distribution system performance in the restoration period were studied for restoring a distribution system. In [53], an approach based on Markov decision making was proposed for calculating restoration policy. In [54], a graph theory based on a particle swarm optimization (PSO) algorithm was presented for distribution system restoration.…”

Deep reinforcement learning driven inspection and maintenance planning under incomplete information and constraints

“…These approaches are also applicable to infrastructure problems beyond inspection and maintenance planning, such as post-disaster recovery, e.g. in [19,20,21]. Respectively, admissible solution strategies to the above approaches span from exhaustive policy enumeration, and genetic algorithms, to gradient-based schemes, and dynamic programming.…”

Self‐healing optimization in active distribution network to improve reliability, and reduction losses, switching cost and load shedding

“…The Markov decision process‐based approach is proposed to compute the restoration policies. The proposed method is applied to a realistic community recovery problem …”

“…Another important issue to consider in this article is the use of backup feeders for system restoration. Backup feeders are only used in References for the restoration problem. In Reference , load shedding and islanding state are not considered for restoration.…”