Development of a Methodology for the Resilience Assessment of Alternative Maritime Fuel Transport and Port Infrastructure

Abstract

The implementation of alternative fuel options in the maritime sector is essential, not only from the perspective of emissions and environmental pollution reduction, but also from supply resilience. In this sense, methanol and ammonia are two of the most promising and studied alternatives. However, experience in their utilization as maritime fuels is still limited. Therefore, it is essential to work towards their safe and successful deployment, considering the risks associated with them. Resilience assessment methods provide the opportunity to analyze system performance and reliability in a comprehensive manner, as they focus not only on pre-failure scenarios, but also post-failure phases. Within the existing quantitative resilience assessment methods, one of the most widely used in literature are Bayesian Networks (BNs), which stand out for their ability to handle causalities between different variables in probabilistic terms, as well as for their adaptability and capacity to evaluate system behavior over time. In this context, the main objective of this work was to develop a methodology to quantitatively analyze the resilience of the ammonia and methanol transport, storage and bunkering stages of the supply chain for the maritime sector based on the creation of a BN model. This was achieved via the derivation of a bow-tie (BT) model, from which indicators for possible disruptions to the systems, as well as for their resilience attributes (i.e., their absorption, adaptation and restoration capacities) were identified. All of these elements were incorporated into two BN models: one for methanol- and another for ammonia-based systems. While the developed models still require the input of the prior and conditional probabilities of the elements included within them before being able to be applied, they constitute a base that can be utilized to structure and program the BNs using software solutions, allowing to perform probabilistic resilience assessments in the future. The derivation of the models also led to the identification of central disruptions that the analyzed systems might incur into, with fuel releases standing out as the most prominent within them, especially when originating from damages to the fuel transfer or storage infrastructure. Additionally, it was possible to understand the differences and similarities between methanol- and ammonia-based systems, as well as to generate recommendations for improving their resilience based on the analysis of their attributes.Compared to the existing work in literature, the methodology presented offers the advantage of incorporating processes from multiple stages of the supply chain, whereas previous analyses have been limited to specific processes or supply chain stages, such as fuel bunkering or storage only. Additionally, it was developed specifically for resilience assessment, rather than risk assessment, which focuses exclusively on predisruption stages. Future lines of work for the models include their application to analyze specific case studies of interest; to consider possible interdependencies between different supply chain stages; to convert the models from static to dynamic by incorporating a temporal dimension as well as the learning capacity of the system; and lastly to explore the use of artificial intelligence (AI) and machine learning (ML) tools to enhance the resilience assessment of the systems.