Analysis of ammonia dispersion behavior under the influence of operational and weather conditions during ship bunkering

Background

Maritime transport plays a significant role in global pollution and greenhouse gas (GHG) emissions, contributing to about 3.3% of the world’s fuel consumption emissions according to the International Energy Agency. Recognizing the urgent need to reduce carbon dioxide emissions to combat climate change, the international shipping industry is stepping up its efforts. The International Maritime Organization (IMO) has introduced a revised strategy to reduce GHG emissions from international shipping, setting the ambitious goal of reaching net-zero emissions by 2050.

In alignment with the IMO’s objectives, ammonia is emerging as a prominent choice for shipping fuel due to its carbon-free nature. With an energy density of 22.5MJ/kg, it is comparable to carbon-based fuels like methanol (22.7MJ/kg) and ethanol (29.7MJ/kg). It can be efficiently liquefied by compressing it to 0.8MPa at 20 ◦C or by cooling it to -33 ◦C at atmospheric pressure, facilitating its storage and transport. The infrastructure for ammonia’s production, storage, and distribution is well-established and reliable, with a global production reaching 150 million tonnes in 2019. However, the primary concern with the use of ammonia lies in its toxicity. At various acute exposure guideline levels (AEGLs), ammonia presents significant health hazards, with effects ranging from transient (AEGL-1, 30 ppm) to irreversible (AEGL-2, 160 ppm), and potentially life-threatening or fatal (AEGL-3, 1100 ppm) within one hour. The dispersion of ammonia release can be influenced by various factors, including meteorological conditions (such as wind speed and direction, and relative humidity) and release characteristics (such as duration and flow rate). However, how these factors influence ammonia dispersion is yet to be fully understood. Further research into the dispersion characteristics of ammonia and its time-dependent plume behaviour under various conditions is essential for effective risk assessment and management.

A wide range of atmospheric models are available for particle dispersion, which are derived based on the mathematical equations describing the atmosphere, dispersion and chemical and physical processes within the plume to calculate the concentrations at various locations. Box models and Gaussian models are typical atmospherical models used in dispersion modelling. They can provide fast estimation of dispersion without dynamically resolving the flow, and the results are relatively reliable when describing unobstructed gas flow over flat terrain.

In recent years, the computational fluid dynamic (CFD) approach has been growing fast with the rapid advances of high-performance computing systems. It solves the conservation equations of mass, momentum and energy, hence the dynamics of the flow are obtained and the results are expected to be more accurate, especially for dispersion of gas flow in complex geometry domain. Fig. 1 presents an example of dispersion comparison between LNG and ammonia from CFD simulations. Alternatively, many researchers have introduced machine learning (ML) models into atmospheric dispersion prediction, such as the artificial neural network (ANN). They have been used to enhance the accuracy of emission inventories and the performance of air quality models through a back-propagation approach that adjusts the gradient of the loss function, which measures the deviation between predicted and observed contaminant concentrations.

Research Objectives

The primary goal of this project is to enhance our understanding of how ammonia spreads under various operational and weather conditions during ship bunkering, which will help in assessing and mitigating the risks related to ammonia operations for marine vessels. A high-fidelity numerical model will be implemented in OpenFOAM, a widely recognized open-source CFD library. It provides powerful flow modelling tools to academic and industry users and offers researchers a high-level programming platform to customise and develop CFD codes freely. The data generated by this CFD model will be utilized to train a machine-learning (ML) model for fast prediction of ammonia dispersion. This dual approach, combining high-fidelity numerical modelling with ML-driven predictions, promises to offer unparalleled insights into ammonia dispersion dynamics, contributing to the improvement of risk assessment and industrial practices.

Supervision Team

• Dr Hao Chen

• Prof. Cheng Siong Chin

• Ros Blazejczyk

• Simon Hindley