Session: 08-01: Poster Session

Paper Number: 148393

148393 - Cycle-Resolved Control of Ammonia Dual-Fuel Engine Using Spiking Neural Networks With Online Learning 

Abstract: 

The heavy-duty demands and long travel distances of marine shipping, locomotive, and many off-road applications present a barrier to electrification. Therefore, these sectors will continue to depend on liquid fuels for the foreseeable future, making the adoption of low-lifecycle-carbon fuels (LLCF) essential for their decarbonization. Green hydrogen, methanol, and ammonia sourced from renewable electricity are some of the LLCFs that will play a crucial role in this transition. In the marine sector, a dual-fuel approach is expected to be employed, where the majority of the fuel energy will come from LLCFs, supplemented by a small diesel pilot injection to initiate combustion. However, incorporating new LLCFs, particularly ammonia, poses challenges due to their substantially different combustion properties compared to traditional fuels. To enable effective and clean combustion of these difficult-to-burn LLCFs like ammonia, control parameters such as fuel quantity, injection timing, and pilot injection need to be adjusted.

To address this issue efficiently, advanced machine learning algorithms with online learning capabilities are used to adjust control parameters on a cycle-by-cycle basis. Neuromorphic computing techniques such as spiking neural networks (SNN)—which are well-suited to real-time time-series control problems—are implemented to adjust the diesel injection quantity and timing to maintain combustion stability while maximizing efficiency and minimizing emissions. The dynamic nature of online learning enables continuous adaptation and optimization, making it easier to cope with the varying fuel availability at ports, especially in the marine sector.

Presenting Author: Brian Kaul Oak Ridge National Laboratory

Presenting Author Biography: Dr. Kaul received his Ph.D. degree in Mechanical Engineering from Missouri University of Science and Technology in 2008. He currently holds the position of Senior Research Staff in the Buildings and Transportation Science Division at Oak Ridge National Laboratory and has coauthored over fifty refereed publications and a book chapter. His current research includes projects related to the characterization and control of cyclic combustion instabilities, marine diesel engine efficiency and alternative low-carbon fuels, and thermodynamic analysis of engine and powertrain systems.

Authors:

Brian Kaul Oak Ridge National Laboratory
Bryan Maldonado Oak Ridge National Laboratory
Catherine Schuman University of Tennessee