Johnson Matthey: Dynamic Modeling for Green Energy Production: Paving the Way to a Net-Zero Future

This project addresses the challenge of fluctuating renewable energy sources and the need for flexible energy storage solutions in a net-zero society. Focusing on green fuels like electrolysis-based hydrogen converted to methanol or ammonia, the study aims to develop a dynamic model for methanol production, responding to changes in inputs and disturbances. By analyzing a comprehensive multivariate time-series dataset from JM, we applied involves feature analysis, uncertainty quantification, and the use of neural network methods like LSTM, NARX, and Transformers. Our model leverages dynamic simulation and machine learning to predict fluctuations and trends in methanol production, crucial for Johnson Matthey’s operational efficiency. By integrating these models, JM gains a significant advantage in forecasting and managing the complexities of their chemical production, enhancing both the precision and reliability of their industrial processes in a sustainable and energy-efficient manner.

Joint work with Johnson Matthey and The Alan Turing Institute DSG, 2023. The final report is under review.