2025 Spring Meeting and 21st Global Congress on Process Safety

(169a) Optimization of Blue Hydrogen Plant Operation Via Hybrid Plant Model

Authors

Vladimir Mahalec - Presenter, McMaster University
Raunak Pandey, McMaster University
Farbod Maghsoudi, McMaster University
Blue hydrogen plant produces hydrogen form natural gas via auto-thermal (ATR) and water-gas shift (WGS) reactors. The plant includes a combined heat and power (CHP) unit which produces electricity and steam for the plant needs. Electricity can also be exported or imported to the grid. Oxygen used in ATR is produced by an air separation unit. Hydrogen is shipped via hydrogen gas pipeline to the customers or is lifted via trucks as liquid hydrogen for local consumption. Electricity prices change frequently and have a significant impact on plant operation.

Hybrid model of this hydrogen plant has been built based on a novel modelling paradigm which replaces mole flows and mole fractions by mass flows of components. This has enabled the use of accurate local approximations of physical properties (unit enthalpy, heat capacity, heat of vaporization) at local stream conditions. Mass based component flows have also made it possible to model the ATR and the WGS reactors by a set of linear equations instead of using RGIBBS mole-fractions based nonlinear model. The entire model is almost linear (only a handful of bilinear terms).

Accuracy of the hybrid plant model has been assessed by comparing it with an AspenPlus model.

Novel, multi-phase composite optimization algorithm has been implemented. Convergence towards a solution starts with a solution of mass balances and energy balances with stream enthalpies fixed at the stream conditions. This is followed by calculation of energy balances at temperatures corresponding to the actual plant operation. Finally, in the outer loop, stream properties are updated from a rigorous thermophysical properties package and the model is re-optimized. It is shown that the optimization results have accuracy comparable to AspenPlus.

Rapid execution times, linearity of the model, and its accuracy, make it suitable for constructing multiperiod planning models or to use it for real-time optimization.