Dynamic-model-based artificial neural network for H2 recovery and CO2 capture from hydrogen tail gas

Abstract

Herein, we developed an integrated process for H2 recovery and CO2 capture from the tail gas of hydrogen plants. The front-sector system (cryogenic, membrane, and compressor units) involved CO2 capture and supply of H2-rich gas to the rear-sector system (heat exchanger (HX) and pressure swing adsorption (PSA) unit) for H2 recovery. The developed dynamic model of the integrated process was validated through reference data. The parametric study highlighted the potential of the developed process for high-purity H2 recovery and CO2 capture. Owing to the complexity of the interconnections, a dynamic-model-based artificial neural network (ANN) for the integrated process was developed to optimize the process performance. The synthetic datasets for the ANN were analyzed by singular value decomposition, and the ANN models for the cryogenic, membrane, and PSA units were trained and tested within a marginal error (<2%). Subsequently, a process-driven model (the integration of the ANN models with the algebraic equations (compressor, HX, and economic evaluation)) was validated through minute deviations from the reference data. The optimization, formulated based on the process-driven model, was conducted using differential evolution. The optimum cost (2.045 $/kg) of recovered H2 (99.99%) was economically comparable to the reference values for H2 production from natural gas. Furthermore, the cost was covered for 91% CO2 capture with 98.6 vol.% CO2. Thus, the result can bridge the gaps in research, development, and implementation and between fossil and renewable energy. Dynamic-model-based ANN can precisely predict the dynamic behavior and optimum performance of an integrated process at a low computational cost.