Numerical simulations of industrial-scale packed-bed adsorbers

Licentiate thesis, 2020

A common technique for removing harmful substances, for either persons or equipment, from a stream is to use the phenomenon of adsorption. This technique is used in, for example, water purification, personal protection, chemical and pharmaceutical production and biomass gasification. In a biomass gasification plant called GoBiGas in Gothenburg, Sweden, beds packed with activated carbon were used to remove light tars from the product gas, where the tars adsorbes to the carbon. This was a step in the gas cleaning train used for the production of bio-methane from forest residue streams in an effort
towards a more carbon neutral future. However, the operation of packed beds for continuous substance removal both requires energy, and puts certain technological demands on the plant. For the production of bio-methane, all losses in the plant influence not only the economical aspects, but also the carbon footprint of the end product. Since carbon neutrality is a compelling reason for using bio-methane, this puts further demands on the bed operation. Here, numerical simulations offer increasing opportunities, both with respect to energy optimization, but also to bed design and operation.

In this work, we formulate a numerical model for an industrial sized adsorber, used in GoBiGas for benzene removal. The end goal includes an increased general understanding of the requirements for a successful numerical model of real-world, industrial conditions. This is done in order to be able to better design and optimize packed bed setups for industrial conditions before the actual facilities are built. However, the work also allows to better understand and optimize setups already online. The work presented here includes both analysis of industrial data from GoBiGas and an establishment of how a baseline numerical model performs.

The numerical model is based on solving the governing equations for the system, with no industry-specific parameter tuning. This is important in order to be able to use the models as a predictive tool, useful in e.g. bed design. A finite volume method is used to numerically simulate the flow, mass- and heat-transport in both time and space. The temperature at different axial positions in the bed is used to compare the numerical simulations with the industrial data. We show that a baseline formulation captures the main characteristics of the temperature signals in a bed but there are dynamics of the
industrial data that are not captured. Three areas are identified that require additional development for a better predictability. Those are that a more complete description of the actual gas composition and a more realistic evaporation rate are required and a model for water drainage would benefit the model.