Introduction

With an increasing energy demand and a world running short in fossil energy sources, sustainability and efficiency are nowadays requirements in a modern provision of energy. In this context, hydrogen is often mentioned as an important energy carrier of the future (Ehret and Bonhoff 2015), and in order to that, one of the most promising alternatives to fossil fuels (Saxena et al. 2009). Hydrogen is also frequently associated with its positive effect on global warming and greenhouse gas emissions, which are growing issues on technological as well as societal and political levels (Bulatov and Klemeš 2011). Unfortunately, as Duić (2015) states, the effect of climate change is still not strong enough a reason to facilitate the development of alternatives, because the user still has the benefit from using fossil fuels whereas the resulting damage is a global one. But with an undeniable finiteness of fossil fuels, the need for a secure and renewable energy supply system may be a much stronger incentive to push the development of alternatives like hydrogen (Duić 2015). In a recent study on hydrogen as the future transportation fuel by Sharma and Ghoshal (2015) it is seen to hold the promise to be the fuel of the future, provided that the major technological barriers will be resolved. The majority of these barriers are cost based, whereas the most important barrier relates to the provision of a sustainable production route.

The state of the art in industrial hydrogen production is natural gas steam reforming (Sharma and Ghoshal 2015), which is a fossil fuel-based pathway. With an estimated annual production close to 27 Mt, it accounts for 50 % of the globally produced hydrogen (Kalamaras and Afstathiou 2013). In total, 95 % of global hydrogen productions are fossil fuel-based processes (Balat and Kirtay 2010). Another commonly used hydrogen production method is the water electrolysis, which was the first commercially available technology, dating from the late 1920s (Riis et al. 2006). This reveals the problem with common production methods; they are either based on fossil fuels or the energy-intensive splitting of water. Even though a lot of research is done on developing and improving the technologies for utilization, storage, and distribution of hydrogen, the production methods are still contradictory to the requirements mentioned in the beginning (Dalebrook et al. 2013). Building the hydrogen economy based on these methods the energy problem will not be solved in a suitable way (Bossel 2006).

The vital demand in sustainability can be met via the production of hydrogen from renewable feedstock via biotechnological pathways. In this context, biomass is often mentioned to be an ideal feedstock because it can be utilized in a sustainable way and has a great energy production potential (Uddin and Daud 2014). Chaubey et al. (2013) states in his review that the methods utilized for the hydrogen production from renewable sources like biomass are either biological or photobiological. One of these methods is the conversion of biomass to hydrogen via dark fermentation, which is considered to be the most environmentally friendly alternative in hydrogen production according to Urbaniec and Bakker (2015). In this particular case, biomass residues as well as second-generation biomasses can be fermented to produce a gas mainly containing hydrogen and carbon dioxide. The share of hydrogen in the fermentation gas varies from 40 vol% (Ren et al. 2006) up to 67 vol% (Zhang et al. 2006), with respect to the applied conditions. This assumption is also confirmed in the work of Al-Shorgani et al. (2014), where H2-purities of 58 vol% occur during the fermentation of agroindustrial wastes. As the H2-purity of the fermentation gas is restricted, a subsequent separation of CO2 is inevitable to obtain the required hydrogen purity in the final product. Commonly used gas upgrading methods in this context are amine absorption and pressure swing adsorption (Niesner et al. 2013). The latter is the most widely used method for the industrial purification of gases (Lestinsky et al. 2015). To ensure an efficient overall process the energy requirement of the upgrading step needs to be low. Utilizing gas permeation could therefore have a positive effect on the energy demand of the process compared to the common upgrading methods.

Hence, the aim of this project is to develop an innovative small-scale process to upgrade hydrogen-rich gas to a defined H2 product purity of higher 95 vol% using membrane technology. Simulation will point out which process conditions result into a robust, effective, and economic separation of hydrogen and carbon dioxide during fermentative hydrogen production.

Materials and methods

The project consisted of a simulative and an experimental part, which are interlinked. As the first step, a process simulation tool was used for the design of a single-stage membrane unit operation. The obtained data were then further used for comparison with the commercially available membrane material. This initiated the practical part of the project, with the assembly and screening of membrane modules in a laboratory scale, followed by distinctive testing with two different gas mixtures. As the next step, a small-scale pilot plant was designed and set up to further test the modules on a field scale. This was done by connecting the small-scale pilot plant to a real hydrogen fermentation reactor. The data obtained from field and laboratory testing were used to evaluate and validate the single-stage simulation model. With the validated single-stage model different two-stage separation processes were simulated to investigate the best process separation efficiency and performance. The parameters for efficiency and performance were defined as the specific energy demand, the hydrogen purity, and the hydrogen recovery. The specific energy demand is the energy needed per volume of hydrogen produced.

Experimental approach

The used membrane material was provided by our project partner AXIOM Angewandte Prozesstechnik GmbH. These hollow fiber membranes are on a polyimide basis and characterized as H2-selective (Shao et al. 2009). To confirm this assumption, the provided material was installed into modules which were further tested in the laboratory. The experimental setup used for the tests is described in detail by Makaruk and Harasek (2009).

Pure gases of hydrogen and carbon dioxide, as well as nitrogen as a reference gas, were used for the membrane screening. Furthermore, tests with two different gas mixtures containing H2, CO2, and N2 were executed. The respective fractions of H2/CO2/N2 for the two gas mixtures were: 30/30/40 and 66/34/0 (all vol%). For all membrane screening experiments different pressure and temperature levels were applied. In Table 1 the respective variation parameters are listed. It has to be mentioned that the operational mode during the pure gas tests of the membrane module was a dead-end mode. Using this mode the maximum transmembrane flowing at a certain pressure and temperature was determined. Furthermore, it was possible to calculate the ideal H2/CO2-selectivity at the respective conditions.

Table 1 Applied parameter variation during membrane screening with pure gases
Full size table

In contrast to the pure gas tests, the modules were operated in a counter-current mode during the screening with gas mixtures. However, the same parameter variation for pressure and temperature was applied. These tests gave information about the membrane’s performance under operational conditions, especially the effect on the permeance, and the influence of a changing CO2 content.

The next step in the experimental work was the design and setup of a small-scale pilot plant for the field tests of the membrane modules. Figure 1 pictures the pilot plant and its respective flow sheet. It was designed as a single-stage gas separation unit, which was connected to a hydrogen fermenter. Thereto, gas pre-treatment steps like drying, desulfurization with iron oxide, and adsorption with activated carbon had to be applied. These steps were supposed to prevent the membrane material from destructive residuals which could have been generated in the fermenter.

Fig. 1
figure 1

Flow sheet and picture of the small-scale pilot plant. The apparatus is connected to a hydrogen fermenter during field tests of the H2-selective membrane modules

Full size image

The objective of the field tests was to determine the boundaries for a stable product gas flow and a stable gas quality. By changing the conditions in the membrane separation unit, the membrane module behavior along with the conditions for stable operation were investigated. Furthermore, it should be demonstrated, that the utilization of membranes for the continuous purification of fermenter gas is an attractive option.

The fermenter system utilized for hydrogen production via dark fermentation has been a part of a previous research (Foglia et al. 2011). It was characterized as a 600 l fermenter, operated under thermophilic conditions, and with molasses as substrate. As the purification process was the emphasis of this research, we refer to literature (Azwar et al. 2014; Saratale et al., 2013) for more detailed information about dark fermentation.

Simulative approach

Based on previous simulation work from Rodrigues (2009), Makaruk and Harasek (2009), and Rom et al. (2014) a gas permeation unit operation was developed with the modeling tool Aspen Custom Modeler® (ACM, V7.3, Aspen Technology, Inc., Burlington, USA, 2011). This unit operation was implemented into the Aspen Plus® (V7.3.2, Aspen Technology, Inc., Burlington, USA, 2013) Model Library for further utilization in the process flow sheet.

A certain share of the data obtained from experimental work was utilized for regression and further development of the ACM single-stage model. After implementation into Aspen Plus®, the membrane model was then used to investigate and sharpen the given permeances from the manufacturer. Typical H2-permeances for polyimide membranes are in the range of 80–120 gpu, with a H2/CO2-selectivity of 2.4 (Makaruk et al. 2012). Compared to that, metallic membranes and ceramic membranes have respective H2-permeances of 4700–25,000 and 20–180 gpu (Adhikari and Fernando 2006). Unfortunately, these membranes are not feasible for the considered task, with respect to their operational parameters. The permeances in all subsequent simulations were assumed to be constant.

The next step was to determine the selectivity for a multi component gas mixture from the single-stage simulation results. The remaining data were then used for comparison with the simulation to evaluate the accuracy of the model. For the investigation of a multi-stage process, three different setups were chosen. Figure 2 pictures the respective flow sheets. For all multi-stage simulations only the 66/34/0 gas mixture was considered. The detailed feed gas stream conditions were set as listed in Table 2, in which the reference pressure level is defined as 1 bar.

Table 2 Feed gas conditions defined for the Aspen Plus® multi-stage simulation
Full size table

Design of the two-stage processes

In Setup 1 the feed gas is compressed to 10 bar, and together with the retentate from the second stage (recycle) introduced into the first membrane. The retentate from the first stage is defined as the CO2-rich off gas. The resulting permeate from the first stage must be compressed again to 10 bar before it can be fed into the second membrane. As mentioned before, the retentate is used as a recycle, and the permeate from the second stage represents the H2-rich product gas.

Setup 2 is characterized by the need of only one compression stage. In this case, the feed gas is mixed with the permeate from the second stage (recycle) and compressed to 10 bar. In this setup the permeate from the first stage is defined as the H2-rich product gas. The pressure-prone retentate from the first stage is directly fed to the second membrane. The resulting permeate from this second stage is added as a recycled feed gas and the resulting retentate is the CO2-rich off gas.

Setup 3 consists of the same arrangement as Setup 2, but the utilization of reverse-selective membrane material is assumed. As a result, the H2-rich gas stream remains always on the pressure-bearing retentate side of each stage. The permeate from the first membrane is defined as the CO2-rich off gas and the permeate from the second membrane is used as recycle stream.

The effect of CO2-selective membranes on the process was investigated by implementing the respective permeances taken from literature (Car et al., 2008) into the simulation. The assumed CO2/H2-selectivity for the reverse-selective material is 8.4 with a given permeance of 105 gpu. Table 3 lists the assumptions taken for the simulations of the single-stage membrane unit, as well as for the three different multi-stage setups.

Table 3 Assumptions for the simulation in Aspen Plus®, with relation to single-stage and multi-stage calculations
Full size table

Results

The membrane screening with pure gases results in an ideal H2/CO2-selectivity of 3.3 at a pressure difference of 9 bar and at a temperature of 30 °C. Table 4 shows that, at the same operational conditions, the investigation of mixed gases results into a reduction of H2/CO2-selectivity to 2.5 and 2.2 for the respective 30/30/40 and 66/34/0 compositions.

Table 4 Selectivity calculated with Aspen Plus®, based on the mixed gas laboratory tests; valid for a pressure difference of 9 bar, and at a temperature of 30 °C
Full size table

Furthermore, the resulting selectivites with respect to nitrogen are listed. All values reflect the findings for polyimide membrane material from literature (Shao et al. 2009). The slight H2/CO2-selectivity difference can be explained by the CO2-induced plasticization effect going along with the difference in the CO2 partial pressure of the feed stream. As the permeances applied in the simulation are based on the laboratory and field results, it can be assumed that the plasticization effect is taken into account to a certain extent.

The analysis of the fermenter gas during field tests showed that admixing of nitrogen as sweep gas did not affect the measurements or gas analysis. Furthermore, a slight variation in the hydrogen to carbon dioxide ratio was detected. It is assumed that the 30/30/40 feed gas mixture from the laboratory tests resembles the composition of the fermenter gas.

Figure 3 illustrates the results for the single-stage setup investigated through laboratory work, field tests, and Aspen Plus® simulation. The primary vertical axis represents the hydrogen recovery, and the secondary vertical axis depicts the hydrogen purity in the permeate. All gradients pictured are depending on the stage cut, which is defined as the permeate to feed flow ratio. It can be seen that, the single-stage Aspen Plus® membrane model predicts the measurements for both feed compositions very well. Furthermore a trade-off between H2-purity and H2-recovery is apparent. With increasing stage cut the H2-recovery also increases, but it results in a decrease of H2-purity in the permeate. The other way round, lowering the stage cut generates an increase in H2-purity, but it implies also a reduction of transmembrane flow. Furthermore, the trend of the H2-purity in the permeate approaches a certain maximum with decreasing stage cut. The results for the single-stage setup go along with the findings from Bakonyi et al. (2013), who identifies the gas composition, temperature, and stage cut as important process variables that could affect the hydrogen separation efficiency. It has to be noted that they also tested polyimide membrane material in their work, but at a lower feed pressure.

Figure 3 also shows that the maximal reachable value is strongly dependent on the H2-concentration in the feed stream. For the given 30/30/40 and 66/34/0 mixtures in the feed stream the maxima are at the respective H2-purities around 67 and 82 vol%. The findings from the membrane tests therefore confirm the assumption that a desired H2-purity in the product of at least 95 vol% cannot be reached by the utilization of a single-stage membrane module. This is somewhat typical for membrane processes, as the selectivity is often not high enough for a complete removal of the contaminant (Bakonyi et al. 2013). However, the implementation of a second stage could lead to the necessary improvement in product quality.

In Table 5 the results for the simulative investigated two-stage processes are listed. These results refer to the process setups as pictured in Fig. 2. Setup 1 is not taken into further consideration, because the utilization of H2-selective membrane material in Setup 2 leads to higher values for H2-purity and H2-recovery. In this scenario the compressor energy of 191 kW and a total membrane area of 775 m2 are needed, which results in a specific energy demand 0.4 kWh/Nm3H2. Still, a further purification step will be necessary to reach the desired targets. Compared to that, when using CO2-selective membrane material the membrane area increases to 1310 m2. Furthermore, it alters the H2-recovery while meeting a product H2-purity of 95.5 vol% and reducing the specific energy demand. Unfortunately, the CO2-selective membrane material utilized for simulation is not yet commercially available (Fig. 3).

Table 5 Comparison of the H2- and CO2-selective membranes applied in the respective two-stage simulation (Setup 2 and Setup 3)
Full size table
Fig. 2
figure 2

Flow sheets of the simulated two-stage separation processes; Setups 1 and 2 utilizing H2-selective membranes, Setup 3 utilizing CO2-selective membranes

Full size image
Fig. 3
figure 3

H2-recovery and H2-purity in the product for a single-stage process depending on feed composition and stage cut; data taken from field tests, laboratory tests, and simulation results

Full size image

Conclusions

This work shows that an actual online gas upgrading is possible. Furthermore, that the potential of a stable separation process depends not only on the membrane material but also on the stability of the fermentation. The field tests were successfully executed and confirm the findings from laboratory experiments. However, the commercially available H2-selective membrane material is the limiting factor for the application of a single-stage setup in hydrogen purification. The low H2/CO2-selectivity of the polyimide makes it impossible to reach the targeted H2-purity with the corresponding H2-recovery.

The design of different two-stage processes in Aspen Plus® is used to determine the most effective and economic separation setup. Implementing a second stage into the process increases the H2-recovery compared to a single stage. But still, when using H2-selective membranes the increase in product quality is unsatisfactory and a subsequent purification step will be necessary. Simulations of the two-stage process show that the utilization of CO2-selective membranes would result in a higher purity of the product going along with a decrease in specific energy demand. To ensure an efficient purification process, it is therefore important to improve the H2-selective material with regard to their selectivity, as well as the further development of CO2-selective material.