Hydrogen production using ethylene glycol steam reforming in a micro-reformer: Experimental analysis, multivariate polynomial regression and genetic programming modeling approaches

Highlights

• Ethylene glycol steam reforming in a micro-reformer for hydrogen production.

• High values of EG conversion and H₂ yield for the NPGA bimetallic catalyst.

• High hydrogen productivity at 450°C and WHSV of 80.8 h⁻¹.

• Using of MPR and GP approaches to modeling of the EGSR process.

• GP model as the optimal model with the highest prediction power.

Abstract

Three types of catalysts, i.e. Pt/γ-alumina (PGA), Ni/γ-alumina (NGA) and Ni–Pt/γ-alumina (NPGA), were prepared by incipient wetness impregnation (IWI) and deposited on a micro-reformer for hydrogen production by ethylene glycol steam reforming (EGSR). Multivariate polynomial regression (MPR) and genetic programming (GP) approaches were used to model the EGSR process based on experimental data. In these models, temperature and weight hourly space velocity (WHSV) as independent variables and ethylene glycol (EG) conversion, H₂ selectivity, H₂ yield and CO selectivity were considered as target functions. Based on the results, the GP model predicts objective functions with the highest prediction power and this model was selected as the optimal model. For example, for the NGA catalyst and the dependent variable of EG conversion, the values of correlation coefficient (R²) and root mean squared error (RMSE) were 0.9980 and 1.3191, respectively based on the GP model while for the best MPR (cubic) model; these parameters were 0.9735 and 4.2476, respectively. The results showed that the EG conversion values for the NPGA bimetallic catalyst were higher than for the PGA or NGA monometallic catalysts. The maximum values of EG conversion, H₂ selectivity and H₂ yield for all the catalysts were obtained at a temperature of 450°C and at a WHSV of 80.8 h⁻¹.

Introduction

Nowadays, many researches in energy field have been focused on hydrogen production from biomass-derived sources due to rising demand for clean energy, fossil fuel reserves reduction and increasing greenhouse gas emissions [1,2]. Fuel cells are the efficient devices that generate electricity from hydrogen as a sustainable and environmentally friendly energy source [3,4]. The main processes for hydrogen production are auto-thermal reforming (ATR) [5], steam reforming (SR) [6], sorption-enhanced steam reforming (SESR) [7], aqueous phase reforming (APR) [8], partial oxidation (POX) [9], and supercritical water gasification (SWG) [10]. Recently, attentions focused on biomass derived oxygenated source for hydrogen generation due to their carbon neutral, renewable and accessible properties [11]. Ethylene glycol (EG) is the most plentiful product of catalytic conversion of cellulose (the main component of biomass) which can be considered as a non-volatile hydrogen source through two main approaches; SR [12] and APR [13] processes. The APR reaction occurs at low temperatures (150°C to 270°C) and pressures in the range of 1500 to 5000 kPa. However, the APR reaction rate and H₂ selectivity are low and in order to inhibit catalyst deactivation, the EG concentration in the feedstock should be low. Compared to the APR, the SR process takes place at higher temperatures and at atmospheric pressure which high energy cost of the SR reaction can be balanced by the use of inexpensive catalysts and considering higher H₂ selectivity [14]. In ethylene glycol steam reforming (EGSR) process, several chemical reactions should be considered, which include the overall SR (Eq. (1)), EG decomposition (Eq. (2)), methanation (Eq. (3)) and water gas shift (WGS) (Eq. (4)) reactions, as follows [15]:

$${\mathrm{C}}_2{\mathrm{H}}_6{\mathrm{O}}_2+2{\mathrm{H}}_2\mathrm{O}\leftrightarrow 2\mathrm{C}{\mathrm{O}}_2+5{\mathrm{H}}_2\Delta \mathrm{H}{°}_{298}=+91\text{kJ}\text{mo}{\mathrm{l}}^{-1}$$

$${\mathrm{C}}_2{\mathrm{H}}_6{\mathrm{O}}_2\leftrightarrow 3{\mathrm{H}}_2+2\text{CO}\Delta \mathrm{H}{}_{298}^{\circ }=+173\text{kJ}\text{mo}{\mathrm{l}}^{-1}$$

$$3{\mathrm{H}}_2+\text{CO}\leftrightarrow \mathrm{C}{\mathrm{H}}_4+{\mathrm{H}}_2\mathrm{O}\Delta \mathrm{H}{}_{298}^{\circ }=-206\text{kJ}\text{mo}{\mathrm{l}}^{-1}$$

$$\text{CO}+{\mathrm{H}}_2\mathrm{O}\leftrightarrow \mathrm{C}{\mathrm{O}}_2+{\mathrm{H}}_2\Delta \mathrm{H}{}_{298}^{\circ }=-41\text{kJ}\text{mo}{\mathrm{l}}^{-1}$$

The elements in group VIII e.g. rhodium [16], ruthenium [17], platinum [18], and palladium [19] have been introduced as promising catalyst active phases for hydrogen generation from hydrocarbon reforming processes. However, noble metallic particles have high selectivity towards hydrogen but their cost is high. Another alternative for reforming reaction are nickel based catalysts due to their low price, high availability and excellent activity for carbon-carbon bonds [20,21]. As a result, the use of Ni-noble metal bimetallic catalysts in various reactions including reforming, hydrogenation and dehydrogenation has been the subject of many studies compared to single-metal catalyst [20] due to their unique properties such as excellent resistance to coke formation and high catalytic activity (H₂ selectivity and feed conversion) [22,23].

In most previous research studies in the EGSR field, the process was performed in a conventional macro scale reactor (fixed bed reactor) with mass and heat transfer limitations which are critical issues for fuel cell applications. Micro-reformer (including sub-millimeter channels) deposited by the catalyst is an attractive light-weight device for portable fuel processing as hydrogen source [24,25]. Due to the endothermicity of steam reforming process, accurate monitoring and control of the reaction temperature in these compact fuel processors resulted in reduction of hot spot regions and inhibition of catalyst deactivation [26]. Izquierdo et al. [27] performed the EGSR process over Rh-based catalysts loaded on La₂O₃ and CeO₂ modified α-Al₂O₃ in a micro-reactor.

Mathematical modeling is a useful tool for the design and control of chemical processes, in particular for steam reforming in a micro-reformer [28,29]. It should be noted that, very few studies have been performed on the modeling of the EG reforming process, and all of them are limited to kinetic and micro-kinetic modeling over Pt catalyst [30]. In recent years, considerable attention has shifted to modeling of chemical processes with multivariate polynomial regression (MPR) and artificial intelligence methods including artificial neural networks (ANNs), genetic algorithm (GA) and genetic programming (GP) approaches [31], [32], [33], [34]. These models (MPR, ANN and GP) do not use any transport equation and only require experimental input (independent variables) and output (dependent variables) data of process. Omata et al. [35] used an experimental design procedure coupled with ANNs to investigate the influence of preparation parameters of Co–MgO catalysts in methane dry reforming process. Ayodele and Cheng [36] applied the Box–Behnken design and ANNs together to model and optimize the input variables (CH₄ partial pressure, temperature of reaction and reactant feed ratio) to achieve desired conversion of CH₄ and CO₂ and syngas ratio. Hossain et al. [37]. used ANNs for modeling of H₂ and CO yields as well as CH₄ and CO₂ conversions for methane dry reforming over Ni/CaFe₂O₄ catalysts. Azarhoosh et al. [5,33,38] used GA for optimization and simulation of a horizontal ammonia synthesis reactor and hydrogen generation by low-pressure ATR of natural gas. Optimization of hydrogen production from toluene steam reforming by ANN coupled GA (ANN-GA) and response surface methodology (RSM) models studied by Yahya et al.[39]. Recently, Lotfi et al. [40] applied 2-dimentinal heterogeneous model and a modified ANN and non-dominated sorting genetic algorithm-II (NSGA-II) to maximize methane conversion in the industrial combined dry and steam reforming (CDSR) process.

Notably, the GP model has not been used for the reforming process, so far. For other processes, Azarhoosh et al. [41] used GA and GP approaches for presenting a new kinetic model for methanol-to-light olefins (MTO) process in the presence of the hierarchical SAPO-34 catalyst. A comparison between the experimental and predicted results of the new kinetic model showed that the proposed model has good accuracy. In another work, they modeled the performance of hierarchical SAPO-34 catalyst in the MTO process using multiple linear regression (MLR) and two intelligent methods, i.e., GP and ANN [34]. According to their finding, the models achieved using the GP method had the highest accuracy for training and test data compared to other models.

In previous work, we studied the EGSR process over Ni, Pt and NiPt catalysts loaded on γ-alumina support in a micro-reformer, experimentally. The γ-alumina was chosen as the catalyst support because of its perfect properties such as good thermal stability, high specific surface area, adjustable pore structure and good adsorption performance [31,42,43].

To the best of our knowledge, the MPR and GP approaches have not yet been used to model the efficiency of a micro-reformer, in particular for the EGSR reaction. Therefore, in this research work, the main objective is the development of MPR (linear, quadratic and cubic) and GP models which can predict the behavior of EGSR process in a micro-reformer for hydrogen production in the presence of three types catalysts (Pt/γ-alumina (PGA), Ni/γ-alumina (NGA) and Ni-Pt/γ-alumina (NPGA)), followed by validation with experimental data. Finally, by selecting the optimal model from the MPR or GP models, a model was used to study the effects of operating parameters (temperature and WHSV) on the target functions i.e. EG conversion, H₂ selectivity, H₂ yield and CO selectivity with the best accuracy and prediction power.