Transesterification of canola, palm, peanut, soybean and sunflower oil with methanol, ethanol, isopropanol, butanol and tert-butanol to biodiesel: Modelling of chemical equilibrium, reaction kinetics and mass transfer based on fatty acid composition

Applied Energy 123 (2014) 108–120

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

Transesterification of canola, palm, peanut, soybean and sunflower oil with methanol, ethanol, isopropanol, butanol and tert-butanol to biodiesel: Modelling of chemical equilibrium, reaction kinetics and mass transfer based on fatty acid composition Blazˇ Likozar ⇑, Janez Levec a b

Laboratory of Catalysis and Chemical Reaction Engineering, National Institute of Chemistry, Hajdrihova 19, 1000 Ljubljana, Slovenia Faculty of Chemistry and Chemical Technology, University Ljubljana, Aškercˇeva 5, 1000 Ljubljana, Slovenia

h i g h l i g h t s

g r a p h i c a l a b s t r a c t

 Catalysed transesterification to

Modelling of chemical equilibrium, reaction kinetics and mass transfer for triglyceride transesterification with different alcohols based on fatty acid composition.

biodiesel with various oils, alcohols and catalysts.  Analysis of components and reactivity based on fatty acid composition of all species.  Simultaneous modelling of mass transfer, reaction kinetics and chemical equilibrium.  Diffusivities, distribution and mass transfer coefficients for individual components.  Correlation of kinetic parameters with molecular structure of reactants and products.

a r t i c l e

i n f o

Article history: Received 8 October 2013 Received in revised form 10 February 2014 Accepted 11 February 2014

Keywords: Biodiesel production optimization Chemical kinetics Diffusion Mathematical model Waste oil Renewable energy and fuels

a b s t r a c t Mechanism of alcoholysis (e.g. methanolysis) using different oils, alcohols and homogeneous base catalysts was utilized to devise chemical kinetics and thermodynamics based on fatty acid composition, differentiating among triglycerides, diglycerides, monoglycerides and fatty acid alkyl esters (e.g. fatty acid alkyl esters, FAME) with bonded gadoleic, linoleic, linolenic, oleic, palmitic and stearic acid-originating substituents. Their concentrations were measured using an optimized high-performance liquid chromatography (HPLC) method. Hydrodynamics and diffusion limitations in emulsion were considered in overall model by determining diffusivities, distribution coefficients, molar volumes, boiling points and viscosities of individual components. Pre-exponential factors and activation energies were related with structure of reactants, intermediates and products acknowledging number of carbons, double bonds and alkyl branches by linear and mixed response surface methodology. Developed model may be used with batch and continuous flow reactors, e.g. for novel micro-structured or industrial-scale process intensification, different vegetable or non-edible oils (waste cooking Jatropha or microalgae lipids). Ó 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Laboratory of Catalysis and Chemical Reaction Engineering, National Institute of Chemistry, Hajdrihova 19, 1000 Ljubljana, Slovenia. Tel.: +386 1 4760283; fax: +386 1 4760300. E-mail address: [email protected] (B. Likozar). http://dx.doi.org/10.1016/j.apenergy.2014.02.046 0306-2619/Ó 2014 Elsevier Ltd. All rights reserved.

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

1. Introduction The transesterification reaction of different oils and alcohols has gained much attention recently because of its use in biodiesel industry. The continuous production of algal, palm, peanut, canola, soybean, sunflower, and other oil-originating biodiesel with different alcohols in various reactor configurations demands a process model, acknowledging chemical equilibrium, reaction kinetics and mass transfer, regardless of fatty acid composition of oil. Most of the studies investigate the biodiesel production process of a given oil and alcohol (most frequently methanol) [1–6], nonetheless; a variation in oil resource compound (e.g. waste oil mixtures), in the fatty acid composition of a single oil type (e.g. canola), or even in the alcohol, utilized for the transesterification (e.g. the partial substitution of methanol with bioethanol) may jeopardize process monitoring, regulation, optimization, or intensification. Mass transfer is often not accounted for in the process model, although it plays an important role in batch reactors, e.g. at low temperatures and poor mixing, and a predominant one in continuous reactors. Some studies mention the process mechanism consisting of an initial mass transfer-controlled region followed by a kineticscontrolled region for the transesterification of palm [7], canola [8,9], soybean [10], and sunflower [9,11] oils with methanol, however; mass transfer is seldom incorporated into the overall process model [7,8,10,11]. When other alcohols are applied in biodiesel synthesis, the literature considering mass transfer resistances between alcohol and oil phase is even more scarce. Fluid mechanics of ethanol/oil emulsions were investigated by Duangsuwan et al. [12], nonetheless; kinetics and mass transfer during the transesterification were not studied. Mass transfer is at least qualitatively treated for the supercritical [13], acidic [14], heterogeneously-catalysed [15,16], and enzymatic [17,18] transesterification with ethanol [13,14,17], isopropanol [14,15, 18], butanol [14,16], and 2-butanol [18], however; is not directly coupled with reaction kinetics into the overall process model. Analogously may be concluded for the conventional base-catalysed homogenous transesterification with ethanol [19,20], isopropanol [20], and butanol [20]. The latter type of catalysis is the most widespread in biodiesel production process to date and should be suitably described. Reaction kinetics were extensively studied for the transesterification of different oils with methanol, but the main disadvantage remains that the utilization of the determined kinetic parameters often remains limited to the investigated oil, as the fatty acid composition usually varies even for one oil type (e.g. canola) [9–11]. In the case of the transesterification with other alcohols (e.g. ethanol), the kinetics is often only studied in a simplified manner [17,19,20]. The aim of this study is to present a model for the transesterification of different oils and alcohols, based on fatty acid composition of tri-(TG), di-(DG) and monoglycerides (MG), and alkyl esters (AE) (biodiesel), acknowledging chemical equilibrium, reaction kinetics and mass transfer.

109

660:2009, ISO 3657:2008, and ISO 3961:2000 official methods. For the transesterification, a certified methanol of 99.8 wt.% and 99.9 wt.% purity was purchased from Sigma–Aldrich (Germany) and Merck (Germany), respectively, ethanol of 99.5 wt.% purity from Sigma–Aldrich (Germany), isopropanol of 99.9 wt.% purity from Sigma–Aldrich (Germany), butanol of 99.5 wt.% purity from Merck (Germany), and tert-butanol of 99 wt.% purity from Merck (Germany). KOH and NaOH pellets of 88 wt.% and 99 wt.% purity were purchased from J.T. Baker (Holland) and Merck (Germany), respectively. Solvents, specifically, acetonitrile (gradient grade; 99.9 wt.%; hypergrade for LC–MS; 99.9 wt.%), methanol (gradient grade; 99.9 wt.%; for liquid chromatography; 99.8 wt.%), n-hexane (for high-performance liquid chromatography (HPLC); 97 wt.%; for liquid chromatography; 98 wt.%), and isopropanol (for HPLC; 99.9 wt.%; gradient grade for liquid chromatography; 99.9 wt.%), all of HPLC grade (Chromasolv; LiChrosolv) and used without purification, were obtained from Sigma–Aldrich (Germany) and Merck (Darmstadt, Germany). The HPLC reference standards for fatty acid methyl, ethyl, isopropyl, butyl, and tert-butyl esters (FAME, FAEE, FAiPE, FABE, and FAtBE) containing methyl, ethyl, isopropyl, butyl, and tert-butyl esters of gadoleic (G), linoleic (L), linolenic (Ln), myristic (M), oleic (O), palmitic (P) and stearic (S) acids (not all combinations of esters), and corresponding tri-(trilinolein, trilinolenin, triolein, tripalmitin, and tristearin), di-(1,2-dilinolein, 1,3dilinolein, 1,2-dilinolenin, 1,3-dilinolenin, 1,2-diolein, 1,3-diolein, 1,2-dipalmitin, 1,3-dipalmitin, 1,2-distearin, and 1,3-distearin) and monoglycerides (1-monolinolein, 2-monolinolein, 1-monolinolenin, 2-monolinolenin, 1-monoolein, 2-monoolein, 1-monopalmitin, 2-monopalmitin, 1-monostearin, and 2-monostearin) were purchased from Sigma–Aldrich (Germany) and Nu-Chek Prep (USA). 2.2. Batch reactor The reactions were carried out in a 0.6 L glass reactor equipped with the Rushton-type turbine (Figs. SD.1 and SD.2, Supplementary Data). The impeller diameter and blade width were 25 and 6 mm, respectively. The impeller was centrally placed at 50 mm from the bottom. The reactor was equipped with glassy double jacket filled with silicone oil circulating from a thermostat bath by means of a pump. The reactor was filled with 272 mL of emulsion (the emulsion height was 75 mm). 2.3. Process conditions The 6:1 M ratio of alcohol to vegetable oil was used in all experiments. KOH and NaOH (0.8 g per 100 g of oil) were dissolved into alcohol before use. The experiments were carried out at 40, 50 or 60 °C, and atmospheric pressure. The impeller speed of 400 rpm (power input per unit volume was 1.5 W/m3 and 295.3 W/m3 for the canola oil mixing at 50 °C using 100 rpm or 600 rpm, respectively) was applied to produce a uniform dispersion of alcohol into oil.

2. Material and methods

2.4. Process procedure

2.1. Materials

The transesterification reactions were performed with canola, palm, peanut, soybean and sunflower oil, and methanol, ethanol, isopropanol, butanol and tert-butanol in the proportions of 1:6 (mol/mol) using KOH and NaOH for approximately 30 min (canola, peanut, soybean, and sunflower oil) or 40 min (palm oil) to obtain a mixture of alkyl esters, glycerol, diglycerides, monoglycerides, and unreacted oil and alcohol at the temperatures of 40, 50, and 60 °C. The reactor was initially charged with 158–206 g (depending on materials and process conditions) of oil, placed in the reactor and heated to the desired temperature, which was then maintained

Commercial refined and edible-grade canola (Slovenia), palm (Slovenia), peanut (Italy), soybean (Bosnia and Herzegovina), and sunflower (Bosnia and Herzegovina) oils were used. The acid, saponification and iodine values of the oils were 0.3 wt.%, 0.6 wt.%, 0.6 wt.%, 0.8 wt.% and 0.7 wt.% (acid), 177, 220, 137, 190 and 193 mg KOH/g (saponification), and 116, 51, 95, 128 and 135 g I2/100 g (iodine) for canola, palm, peanut, soybean, and sunflower oil, respectively, determined according to the ISO

110

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

at a constant value. Alcohol (43–82 g; depending on materials and process conditions) with dissolved potassium or sodium hydroxide (1.5–1.8 g; depending on materials and process conditions), which was heated separately, was added to the reactor. The mechanical stirrer was turned on during oil heating and as soon as alcohol was added to oil, the reactions were timed. For studying the equilibrium, kinetics, and mass transfer, the samples (2 mL) were removed from the reaction mixture during the progress of the reactions, immediately quenched by adding an aqueous hydrochloric acid solution (1.0:1.7 (w/w); 0.02 mL) and vigorously shaken (manually for 1 min).

image processing and analysis (ImageJ, Bethesda, MA, USA). From 25 to 50 (I) droplets were digitally measured in photographs with the progress of the alcoholysis reactions for each operational condition. The Sauter mean diameter of droplets, d32, was calculated using Eq. (2) (di is individual droplet diameter).

2.5. Analytical methods

Biodiesel production process occurs for all mentioned alcohols except for methanol within a single phase (pseudo-homogeneous regime). In this case the regime is determined solely by reaction kinetics and may be represented by the kinetic scheme of Eq. (3), with TG, DG, MG, G, A, AE, and OH being tri-, di- and monoglyceride, glycerol, alcohol, alkyl ester, and hydroxide ion (catalyst), respectively. In Eq. (3), k1k6 represent reaction rate constants.

2.5.1. Determination of concentration of reactants, intermediates and products The samples, removed from reaction mixtures, formed two layers, which were blended by shaking, and a part of the samples was withdrawn, dissolved in isopropanol/n-hexane (5:8 (v/v)) in an appropriate ratio, always obtaining the sample-to-solvent ratio of 1:30 (w/w). This procedure was used in accordance with the rule that the sample must be diluted in the solvent used for elution. The resulting mixture was used to prepare the samples for the tests that is, for HPLC analysis. The composition of the samples of the reaction mixture was determined by HPLC, as described elsewhere [21], using the optimal method, obtained by statistical analysis, setting the parameters each time according to the ones, which were calculated by the full factorial design method [21] (Tables SD.1, SD.2 and Fig. SD.3, Supplementary Data). The analyses were conducted with an Agilent (Hewlett Packard) (Santa Clara, CA, USA) 1100 Series HPLC equipped with two G1312A solvent delivery units for binary gradient elution, a model G1315A UV–Vis detector, an automatic G1313A sample injector, a model G1316A columns oven for precision temperature control above ambient temperatures, and a ChemStation for LC 3D software. The constant parameters for the analyses were as follows; the temperature of the thermostatically controlled compartment of the column set at 30 °C to avoid alcohol evaporation problems, and the wavelength of the UV detector set at 210 nm to ensure the transparency of the mixtures used as solvents without exceeding 220 nm, a value nearing the wavelength limit for the analysis of FAME FAEE, FAiPE, FABE, FAtBE, and glycerides. In fact, the UV spectrum of a mixture of FAME declines beyond 220 nm [21]. A Phenomenex (Torrance, CA, USA) Synergi™ 4l Hydro-RP 80 Å analytical column, with the internal diameter of 4.6 mm, length of 25 cm, and particle size of 4 lm was used for the analyses. For the analysis of the standards in liquid and solid phase and the canola oil of known fatty acid composition, the solutions of 1:30 (w/w) were prepared in isopropanol/n-hexane (5:8 (v/v)), then diluted to obtain the solutions of 1:36 (w/w), 1:45 (w/w), 1:60 (w/w), 1:90 (w/w), and 1:180 (w/w). The higher dilution ratio of the standards for the samples used in the analyses was justified by the need to identify and quantify the main compounds involved in transesterification. The molar fraction of TG, DG, and MG (XCOMPONENT) was calculated from the concentration of TG, DG, and MG in the fatty acid alcohol ester/ oil fraction of the reaction mixture, cCOMPONENT (mol m3), by the following equation, in which cTG0, cDG0, cMG0, and cAE0 stand for the initial TG, DG, MG, and AE concentrations (mol m3), respectively.

X COMPONENT ¼

cCOMPONENT cTG0 þ cDG0 þ cMG0 þ cAE0 =3

ð1Þ

2.5.2. Determination of dispersed phase droplet size The technique used for droplet size measurement was digital image acquisition (Nikon D3100, Tokyo, Japan) and the procedure,

d32

I X 3 ¼ di i¼1

, I X

2

di

ð2Þ

i¼1

3. Theory

k1

TG þ A DG þ AE k2

k3

DG þ A MG þ AE k4

k5

MG þ A G þ AE k6

ð3Þ

The differential mass balances of the above mentioned components for a single phase are as follows, with cx being the bulk concentration of component x (mol m3) and t time (s).

dcTG ¼ k1 cOH cTG cA þ k2 cOH cDG cAE dt

ð4Þ

dcDG ¼ þk1 cOH cTG cA  k2 cOH cDG cAE  k3 cOH cDG cA dt þ k4 cOH cAG cAE

ð5Þ

dcMG ¼ þk3 cOH cDG cA  k4 cOH cMG cAE  k5 cOH cMG cA dt þ k6 cOH cG cAE

ð6Þ

dcG ¼ þk5 cOH cMG cA  k6 cOH cG cAE dt

ð7Þ

dcA ¼ k1 cOH cTG cA þ k2 cOH cDG cAE  k3 cOH cDG cA dt þ k4 cOH cMG cAE  k5 cOH cMG cA þ k6 cOH cG cAE

ð8Þ

dcAE ¼ þk1 cOH cTG cA  k2 cOH cDG cAE þ k3 cOH cDG cA dt  k4 cOH cMG cAE þ k5 cOH cMG cA  k6 cOH cG cAE

ð9Þ

In the case of methanol utilization, two phases exist initially (Fig. 1; cx,y and cx,y,i denote the bulk and interface concentration of component x in phase y (mol m3), kc,x,y the mass transfer coefficient of component x in phase y (m s1), Dx,alcohol/oil the distribution coefficient of component x between alcohol and oil phase (/), and jm,x the molar flux of component x (mol s1 m2)), thus the differential mass balances of components have to be extended for both of them, including mass transfer resistance terms (Eqs. (SD.1–12), Supplementary Data). The balances consequently simplify to Eqs.

111

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

Fig. 1. Physical models describing the reaction medium; either continuous phase is nonpolar oil phase and dispersed phase is polar alcohol phase (oil and methanol) or homogeneous nonpolar oil/alcohol phase is formed (oil and ethanol, isopropanol, butanol, and tert-butanol) (50 °C, 400 rpm, oil/alcohol ratio of 1:6 (mol/mol), and 0.8 wt.% catalyst per oil weight).

(3)–(9), when the transition to the pseudo-homogeneous regime occurs (reaction kinetics-determining region without mass transfer limitations). 4. Results and discussion 4.1. Mixing, phase separation and mass transfer in biodiesel production process As shown in Fig. 1, initial phase separation occurs only when methanol is used, as far as alcohol as reactant is concerned; ethanol, isopropanol, butanol and tert-butanol are less polar, and hence formed a homogeneous system to begin with, which was also noted in other studies [19,20]. While catalyst did not drastically influence the distribution of the dispersed phase within the continuous upon the utilization of methanol, an interesting observation was that homogeneous solution was initially present solely when palm oil was applied. All other oils formed the biphasic system with methanol. This was not insomuch related to oil type itself, but to a higher conversion of triglycerides being the result of a lar-

ger initial content of diglycerides (Fig. SD.4, Supplementary Data). Diglycerides were in addition to MG and AE responsible for the transition to the pseudo-homogeneous regime (Fig. 1) through the emulsification of the dispersed phase for other oil types as well, which may be seen at maximally 3 min for the lowest DG content (sunflower oil). This further promoted the description of mass transfer mechanisms based on the fatty acid composition of species, deriving overall mass transfer coefficient, Kc,x,m (Eqs. (SD.1– 12), Supplementary Data).

K c;x;m ¼ ð1=kc;x;m þ Dx;m=o =kc;x;o Þ

kc;x;m

d32 ¼ ln 6t

1

1 6 X 1 4p2 n2 De t exp 2 p2 n¼1 n2 d32

kc;x;o ¼ 0:725Re0:43 Sc0:58 V t ð1  /m Þ

ð10Þ !! ð11Þ

ð12Þ

Kc,x,m composes of the mass transfer coefficients of component x in methanol (kc,x,m) and oil (kc,x,o) and its distribution coefficient between phases (Dx,m/o). The latter lies within 103.60 (SSS) and

112

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

Fig. 2. Density and viscosity of used canola, palm, peanut, soybean and sunflower oil, and distribution coefficients of components between methanol and canola, palm, peanut, soybean and sunflower oil versus temperature, utilized for modelling of alkali-catalysed biodiesel production in batch process.

101.78 (glycerol) (Fig. 2), and the mass transfer resistance in both oil and methanol has to be accounted for and is described using Eqs. (11) and (12) [23,24]. The diffusivities in the concentrated mixtures of x in y (Dxy) were calculated from the ones, used for the dilute mixtures, according to Siddiqi and Lucas [25] (Supplementary Data). Re, Sc, Vt, and m represent the Reynolds and Schmidt numbers for drops (defined through Eqs. (SD.17) and (SD.20), Supplementary Data), the terminal velocity of methanol phase drops, and the volume fraction of methanol phase, in this order. The effective diffusion coefficient within methanol drops (De, Eq. (11)) was calculated from Dxm, corrected for the interface adsorption of hydroxyl ions (Eqs. (SD.15SD.19), Supplementary Data). All the details for the calculation of density, viscosity and distribution coefficients (Fig. 2) as well as other properties, needed for the inclusion of the mass transfer into the process model, from the fatty acid composition of species according to the literature [26– 28] may be found in Supplementary Data. The transition from the emulsion of alcohol in oil, where both mass transfer and reaction kinetics determine the rate of transesterification (modelled using Eqs. (SD.1–20) and (10)–(12), and the properties in Fig. 2), to pseudo-homogeneous regime (modelled using Eqs. (3)–(9)), where reaction kinetics present the sole determining step, is schematically represented in Fig. 1, and is reflected through the rates of the depletion and evolution of individual and overall components in Figs. 3 and 4, SD.4 and SD.5 (Supplementary Data). As far as oil is concerned, only palm oil (Figs. 3 and 4) did not

exhibit heterogeneous stage. The reason for this behaviour may not be straightforwardly deduced from the FAME conversion in Fig. 3, but is somewhat more evident from Fig. SD.4. The width of pseudo-homogeneous regime may thus not be directly related to oil type, but predominantly to originating TG conversion, as initially-present DG emulsify the system and usher the transition to homogeneous regime. This behaviour may be verified, with the only exception of sunflower oil at 50 °C, where evidently oil properties as the other influencing factor minutely shorten initial regime; nonetheless, do not distort the correlation between original TG conversion and its analogue at 1 min, observed in Fig. SD.5. These findings reflect the imperative to relate the influences of the mixing, phase separation and mass transfer in biodiesel production process with fatty acid composition distribution, and not with individual oil types, as the latter may vary in composition, e.g. a canola oil may initially contain more DG, and hence exhibit a narrower heterogeneous stage, than the palm oil, used in this study. As far as base catalyst is concerned, the selection of either KOH or NaOH did not largely affect the width of initial regime. From the point of view of reactants and catalyst, the choice of oil predominantly affects the breadth of mass transfer-determining step (Figs. 4 and SD.4); nonetheless, this can hardly be compared to the effect of the rotational speed of impeller, and since the increase of the latter is from process costs perspective negligible in comparison to resource price [29], the utilization of intensive mechanical

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

113

Fig. 3. Mass transfer/kinetics/equilibrium modelling curves and experimental points for composition of reaction mixture during transesterification (methanolysis) of palm oil at temperature 50 °C, phase fraction 1:6 (mol/mol) (oil to methanol), catalyst concentration 0.8 wt.% (per palm oil weight) and impeller speed 400 rpm. Different overall and individual predominant components (a), triglycerides (b), diglycerides (c), and monoglycerides and methyl esters (d). See Section 2. Material and methods for abbreviations.

mixing or ultrasound irradiation [30,31] is vital to decrease the duration of transport phenomena-affected region, even though some extent of variation among different oils would still exist. When replacing a homogeneous catalyst with a heterogeneous one [32,33], an additional mass transfer limitation arises in the form of the liquid film around catalyst particles, and while the latter may also be partially decreased by mixing (film thickness), the increase of catalyst surface presents itself as preferable option. 4.2. Oil feedstock in biodiesel production process The model acknowledging chemical equilibrium, reaction kinetics and mass transfer (Eqs. (10)–(12) and Eqs. (SD.1–20) (Supplementary Data) for the emulsion of alcohol in oil, and Eqs. (4)–(9) for the pseudo-homogeneous solution in Fig. 1) was fitted to measurements, in which different oils, alcohols, catalysts and experimental conditions were applied. Specific measurements were performed even in a wider range of conditions than described in Materials and Methods section in order to grant a good model validity and general applicability, that is within 30–70 °C, 100– 600 rpm, 1:3–1:8 (mol/mol) (oil to methanol), and 0.2–1.2 wt.% (catalyst per oil). Mass transfer characteristics were assessed as described in previous section (selected properties are presented in Fig. 2), while activation energies and pre-exponential factors were determined using nonlinear regression (Levenberg–Marquardt algorithm, 105 tolerance) and the model solved by the Runge– Kutta method with the time step of 0.001 min. Parameter determination was performed simultaneously for all experiments employing the same alcohol, and regression was exe-

cuted several times using different initial approximations of parameters, obtained from the literature [9–11], to reach the global minimum of objective function. Initial approximations were randomly redistributed prior to each regression run and no constraints were applied during their determination procedure. The values of activation energies and pre-exponential factors are presented in Tables 1, 2 and Figs. SD.3, SD.6 (Supplementary Data), while the correlation of these values with the structural characteristics of oil-originating components and alcohols using response surface methodology are depicted in Figs. 5 and 6 (linear/mixed model), and Fig. SD.7 (Supplementary Data) (linear model). Measured data and model predictions for all process stages (Fig. 1) are presented in Figs. 3, 4, SD.4, and SD.5 (Supplementary Data) and Table 3. An excellent agreement was achieved for different oil feedstock and process conditions (Figs. 4 and SD.4) as well as for all oil-originating components (the example for the palm oil in Fig. 3) using the same set of kinetic parameters, based on the fatty acid composition of individual species. Mass transfer-controlled stage is primarily influenced by initial diglyceride content (Fig. SD.4), while overall FAME conversion rate is the highest for soybean oil (Fig. 4), which may be explained by the largest concentration of Ln in this oil in addition to the largest A5 and lowest Ea5 for Ln ? MLn reaction (Table SD.3 and Fig. SD.6), respectively. Even though reaction system is much more complex (reversible, parallel and consecutive reactions), this gives a good illustration of the additional advantages of the model, as the latter observation could not have been deduced from apparent kinetics, treating TG, DG, MG and AE as overall concentrations. The result of this treatment

114

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

Fig. 4. Effect of oil type at 40 °C (a) and 50 °C (b), alcohol type (canola oil) (c), catalyst type and temperature (canola oil) (d), and reaction time on dependence of overall conversion to alkyl esters (FAME, FAEE, FAiPE, FABE and FAtBE) for base-catalysed transesterification reactions of different oils with different alcohols using mechanical mixing at 50 °C, impeller speed of 400 rpm, molar ratio of 1:6 (mol/mol) (oil to alcohol) and catalyst concentration of 0.8 wt.% (per specific oil weight) unless specified otherwise.

is the mismatch in determined kinetic parameters, these being dependent on resources and conditions employed [9–11]. As far as chemical equilibrium is concerned, Table 3 reveals that the highest final conversion is achieved in the case of palm oil, regardless of synthesis temperature. This may not be straightforwardly anticipated taking only Fig. 4 into consideration, as the conversion to AE using palm oil is in fact the lowest at 60 min among investigated oils; nonetheless, Fig. SD.4 gives an insight that the highest final conversion to AE is related to the largest extent of TG transesterification. Even though the reaction system is complex, as mentioned, a simplified explanation of observed behaviour may be as follows. If the initial reaction kinetics of overall AE conversion are the highest for soybean oil due to highest Ln content, as mentioned before, the highest AE conversion is achieved with the studied palm oil, owing to the largest content of S (4.71 mol.%) in comparison to other oils. When approaching equilibrium, the effect of third transesterification backward reaction becomes more pronounced, and if k5/k6 is 4.03 and 4.78 for O, it is 157.84 and 167.62 for S, at 40 and 50 °C. Figs. 5 and 6 (linear/mixed models) and SD.7 (linear models) illustrate the correlation between the kinetic parameters and structural characteristics of oil-originating components. The linear models still represent a satisfactory prediction of pre-exponential factors, even though a better agreement is obtained upon the utilization of linear/mixed models. These may be readily used for the prediction of the kinetic parameters with known component structures.

Developed model may thus be used for the transesterification process description for a wide range of the conditions by knowing only kinetic parameters (Tables 1 and 2) and the initial composition of a given vegetable oil, e.g. different sunflower oils [30,31], the latter being the most abundant in Europe. The advantages of the model become even more pronounced when dealing with vegetable oil mixtures [31], or even waste cooking oils [32], as the latter possess an utterly dissimilar composition and properties than any individual oil they comprise of. With the advent of the preference of non-edible (algal, Jatropha Calophyllum, etc.) oils [33–35] over conventional vegetable alternatives, the correlation of new glyceride and alcohol molecule structures with reaction kinetics is also needed for rather different fatty acid compositions, which may be straightforwardly achieved by methodology, outlined above, even for fatty acids, not taken into account in this study (erucic, lauric, etc.). 4.3. Alcohol in biodiesel production process and effect of its type on biodiesel synthesis and yield If oil feedstock primarily affects transesterification through the fatty acid composition of individual species, alcohol influences the mass transfer rate-determining region through its physical properties and the reaction rate-determining region depending on its chemical properties and reactivity. As illustrated in Fig. 1, all investigated alcohols save for methanol are non-polar enough

115

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

Table 1 Pre-exponential factors of reaction rate constants for catalysed transesterification reactions of TG, DG and MG with ethanol, regressed for different temperatures, mixing intensities, phase fractions, catalyst contents, oils and catalysts. Pre-exponential factor, A1 (107 m6/(kmol2 min))

Pre-exponential factor, A2 (106 m6/(kmol2 min))

Pre-exponential factor, A3 (108 m6/(kmol2 min))

OOO ? OO OOL ? OL OOL ? OO OOLn ? OLn OOLn ? OO OOP ? OP OOP ? OO OOG ? OG OOG ? OO OLL ? LL OLL ? OL OLLn ? LLn OLLn ? OLn OLLn ? OL OLP ? LP OLP ? OP OLP ? OL OLS ? LS OLS ? OS OLS ? OL OLnLn ? LnLn OLnLn ? OLn LLL ? LL

OO ? OOO OL ? OOL OO ? OOL OLn ? OOLn OO ? OOLn OP ? OOP OO ? OOP OG ? OOG OO ? OOG LL ? OLL OL ? OLL LLn ? OLLn OLn ? OLLn OL ? OLLn LP ? OLP OP ? OLP OL ? OLP LS ? OLS OS ? OLS OL ? OLS LnLn ? OLnLn OLn ? OLnLn LL ? LLL

0.73 ± 0.03 0.91 ± 0.01 0.86 ± 0.03 1.07 ± 0.03 0.9 ± 0.1 0.481 ± 0.005 0.58 ± 0.04 0.66 ± 0.06 0.736 ± 0.004 1.2 ± 0.1 1.1 ± 0.1 1.24 ± 0.06 1.238 ± 0.002 1.22 ± 0.05 0.65 ± 0.02 0.55 ± 0.05 0.77 ± 0.03 0.701 ± 0.005 0.58 ± 0.04 0.84 ± 0.08 1.47 ± 0.03 1.39 ± 0.02 1.30 ± 0.06

OO ? O OL ? L OL ? O OLn ? Ln OLn ? O OP ? P OP ? O OG ? G OG ? O OS ? S OS ? O LL ? L LLn ? Ln LLn ? L LP ? P LP ? L LS ? S LS ? L LnLn ? Ln PP ? P SS ? S Pre-exponential factor, A4 (105 m6/(kmol2 min))

4.10 ± 0.06 4.40 ± 0.09 4.9 ± 0.1 4.6 ± 0.1 5.5 ± 0.4 3.8 ± 0.1 3.1 ± 0.2 4.26 ± 0.04 4.3 ± 0.3 4.1 ± 0.4 5.6 ± 0.4 4.4 ± 0.3 5.20 ± 0.09 5.93 ± 0.04 3.8 ± 0.3 4.6 ± 0.2 3.3 ± 0.2 4.0 ± 0.2 4.7 ± 0.2 3.47 ± 0.01 4.5 ± 0.3 6.2 ± 0.1 5.5 ± 0.3

1.23 ± 0.07 1.74 ± 0.01 1.204 ± 0.009 2.2 ± 0.2 1.26 ± 0.04 0.48 ± 0.06 1.15 ± 0.03 1.22 ± 0.01 1.009 ± 0.004 0.561 ± 0.008 1.04 ± 0.09 1.84 ± 0.06 2.29 ± 0.09 2.0 ± 0.2 0.57 ± 0.02 1.704 ± 0.001 0.590 ± 0.009 1.79 ± 0.07 2.3 ± 0.1 0.45 ± 0.07 0.537 ± 0.002

LLLn ? LLn LLLn ? LL LLP ? LP LLP ? LL LLS ? LS LLS ? LL LLnLn ? LnLn LLnLn ? LLn LnLnLn ? LnLn PPP ? PP SSS ? SS Pre-exponential factor, A5 (108 m6/(kmol2 min))

6.0 ± 0.5 6.14 ± 0.02 4.6 ± 0.4 3.38 ± 0.07 5.1 ± 0.2 3.5 ± 0.2 5.3 ± 0.3 6.271 ± 0.003 6.00 ± 0.5 2.7 ± 0.3 2.8 ± 0.2

LLn ? LLLn LL ? LLLn LP ? LLP LL ? LLP LS ? LLS LL ? LLS LnLn ? LLnLn LLn ? LLnLn LnLn ? LnLnLn PP ? PPP SS ? SSS Pre-exponential factor, A6 (104 m6/(kmol2 min))

1.441 ± 0.001 1.3 ± 0.1 0.76 ± 0.04 0.935 ± 0.003 0.79 ± 0.04 0.90 ± 0.06 1.69 ± 0.05 1.49 ± 0.08 1.80 ± 0.03 0.12 ± 0.02 0.11 ± 0.05

O ? OO L ? OL O ? OL Ln ? OLn O ? OLn P ? OP O ? OP G ? OG O ? OG S ? OS O ? OS L ? LL Ln ? LLn

2.21 ± 0.02 2.1 ± 0.2 3.4 ± 0.3 2.20 ± 0.07 3.9 ± 0.1 2.1 ± 0.2 1.1 ± 0.2 2.15 ± 0.08 2.3 ± 0.3 2.1 ± 0.1 1.1 ± 0.2 3.19 ± 0.04 3.0 ± 0.2

O ? EO L ? EL Ln ? ELn P ? EP G ? EG S ? ES

1.90 ± 0.01 2.130 ± 0.003 2.44 ± 0.09 1.45 ± 0.02 1.7 ± 0.2 1.87 ± 0.04

EO ? O EL ? L ELn ? Ln EP ? P EG ? G ES ? S

7.6 ± 0.3 7.92 ± 0.03 7.98 ± 0.08 7.0 ± 0.2 8.92 ± 0.04 7.6 ± 0.2

L ? LLn P ? LP L ? LP S ? LS L ? LS Ln ? LnLn P ? PP S ? SS

3.98 ± 0.07 3.5 ± 0.4 1.0 ± 0.3 3.117 ± 0.003 1.1 ± 0.3 4.4 ± 0.3 1.3 ± 0.2 1.1 ± 0.2

to initially form a homogeneous solution, while the emulsion of alcohol in oil is firstly formed in the case of methanol, and this emulsion gradually enters the pseudo-homogeneous phase with the progress of transesterification reactions. This is reflected through the lack of the initial phase for all alcohols, except for methanol, in Figs. 4 and SD.4, and the lowest conversion of TG and AE at 1 min in the case of the methanol use in Fig. SD.5. The emulsion formation with ethanol is often not reported [17,20], while it is present in acid-catalysed transesterifications [14] due to the lack of emulsifying hydroxyl ions. A homogeneous solution was reported when either butanol or tert-butanol were used [14,16,18]. As far as reaction kinetics are concerned, tert-butanol, isopropanol, methanol (considering only kinetics-determining step, not accounted for in Fig. SD.5), ethanol and butanol exhibit a decreasing initial conversion rate to AE (Fig. 4), which may be extrapolated to TG conversion rate as well, except for isopropanol being slightly more reactive in the first forward transesterification reaction (Fig. SD.4). Table 3 indicates that the conversion in equilibrium mirrors the trend, obtained for initial rate, while a noticeably higher conversion is achieved with methanol. The length of the linear

aliphatic chain in alcohol negatively influences the conversion to AE in equilibrium, which was corroborated in the literature [14,17,20], with conversions decreasing upon methanol, ethanol and butanol use, respectively. A lower conversion in equilibrium when using isopropanol versus methanol agrees with presented results, but low efficiency of tert-butanol is may be due to biocatalysis [18,20]. When correlating the structural characteristics of the studied alcohols with their transesterification kinetic parameters (Tables 1, 2 and Figs. SD.3, Fig. SD.6) it is hard to parallel the results due to the lack of the analogous data in the literature. Regardless of response surface models employed (Figs. 5 and SD.7), alcohol carbon number (CNA) negatively influences A1, while the effect of alkyl branches (ABA) is the opposite; higher alcohols are thus less prone to react with TG due to the apparently-prevailing steric hindrance of hydroxyl, while branched alcohols obviously easier react with the 1- and 3-positions in glycerol-bonded ester groups per comparable CNA value. Both effects seem to be positive for backward reaction (A2). Linear models reveal that for the second (A3) and third (A5) forward reaction, both effects seem to increase the value of pre-exponential factor, steric hindrance being surpasses by the

116

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

Table 2 Activation energies of reaction rate constants for catalysed transesterification reactions of TG, DG and MG with ethanol, regressed for different temperatures, mixing intensities, phase fractions, catalyst contents, oils and catalysts. Activation energy, Ea1 (104 J/mol)

Activation energy, Ea2 (104 J/mol)

Activation energy, Ea3 (104 J/mol)

OOO ? OO OOL ? OL OOL ? OO OOLn ? OLn OOLn ? OO OOP ? OP OOP ? OO OOG ? OG OOG ? OO OLL ? LL OLL ? OL OLLn ? LLn OLLn ? OLn OLLn ? OL OLP ? LP OLP ? OP OLP ? OL OLS ? LS OLS ? OS OLS ? OL OLnLn ? LnLn OLnLn ? OLn LLL ? LL

OO ? OOO OL ? OOL OO ? OOL OLn ? OOLn OO ? ? OOLn OP ? OOP OO ? OOP OG ? OOG OO ? OOG LL ? OLL OL ? OLL LLn ? OLLn OLn ? OLLn OL ? OLLn LP ? OLP OP ? OLP OL ? OLP LS ? OLS OS ? OLS OL ? OLS LnLn ? OLnLn OLn ? OLnLn LL ? LLL

OO ? O OL ? L OL ? O OLn ? Ln OLn ? O OP ? P OP ? O OG ? G OG ? O OS ? S OS ? O LL ? L LLn ? Ln LLn ? L LP ? P LP ? L LS ? S LS ? L LnLn ? Ln PP ? P SS ? S Activation energy, Ea4 (104 J/mol)

5.45 ± 0.02 5.5568 ± 0.0009 4.9 ± 0.1 5.55 ± 0.02 5.07 ± 0.08 5.4 ± 0.2 5.3 ± 0.1 4.68 ± 0.02 4.3 ± 0.1 5.50 ± 0.03 5.50 ± 0.01 5.74 ± 0.02 5.1 ± 0.1 4.9 ± 0.1 5.1 ± 0.1 5.3 ± 0.1 5.6 ± 0.3 4.9 ± 0.2 4.7 ± 0.2 5.2 ± 0.2 5.2 ± 0.1 5.56 ± 0.04 5.70 ± 0.06

3.7 ± 0.3 4.1 ± 0.2 3.7 ± 0.2 3.90 ± 0.04 3.4 ± 0.4 3.8 ± 0.2 4.10 ± 0.03 3.7 ± 0.3 3.3 ± 0.3 4.1 ± 0.2 3.7 ± 0.1 3.98 ± 0.08 3.75 ± 0.08 3.3 ± 0.4 3.8 ± 0.3 3.86 ± 0.06 3.6 ± 0.5 3.94 ± 0.07 3.6 ± 0.3 3.9 ± 0.2 3.6 ± 0.3 3.711 ± 0.004 3.7 ± 0.1

5.33 ± 0.06 5.6 ± 0.2 6.0 ± 0.6 5.8 ± 0.3 5.47 ± 0.03 5.21 ± 0.07 5.6 ± 0.1 5.0 ± 0.3 4.1 ± 0.2 5.0 ± 0.4 5.7 ± 0.4 5.51 ± 0.05 5.4 ± 0.1 5.45 ± 0.07 5.2 ± 0.3 5.7 ± 0.3 5.45 ± 0.02 5.38 ± 0.03 5.52 ± 0.02 4.92 ± 0.02 5.7 ± 0.4

LLLn ? LLn LLLn ? LL LLP ? LP LLP ? LL LLS ? LS LLS ? LL LLnLn ? LnLn LLnLn ? LLn LnLnLn ? LnLn PPP ? PP SSS ? SS Activation energy, Ea5 (104 J/mol)

5.63 ± 0.02 5.66 ± 0.08 4.75 ± 0.07 5.7 ± 0.1 5.30 ± 0.04 5.38 ± 0.08 5.1 ± 0.1 5.1 ± 0.1 5.57 ± 0.01 5.5 ± 0.5 5.1 ± 0.1

LLn ? LLLn LL ? LLLn LP ? LLP LL ? LLP LS ? LLS LL ? LLS LnLn ? LLnLn LLn ? LLnLn LnLn ? LnLnLn PP ? PPP SS ? SSS Activation energy, Ea6 (104 J/mol)

4.00 ± 0.2 3.9 ± 0.2 3.91 ± 0.04 3.9 ± 0.3 3.83 ± 0.06 4.00 ± 0.1 4.00 ± 0.2 3.5 ± 0.1 3.5 ± 0.2 3.8 ± 0.1 4.14 ± 0.05

O ? OO L ? OL O ? OL Ln ? OLn O ? OLn P ? OP O ? OP G ? OG O ? OG S ? OS O ? OS L ? LL Ln ? LLn

3.6 ± 0.4 3.7 ± 0.3 4.2 ± 0.3 4.4 ± 0.3 3.6 ± 0.2 4.4 ± 0.2 3.9 ± 0.4 3.1 ± 0.2 4.1 ± 0.2 3.6 ± 0.3 4.3 ± 0.3 4.2 ± 0.2 3.9 ± 0.1

O ? EO L ? EL Ln ? ELn P ? EP G ? EG S ? ES

5.6 ± 0.3 5.37 ± 0.06 5.4 ± 0.1 4.8 ± 0.3 5.09 ± 0.01 5.2 ± 0.1

EO ? O EL ? L ELn ? Ln EP ? P EG ? G ES ? S

4.08 ± 0.07 3.6 ± 0.2 3.49 ± 0.06 3.42 ± 0.06 4.27 ± 0.06 4.21 ± 0.06

L ? LLn P ? LP L ? LP S ? LS L ? LS Ln ? LnLn P ? PP S ? SS

3.8 ± 0.1 4.41 ± 0.02 3.84 ± 0.01 3.8036 ± 0.0002 4.5 ± 0.3 4.00 ± 0.05 3.1 ± 0.1 4.3 ± 0.3

electronegativity effects of groups, and reactions being favoured for long-chain alcohols. Only ABA negatively influences the second (A4) and third (A6) backward reactions, the steric effect of alcohol branches apparently lowering the reactivity with AE. Upon comparison of linear (Fig. SD.7) and mixed (Fig. 5) models, the only difference is present in the decreasing effect of ABA on A3 and A5 in the latter, since mixed term (CNAABA) compensates for the observed positive influence in the linear model. Similar findings may be deduced for the dependency of the activation energies (Fig. 6) on the structural characteristics (CN, DB and AB) of oil-originating components and alcohols. In addition to methanol, the majority of studies employed ethanol for transesterification. Santori et al. [31] reported that even at less favourable transesterification process conditions (the temperature, alcohol and catalyst concentrations of 25 °C, 400 rpm, 1:4.5 (mol/mol) and 0.5 wt.% in the case of methanol, and 26 °C, 600 rpm, 1:6.0 (mol/mol) ester to hydroxyl group and 1.0 wt.% catalyst per oil weight in the case of ethanol), biodiesel equilibrium yield was still higher upon the use of methanol, which is in agreement with the results, presented in Table 3, both at 40 and 50 °C. This is a relatively general observation that is, although the utilization of ethanol basically eliminates transfer phenomena-determin-

ing step due to a less polar character of this alcohol in comparison to methanol (Figs. 4 and SD.4), chemical equilibrium is shifted towards reactants and reaction intermediates (Table 3). What is more, ethanol causes difficulties during the downstream processing of product due to the formation of the azeotrope with water, which is used for the washing of the soaps from biodiesel [31].

4.4. Catalyst in biodiesel production process and effect of its type on biodiesel synthesis and yield Homogeneous catalyst concentration linearly influences transesterification reaction rate as long as catalyst concentration does not surpass that of alcohol, but the intrinsic catalyst concentration-independent pre-exponential factors are seldom considered [1,6,9,10,20], and even when they are, not acknowledging mass transfer resistance and fatty acid composition [11,16]. The model predictions in Figs. 4 and SD.4 excellently fit measured data, and the determined pre-exponential factors in Tables 1 and SD.3 are independent of hydroxyl ion (catalyst) concentration, while their multiplication with the latter yields their apparent values.

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

117

Fig. 5. Parity plots for linear/mixed model-predicted structure dependencies of pre-exponential factors of TG, DG and MG transesterification reactions (CNTG, CNDG, CNMG, CNAE, CNA, DBTG, DBDG, DBMG, DBAE, and ABA are number of carbons (CN), double bonds (DB), and alkyl branches (AB) in triglyceride, diglyceride, monoglyceride, alkyl ester and alcohol) determined by nonlinear regression method for base-catalysed transesterification using temperature range of 30–70 °C, rotational speed of mechanical mixing range of 100–600 rpm, phase fraction range of 1:3–1:8 (mol/mol) (oil to methanol), and catalyst content range of 0.2–1.2 wt.% (per oil).

118

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

Fig. 6. Parity plots for linear/mixed model-predicted structure dependencies of activation energies of TG, DG and MG transesterification reactions (CNTG, CNDG, CNMG, CNAE, CNA, DBTG, DBDG, DBMG, DBAE, and ABA are number of carbons (CN), double bonds (DB), and alkyl branches (AB) in triglyceride, diglyceride, monoglyceride and alkyl ester) determined by nonlinear regression method for base-catalysed transesterification using temperature range of 30–70 °C, rotational speed of mechanical mixing range of 100– 600 rpm, phase fraction range of 1:3–1:8 (mol/mol) (oil to methanol), and catalyst content range of 0.2–1.2 wt.% (per oil).

119

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

Table 3 Effect of oil, alcohol and catalyst type, and mixture temperature on equilibrium transesterification of different oils and biodiesel sample characteristics at 400 rpm, 1:6 (mol/mol) (oil to alcohol) and 0.8 wt.% catalyst (per specific oil weight) unless specified otherwise.

a b c d e f g

Temperature (°C)

Alcohol (volume fraction) (%)

–log (acid dissociation constant) (/)

Triglyceride fraction (wt.%)

Diglyceride fraction (wt.%)

Monoglyceride fraction (wt.%)

AE fraction (wt.%)

AE conversion (%)

Cetane number (/)g

Oil type 40a 40b 40c 40d 40e 50a 50b 50c 50d 50e

Methanol Methanol Methanol Methanol Methanol Methanol Methanol Methanol Methanol Methanol

15.2 15.2 15.2 15.2 15.2 14.9 14.9 14.9 14.9 14.9

8.7 8.8 10.2 11.4 7.6 4.3 4.2 5.8 6.5 4.1

3.2 1.7 2.7 1.9 3.2 1.5 0.8 1.4 1.1 1.4

1.3 1.0 1.3 1.4 1.5 1.7 1.6 1.9 2.1 1.9

86.7 88.5 85.8 85.3 87.8 92.5 93.4 90.9 90.2 92.5

83.9 86.6 83.1 82.8 84.7 89.7 90.9 87.9 87.1 89.4

45.5 44.8 44.5 42.6 44.2 45.4 44.7 44.4 42.6 44.1

Alcohol type 50a 50a 50a 50a 50a

Methanol (20.7) Ethanol (20.5) Isopropanol (20.5) Butanol (20.5) tert-Butanol (20.5)

14.9 15.1 15.7 15.3 19.5

4.3 8.0 6.3 1.5 0.1

1.5 2.3 1.0 0.6 0.0

1.7 4.7 3.0 10.6 0.1

92.5 85.0 89.7 87.3 99.8

89.7 83.7 86.0 80.2 99.9

45.4 44.7 44.9 44.8 45.3

Catalyst type 40a 50a 40f 50f

Methanol Methanol Methanol Methanol

15.2 14.9 15.2 14.9

8.7 4.3 7.2 3.8

3.2 1.5 2.3 1.2

1.3 1.7 1.5 2.4

86.7 92.5 89.0 92.6

83.9 89.7 86.1 89.0

45.5 45.4 45.5 45.4

(20.5) (20.6) (20.1) (20.5) (20.5) (20.7) (20.6) (20.1) (20.6) (20.6)

(20.5) (20.7) (20.5) (20.7)

Canola oil, KOH. Palm oil, KOH. Peanut oil, KOH. Soybean oil, KOH. Sunflower oil, KOH. Canola oil, NaOH. Calculated from fatty acid composition according to Piloto-Rodríguez et al. [22].

The equivalent molar concentration of any completely soluble homogeneous catalyst should result in the same conversion in equilibrium, which is indeed the case [3–6]; nonetheless, equivalent mass concentrations (as the content of catalyst is usually given per oil weight and the price of catalyst is estimated per its mass) favour the catalyst with a lower molar mass, i.e. better performance of NaOH in comparison to KOH in Figs. 4 and SD.4, and Table 3. In industrial scale, however, molar equivalence relation and described NaOH advantages often do not apply due to saponification, high water and free fatty acid content. The mechanism of base-catalysed transesterification involves the nucleophile attack on the ester bond in glyceride, where nucleophile is represented by alkoxy (e.g. methoxy, ethoxy, etc.) group. Owing to the fact that the chemical equilibrium between alcohol and catalyst is established much faster in comparison to transesterification, and that the amount of catalyst is much lower than stoichiometric, as per alcohol, the derivation of Eqs. (4)–(9) is valid that is the rate of transesterification is linearly dependent on hydroxide ion concentration. The latter agrees rather well with the measured data at different catalyst concentrations [29] and is valid for various temperatures and homogeneous catalysts as well (Fig. 4). Failure to straightforwardly incorporate this linear dependence may perhaps result in a better overall agreement of predicted and measured data, but will inadvertently result is a less comprehensive model, involving rate constants, decreasing with catalyst concentration [31].

in terms of the conversion to AE in equilibrium, regardless of investigated oil and its composition. Even though increasing temperature affects backward reactions as well, the influence on forward reactions is much more pronounced within the temperature window of biodiesel production. Moreover, Table 3 implies that temperature favourably influences biodiesel yield even upon the use of different homogeneous base catalysts or various alcohols (data not shown). The latter observations predominantly apply even upon the utilization of other types of catalysis, such as acidic or enzymatic [1–20]. Temperature, nonetheless, usually has to be below alcohol boiling point, which promotes the use of renewable higher alcohols (bioethanol and biobutanol), should their price become comparable to (relatively inexpensive) methanol. Even though temperature increases the yield of the alkyl esters within the whole temperature interval below methanol (most commonly used alcohol) boiling point [29], which is valid for different oils, alcohols and homogeneous catalysts as well (Table 3), one always has to acknowledge the reversible nature of chemical reactions, pertinent to transesterification, and the Arrhenius law for both forward and backward reactions. The increase of the product yield with temperature thus has to be tested experimentally, especially in cases, in which mechanism is more complex (e.g. adsorption–reaction–desorption mechanisms, related to heterogeneous catalysis [32–34]), or when using the alcohols with higher boiling points than methanol, for which the extrapolation of product conversion with temperature might come to a turning point, when backward reactions commence to prevail.

4.5. Effect of reaction temperature on biodiesel synthesis and yield 5. Conclusions Figs. 4, SD.4 and SD.5 show that higher temperature shortens initial heterogeneous phase, increases the apparent conversion rates of TG and AE, and, as evident from Table 3, improves the biodiesel yield

In this study, physical models describing reaction medium were developed and utilized for the modelling of the alkali-catalysed

120

B. Likozar, J. Levec / Applied Energy 123 (2014) 108–120

biodiesel production in batch process, individually considering the densities and viscosities of different oils used, and the distribution coefficients of the components between methanol and different oils, versus temperature (the systems with other alcohols were homogeneous), and comparing the mass transfer/kinetics/equilibrium modelling predictions and experimental observations of the composition of the reaction mixture during transesterification (alcoholysis) of different oils at various temperatures, phase fractions (oil to alcohol), catalyst concentrations (per specific oil weight) and impeller speeds. The main advantage of the developed model is its acknowledgement of oil composition, while the future work includes its extension for heterogeneous inorganic, polymeric and enzymatic catalysis. Acknowledgements The provision of financial support for the conduct of the research and preparation of the article by Slovenian Research Agency (ARRS) (Program P20152) is gratefully acknowledged. The authors would also like to thank to Matic Skornšek for providing experimental help during research. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.apenergy.2014. 02.046. References [1] Fjerbaek L, Christensen KV, Norddahl B. A review of the current state of biodiesel production using enzymatic transesterification. Biotechnol Bioeng 2009;102:1298–315. [2] Kashid MN, Kiwi-Minsker L. Microstructured reactors for multiphase reactions: state of the art. Ind Eng Chem Res 2009;48:6465–85. [3] Leung DYC, Wu X, Leung MKH. A review on biodiesel production using catalyzed transesterification. Appl Energy 2010;87:1083–95. [4] Qiu ZY, Zhao LN, Weather L. Process intensification technologies in continuous biodiesel production. Chem Eng Process 2010;49:323–30. [5] Lam MK, Lee MT, Mohamed AR. Homogeneous, heterogeneous and enzymatic catalysis for transesterification of high free fatty acid oil (waste cooking oil) to biodiesel: a review. Biotechnol Adv 2010;28:500–18. [6] Shahla S, Cheng NG, Yusoff R. An overview on transesterification of natural oils and fats. Biotechnol Bioproc E 2010;15:891–904. [7] Narvaez PC, Sanchez FJ, Godoy-Silva RD. Continuous methanolysis of palm oil using a liquid–liquid film reactor. J Am Oil Chem Soc 2009;86:343–52. [8] Harvey AP, Mackley MR, Seliger T. Process intensification of biodiesel production using a continuous oscillatory flow reactor. J Chem Technol Biot 2003;78:338–41. [9] Klofutar B, Golob J, Likozar B, Klofutar C, Zˇagar E, Poljanšek I. The transesterification of rapeseed and waste sunflower oils: mass-transfer and kinetics in a laboratory batch reactor and in an industrial-scale reactor/ separator setup. Bioresource Technol 2010;101:3333–44. [10] Noureddini H, Zhu D. Kinetics of transesterification of soybean oil. J Am Oil Chem Soc 1997;74:1457–63. [11] Vicente G, Marchtinez M, Aracil J, Esteban A. Kinetics of sunflower oil methanolysis. Ind Eng Chem Res 2005;44:5447–54. [12] Duangsuwan W, Tuzun U, Sermon PA. Configurations and dynamics of single air/alcohol gas-liquid compound drops in vegetable oil. Chem Eng Sci 2009;64:3147–58.

[13] Da Silva C, de Castilhos F, Oliveira JV, Filho LC. Continuous production of soybean biodiesel with compressed ethanol in a microtube reactor. Fuel Process Technol 2010;91:1274–81. [14] Ataya F, Dube MA, Ternan M. Variables affecting the induction period during acid-catalyzed transesterification of Canola oil to FAME. Energy Fuel 2008;22:679–85. [15] Nawaratna G, Fernando SD, Adhikari S. Response of titanium-isopropoxidebased heterogeneous amphiphilic polymer catalysts for transesterification. Energy Fuel 2010;24:4123–9. [16] Pappu VKS, Yanez AJ, Peereboom L, Muller E, Lira CT, Miller DJ. A kinetic model of the amberlyst-15 catalyzed transesterification of methyl stearate with nbutanol. Bioresource Technol 2011;102:4270–2. [17] Sotoft LF, Westh P, Christensen KV, Norddahl B. Novel investigation of enzymatic biodiesel reaction by isothermal calorimetry. Thermochsim Acta 2010;501:84–90. [18] Fu B, Vasudevan PT. Effect of solvent-co-solvent mixtures on lipase-catalyzed transesterification of canola oil. Energy Fuel 2010;24:4646–51. [19] Marjanovic´ AV, Stamenkovic´ OS, Todorovic´ ZB, Lazic´ ML, Veljkovic´ VB. Kinetics of the base-catalyzed sunflower oil ethanolysis. Fuel 2010;89:665–71. [20] Colucci JA, Borrero EE, Alape F. Biodiesel from an alkaline transesterification reaction of soybean oil using ultrasonic mixing. J Am Oil Chem Soc 2005;82:525–30. [21] Di Nicola G, Pacetti M, Polonara F, Santori G, Stryjek R. Development and optimization of a method for analyzing biodiesel mixtures with non-aqueous reversed phase liquid chromatography. J Chromatogr A 2008;1190:120–6. [22] Piloto-Rodríguez R, Sánchez-Borroto Y, Lapuerta M, Goyos-Pérez L, Verhelst S. Prediction of the cetane number of biodiesel using artificial neural networks and multiple linear regression. Energy Convers Manage 2013;65:255–61. [23] Saien J, Shabnam D. Modelling mass transfer coefficient for liquid-liquid extraction with the interface adsorption of hydroxyl ions. Korean J Chem Eng 2009;26:963–8. [24] Skelland AHP, Tedde DW. Extractionorganic chemicals processing. In: Rousseau RW, editor. Handbook of separation process technology. Hoboken: John Wiley & sons, Inc.; 1987. p. 405–66. [25] Siddiqi MA, Lucas K. Correlations for prediction of diffusion in liquids. Can J Chem Eng 1986;64:839–43. [26] Hilal SH, Karickhoff SW, Carreira LA. Prediction of the vapor pressure boiling point, heat of vaporization and diffusion coefficient of organic compounds. QSAR Comb Sci 2003;22:565–74. [27] Hilal SH, Karickhoff SW, Carreira LA. Prediction of the solubility, activity coefficient and liquid/liquid partition coefficient of organic compounds. QSAR Comb Sci 2004;23:709–20. [28] Sastri SRS, Rao KK. A new group contribution method for predicting viscosity of organic liquids. Chem Eng J Biochem Eng 1992;50:9–25. [29] Likozar B, Levec J. Effect of process conditions on equilibrium, reaction kinetics and mass transfer for triglyceride transesterification to biodiesel: experimental and modeling based on fatty acid composition. Fuel Process Technol 2014;122:30–41. [30] Yin X, Ma H, You Q, Wang Z, Chang J. Comparison of four different enhancing methods for preparing biodiesel through transesterification of sunflower oil. Appl Energy 2012;91:320–5. [31] Santori G, Di Nicola G, Moglie M, Polonara F. A review analyzing the industrial biodiesel production practice starting from vegetable oil refining. Appl Energy 2012;92:109–32. [32] Talebian-Kiakalaieh A, Amin NAS, Zarei A, Noshadi I. Transesterification of waste cooking oil by heteropoly acid (HPA) catalyst: optimization and kinetic model. Appl Energy 2013;102:283–92. [33] Badday AS, Abdullah AZ, Lee KT. Ultrasound-assisted transesterification of crude Jatropha oil using alumina-supported heteropolyacid catalyst. Appl Energy 2013;105:380–8. [34] Dawodu FA, Ayodele O, Xin J, Zhang S, Yan D. Effective conversion of nonedible oil with high free fatty acid into biodiesel by sulphonated carbon catalyst. Appl Energy 2014;114:819–26. [35] Sostaric M, Klinar D, Bricelj M, Golob J, Berovic M, Likozar B. Growth, lipid extraction and thermal degradation of the microalga Chlorella vulgaris. New Biotechnol 2012;29:325–31.