Biomass modelling: Estimating thermodynamic properties from the elemental composition

Fuel 181 (2016) 207217 Contents lists available at ScienceDirect Fuel journal homepage: www.elsevier.com/locate/fuel Full Length Article Biomass modelling: Estimating thermodynamic properties from the elemental composition Emanuela Peduzzi a,, Guillaume Boissonnet b, Franỗois Marộchal a a b ẫcole Polytechnique Fộdộrale de Lausanne, Industrial Process and Energy Systems Engineering, Rue de lIndustrie 17, Case Postale 440, CH-1951 Sion, Switzerland CEA Grenoble DRT/LITEN/DTBH/Laboratoire des Technologies Biomasse, 17 rue des Martyrs, 38054 GRENOBLE cedex 9, France h i g h l i g h t s  Accuracy of heating value correlations on a consistent database of biomass samples  Complete and coherent lignocellulosic biomass model for numerical simulations  Enthalpy and Gibbs free energy as linear correlations of the elemental composition  Exergy in terms of thermodynamic properties and composition of the environment a r t i c l e i n f o Article history: Received 14 September 2015 Received in revised form 14 April 2016 Accepted 21 April 2016 Keywords: Biomass Heating value Modelling Exergy Gibbs free energy a b s t r a c t In the context of modelling biomass conversion processes, the accurate representation of biomass, which is a complex and highly variable material, is of crucial importance This study provides a rather simple and flexible way to represent biomass, especially suited in the context of thermochemical conversion processes The procedure to represent the enthalpy of formation, the Gibbs free energy and the exergy of biomass in terms of its elemental composition (C, H, O, N, S) and moisture content is outlined The correlations relating the heating value to the elemental composition of biomass are evaluated through a database of over one hundred raw and pretreated biomass samples Results show that such correlations can predict the higher heating value (HHV) within an accuracy of 1.93% and 2.38% One of the correlations is then applied to represent the enthalpy of formation of biomass as a linear function of the elemental composition The procedure is extended to estimate the Gibbs free energy of formation and subsequently the exergy of biomass, which are expressed as linear functions of the elemental composition The method proposed for the estimate of exergy allows taking directly into account the composition of the reference environment Results show that the method proposed in this study agrees within 1% accuracy with the widely used correlation proposed by Szargut et al (1988) The values obtained for Exergy, over the range of compositions of the samples considered, vary in general between 105% and 115% of the lower heating value (LHV) and 103% and 107% of the higher heating value obtained using the literature correlation by Boie (1953) On the basis of these correlations, this study provides the thermodynamic properties of C, H, O, N, S and bound water pseudo-compounds that can be used in the thermodynamic properties evaluation packages used in flowsheeting software and in numerical simulations for a coherent description of biomass as a function of its composition ể 2016 Elsevier Ltd All rights reserved Introduction The first challenge in the thermochemical modelling of biomass conversion systems is to model biomass itself, the raw material Biomass is not a standard compound and its chemical, elemental composition, as well as its thermal properties vary significantly Corresponding author E-mail address: emanuela.peduzzi@epfl.ch (E Peduzzi) http://dx.doi.org/10.1016/j.fuel.2016.04.111 0016-2361/ể 2016 Elsevier Ltd All rights reserved Different solutions are adopted or suggested, for example, to model biomass in flowsheeting software In an example bioethanol production unit, modelled in ProSim Plusề [1], biomass is represented as a mixture of its chemical constituents by adapting the properties of glucose, such as the chemical formula, the heat of formation and the molecular weight, to represent cellulose, hemicellulose and lignin Aspen Plusề allows the implementation of organic substances as non-conventional solid compounds through the definition of attributes in terms of ultimate (i.e elemental 208 E Peduzzi et al / Fuel 181 (2016) 207217 Nomenclature Acronyms ar as received basis daf dry ash free basis db dry basis EMC equilibrium moisture content fsp fiber saturation point HHV higher heating value hw hard wood LHV lower heating value MAE mean absolute error MBE mean bias error MC moisture content RMSE root mean square error SI supplementary information SRC short rotation coppice SRF short rotation forestry sw soft wood Roman letters B00 biomass on a dry basis, 0% moisture () Bch standard chemical exergy (kJ mol1) BX biomass on a wet basis, X% moisture () composition) and proximate (i.e fixed carbon, volatile matter and ash content) analysis The heating value, the heat of formation and the heat capacity are then calculated by selecting the relevant literature correlations, generally developed for coal, in the General Coal Enthalpy Model (HCOALGEN property model [2]) The attributes can be changed through the use of subroutines to represent changes in composition during conversion, for example during coal devolatilisation Rửnsch and Wagner [3] analysed the application of the empiric correlations used in AspenPlusề to model wood and straw They showed that the heating value correlations, developed for coal, generally underestimate the values for biomass and they indicated which correlations can predict the heating values of average wood and wheat straw samples within their standard deviation The objective of this study is to propose an accurate and consistent definition of biomass, especially relevant for the simulation of thermochemical conversion processes The goal is to develop a generic representation of the thermodynamic properties of biomass that allows to easily update process performances as a function of the type of biomass, which is considered on the basis of its elemental composition The model developed is general and may be used in any numerical simulation but in this study it is applied to represent biomass in the flowsheeting software Valiề by Belsim [4] Therefore the methodology presented refers rigorously to thermodynamics but the formalism refers sometimes to this software Thermodynamic properties packages offered in flowsheeting software, such as Valiề, allow in fact the definition of new pseudo-compounds, that is, user-defined compounds used to model substances that are not present in the internal database for which thermodynamic properties need to be defined In this study, special attention is given to the coherence of the heating value calculated from the thermochemical properties of the pseudo-compounds The heating value represents the energy content of biomass and is one of the most important properties for the design and simulation of biomass thermochemical conversion systems [5] The steps leading to the representation of biomass are presented in detail First of all a biomass database, obtained in the context of previous work, is used to extract information regarding the elemental composition and heating value of many representative cp;dry cp;wet G H HHV KB LHV Mm Mb M fsp MC R S T T00 heat capacity, dry basis (kJ kgB00 K1) heat capacity, wet basis (kJ kgtot K1) standard Gibbs free energy (kJ mol1) standard enthalpy (kJ mol1) higher heating value (kJ kgB00 or kJ mol1) constant by Battley () lower heating value (kJ kgB00 or kJ mol1) molar mass (g mol ) moisture content referring to bound water (kgH2 O kgB00 ) moisture content after evaporation of free water (kgH2 O kgB00 ) moisture content (kgH2 O kgB00 ) ideal gas constant (J mol1 K1 standard entropy (kJ mol1 K1) temperature (K or C) Torrefied biomass, 0% moisture () Greek letters U humidity (kgH2 O kgtot ) r standard deviation () lignocellulosic biomass samples This information is used as a basis for comparing several correlations available in the literature relating the elemental composition to the heating value in the range of compositions of biomass types considered in this study The systematic comparison of the correlations and corresponding errors allows one to identify the correlation best fitting the experimental data Finally, given a correlation, the approach and the fundamental assumptions used to model biomass are presented The properties considered are the heating value with the corresponding enthalpy of formation, the heat capacity, the heat of adsorption of the moisture content Furthermore the modelling approach is extended to the entropy, the Gibbs free energy and the exergy of biomass In particular, the exergy model presented allows the explicit definition of the exergy of biomass based on its thermodynamic properties and the properties of the reference environment The definition of biomass From a legal standpoint biomass is the biodegradable fraction of products, waste and residues from biological origin from agriculture (including vegetal and animal substances), forestry and related industries including fisheries and aquaculture, as well as the biodegradable fraction of industrial and municipal waste [6] Biomass therefore includes a large variety of materials In this study biomass refers only to lignocellulosic materials from forestry and agricultural products, namely wood and straw of different types Biomass is generally defined by considering its chemical/structural (i.e cellulose, hemicellulose, lignin and extractives), proximate and ultimate analysis The ultimate analysis and the heating value are especially important for the definition of biomass in a thermochemical model, as they provide the basis for the atomic and energy balance of a conversion process In terms of ultimate analysis, biomass is mainly composed of C, H, and O, which define, for the most part, its heating value It also contains small quantities of N, S, Cl These six elements make up the organic phase of biomass The inorganic phase contains Si, Al, Ti, Fe, Ca, Mg, Na, K, S, P, and other minor elements which are E Peduzzi et al / Fuel 181 (2016) 207217 important for ash characterisation Biomass usually also contains traces of heavy metals [7] The heating value of a fuel, such as biomass, represents its energy content It is defined as the heat released by the complete combustion of a unit of volume of the fuel at bar (101325 Pa), considering reactants and products at the same reference temperature [8] The reference temperature, by convention, is considered here at 25 C (T  is 298.15 K) It is possible to distinguish between two types of heating values: the higher heating value (HHV) if the heat of condensation of water generated in the combustion (from the hydrogen originally present in the biomass) is recovered and the lower heating value (LHV) if this water is considered in its vapour state (and therefore its heat of condensation is not recovered) Due to the presence of water and ash in biomass, the HHV and LHV may be referred to on a weight basis: dry basis (db), dry ash free basis (daf ), or as received basis (ar) The relationship between HHV and LHV on a dry basis and the relationship between LHV on a dry basis and LHV on an as received basis are reported in Eqs (1) and (2) respectively ~ ẳ LHV ~ db ỵ M m;H2 O H DH ~ HHV v ap kJ kg ị Mm;H 100 ~ wb ẳ LHV ~  U kJ kg1 ị ~ db Uị DH LHV v ap 1ị 2ị M m;H2 O and M m;H are the molar masses of water and hydrogen, H ~  is the enthalpy is the hydrogen mass percentage in the fuel, DH v ap of vaporisation of water at the reference temperature (in kJ kg1), $ U is the humidity (in kgH2 O kg1 is used to refer to properties tot ) and on a mass basis In this study, unless otherwise specified, all values refer to dry or dry ash free basis The evaluation of the ultimate analysis and the heating value of different biomass types is summarised hereafter with the introduction of the biomass database used in this study Biomass database The biomass database used in this study is a collection of experimental elemental compositions and corresponding HHV and LHV These values are obtained, in the context of previous studies, from the analysis of over one hundred samples, which include both raw and torrefied biomass The raw biomass database relies on the characterisation, carried out by Dupont et al [9], of 92 representative samples The standards used to obtain the ultimate analysis and the heating values are summarised in Table A.1 in the supplementary information (SI) Hereafter, as in the previous study by Dupont et al [9], the data is classified in biomass families Hardwoods (angiosperms) are represented by broad-leafed trees such as oaks, beech trees, hornbeam and lime (basswoods) trees Softwoods (gymnosperms) are represented by conifers such as pine trees, fir trees, and spruce The database includes also short rotation coppice (SRC) and short rotation forestry (SRF) represented by willow and poplar, and agricultural biomass represented by straw from barley, corn, rape, and wheat, and energy crops from alfalfa, miscanthus, fiber and sweet sorghum, switchgrass, tall fescue, and triticale The data relative to torrefied biomass was obtained by Marty [10], by Nocquet et al [11,12] and other previous studies The database was further enriched including biomass chars and, as a reference, coal, from Parikh et al [13] The average values of the ultimate analysis and heating values obtained for the different raw biomass species are summarised in Table Results presented in Table show that biomass composition is similar across biomass types However, agricultural residues and energy crops, display a sensibly smaller heating value and carbon content and higher ash content than other raw biomass types 209 Correlations estimating the heating value The direct measure of the heating value of biomass, through bomb calorimetry, is a complicated and time-consuming procedure It can therefore be interesting to use a correlation to calculate the heating value from other conventional properties of biomass, i.e proximate and ultimate analysis, which can be obtained relatively easily and cheaply [5] An extensive review of correlations relating the heating value of biomass to its ultimate, proximate, and chemical/structural analysis, with attention to biomass type, is presented by Vergas-Moreno et al [7] This review also highlights the lack of information which is sometimes encountered in the literature in terms of the data used to develop and validate the correlations and their basis (dry basis or dry ash free), but also in terms of the transcription errors when correlations are referred to by other authors The first model developed to calculate the heating value from an elemental composition was proposed by Dulong in 1880 and concerned coal samples During the 20th century many models were developed to estimate the heating value of coal and, over the last 30 years, also of biomass [7] A survey of published correlations is also reported by Channiwala and Parikh [14], who reviewed several correlation types and relative basic assumptions Sheng and Azevedo [5] tested the performance of correlations relating the heating value to the elemental, proximate and chemical analysis on a large database of biomass samples collected from the literature They show that the correlations relating the proximate composition and the heating value display low accuracy, but they have recently gained importance among engineers and researchers due to the ease and speed of proximate analysis [7] The correlations based on chemical/structural analysis also present low accuracy because of the variation of biomass components properties and chemical composition On the contrary, the correlations relying on the ultimate analysis are the ones that provide the most accurate results [5] In thermochemical process modelling the use of correlations relating the heating value to the elemental composition is a good compromise between accuracy and ease of computation Given the database presented in Section it is possible to either develop a new correlation or use the biomass database to validate the correlations proposed in other studies The interest in using this database lies in the use of samples which cover a wide range of biomass fuels, from agricultural to woody and torrefied biomass, and in the consistency of the measurement procedures However, given the relatively small composition variations of the biomass samples, the unequal distribution across the heating values and in the interest of a most general approach, correlations from the literature are validated using the database 4.1 Evaluation of literature correlations A survey of correlations from the literature estimating the heating value from the elemental composition is reported in Table These correlations have been developed considering biomass samples or, even if they were developed for coals or other hydrocarbon fuels, are sometimes used in the context of biomass The oldest correlations reported, for reference, were developed for coal and are the ones by Dulong, Eq (3), and by Mott, Eq (4) [15] The correlation by Boie [16], Eq (5), was derived using hydrocarbon fuels with an expected error within 1.8% [14] This correlation is sometimes used to model biomass in flowsheeting software [1719] The relationship proposed by the Institute of Gas Technology (IGT) [20], Eq (6), is very general and was developed on the basis of over 700 coal samples, with an error that lies within 1.2% [14] 210 E Peduzzi et al / Fuel 181 (2016) 207217 Table Average composition and HHV of different types of biomass on a dry basis The number of samples of each biomass type is presented in parenthesis Hardwood (33) Carbon (wt%) Hydrogen (wt%) Oxygenb (%wt) Nitrogen (wt%) Sulfur (mg kg1) Ash, 550 C (wt%) HHVdb (kJ kg1) LHVdb (kJ kg1) a b Mix (3)a Softwood (16) SRC & SRF (11) Ag Biomass (23) Av r Av r Av r Av r Av r 49.7 5.8 42.7 0.2 171.2 1.74 19811 18612 1.0 0.1 1.1 0.1 133.7 0.64 448.4 442.5 51.2 5.9 41.9 0.1 270.9 0.99 20331 19128 1.3 0.2 1.3 0.1 443.8 0.65 704.3 707.2 50.8 5.9 40.8 0.2 243.3 2.33 19947 18740 1.0 0.1 1.3 0.0 108.7 1.31 480.9 477.5 49.1 6.0 41.9 0.3 609.3 2.88 19552 18321 0.3 0.1 1.1 0.2 673.8 1.08 612.6 614.4 46.8 5.8 42.3 0.8 1,355.5 5.14 18458 17263 1.4 0.2 2.4 0.5 645.1 2.08 703.8 703.2 Mix refers to hardwood and softwood samples Oxygen is calculated by difference Table Survey of correlations estimating the HHV from the elemental composition a b c d Authors (year) Correlations: HHV (kJ kg1), dry basisa Dulong (1880)b Mott-Spponer (1940)b Boie (1953) [16]b IGT (1973) [20] Tillman (1978) [21] Grabosky and Bain (1981)c Channiwala and Parikh (2002) [14] Sheng and Azevedo (2005) [5] Friedl et al (2005) [22] 338:29 C ỵ 1442:77 H ỵ 94:2 S 180:36 O 336:20 C ỵ 1419:33 H ỵ 94:2 S 153:24 72:01 O=100 Aịị O 351:70 C ỵ 1162:49 H ỵ 104:67 S 110:95 O ỵ 62:80 N 341:7 C ỵ 1322:1 H 119:8 O ỵ Nị 123:2 S=10000 15:3 A 437:3 C 1670:1 328 C ỵ 1430:6 H ỵ 92:9 S 23:7 N 40110 A=100ị H=C 349:1 C ỵ 1178:3 H ỵ 100:5 S 103:4 O 15:1 N 21:1 A 1367:5 ỵ 313:7 C ỵ 700:9 H 31:8 O0 d 3:55 C2 232 C 2230 H ỵ 51:2 C Hị ỵ 131 N ỵ 20600 (3) (4) (5) (6) (7) (8) (9) (10) (11) C, H, O, N, S, A represent respectively the carbon, hydrogen, oxygen, nitrogen, sulphur, and ash content expressed in % by mass on a dry basis A conversion factor of 2.326 kJ kg1 per BTU lb1 was considered to adapt the correlation as reported by Mason and Gandhi [15] As reported by Channiwala and Parikh [14] Here O0 is the sum of oxygen and other elements in the organic matter (S, N, Cl, etc.) and is calculated by difference O0 = 100 C H A In the context of biomass, Tillman [21] developed a correlation, Eq (7), to calculate the heating value of wood and wood barks from their carbon content and later modified it to extend its validity to the whole range of biomass materials Predictions for this equation are reported to be within 5% error Eq (8) represents the correlation developed by Grabosky and Bain offering predictions claimed to be within 1.5% [14] Channiwala and Parikh [14] developed a generalised and unified correlation, Eq (9), aiming at simultaneously describing the heating value of solid, gaseous and liquid fuels within the range 0% C 92:25%, 0:43% H 25:15%, 0:00% O 50:00%, 0:00% N 5:60%, 0:00% S 94:08%, 0:00% A 71:4%, 4.745 MJ kg1 HHV 55.345 MJ kg1, where C, H, O, N, S and A represent carbon, hydrogen, oxygen, nitrogen, sulphur and ash contents of a fuel, respectively, in wt% (dry basis).1 These fuels include terrestrial and aquatic biomass material, industrial waste and municipal solid waste, refuse and sludge, as well as chars, coals and coke [14] According to the authors this correlation yields a mean absolute percentage error of 1.45% and mean bias error (MBE) error of 0.00% Sheng and Azevedo [5] proposed a new correlation, Eq (10), based on a database of biomass samples collected from the open literature According to the authors this relationship provides 90% of predictions within 5% error More recently, Friedl et al [22] used 122 plant biomass samples and obtained a correlation, Eq (11), by ordinary least squares regression (OLS) and by partial least squares regression (PLS) which allows HHV prediction with a standard error of 2% The correlations reported in Table present different coefficients and therefore yield different results for the same As far as the authors know this is the only correlation, considered in this study, for which a validity range is specified Fig Evaluation of the performance of literature correlations estimating the HHV with respect to the measured values of the biomass database considered in this study The bars represent the difference between the average HHV calculated using the different correlations and the average HHV measured The average measured value is 19.7 MJ kg1 The grey bars represent correlations originally developed for coal and hydrocarbons whereas the green bars represent correlations originally developed for or including biomass The red lines represent the standard deviation of the measured HHV, the error bars represent the standard deviations obtained using the correlations (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) composition Comparing these correlations is difficult because the authors report different accuracies based on different samples and composition ranges Therefore, a comparison of the average results obtained using these correlations to estimate the HHV of the raw samples belonging to the biomass database considered in this study is carried out and reported in Fig Most of the correlations accurately reproduce the average HHV of the biomass samples 211 E Peduzzi et al / Fuel 181 (2016) 207217 Table Accuracy of correlations in literature estimating the HHV for the biomass database considered in this study Excluding char Dulong (1880) Mott-Spooner (1940) Boie (1953) [16] IGT (1978) [20] Tillman (1978) [21] Grabosky et al (1981) Channiwala and Parikh (2002) [14] Sheng and Azevedo (2005) [5] Friedl et al (2005) [22] Including char RMSE (kJ kg1) MAE (%) MBE (%)