
Research Article


Prediction of Heating Values of Oil Palm Fronds from Ultimate Analysis 

Ramzy Elneel,
Shaharin Anwar
and
Bambang Ariwahjoedi



ABSTRACT

The heating value is one of the most important properties
of biomass materials, as it can be used for design calculations or numerical
simulations of the thermal conversion systems of biomass. The heating value
can be determined experimentally or numerically. The determination of the heating
value experimentally (by using a bomb calorimeter) involves laborious measurements
while an ultimate analysis can be performed using automatic equipment. In this
study, a multiple linear regression analysis was used to develop an empirical
equation for the estimation of the higher heating value of Oil Palm Fronds (OPFs)
using an ultimate analysis. An empirical equation based on the main elements
carbon, hydrogen and oxygen (HHV = 0.879C+0.321H+0.056O24.826) was the most
accurate model, with a correlation coefficient (R^{2}) of 0.92, Average
Absolute Error (AAE) of 1.4% and Average Bias Error (ABE) of 0.16%.





Received: August 26, 2012;
Accepted: February 06, 2013;
Published: April 22, 2013


INTRODUCTION
The steep rise in greenhouse emissions (CO_{2}) generated by the combustion
of fossilfuels and the sustainable energy needs for the future motivate development
of renewable energy sources and advance energy technologies. Nowadays, biomass
is getting increased attention as one of the potential sources of renewable
energy. Furthermore, biomass has the flexibility of using different thermochemical
conversion processes such as combustion, pyrolysis and gasification technology
(Li et al., 2009; Gordillo
and Annamalai, 2010). Knowing some of the biomass properties (heating value,
moisture content, elemental composition, ash properties, etc) is essential for
the design and operation of biomass thermochemical systems (Friedl
et al., 2005). Moreover, the proper design and operation of biomass
conversion equipment rely significantly on the biomass properties.
The heating value (calorific value) of a biomass fuel defines its energy content
which is one of the important parameters for designing conversion systems. Generally,
the biomass heating value could be reported on two bases: (1) Higher Heating
Value (HHV) and is in reference to the heat let off with the generated as well
as original water in a condensed state during the fuel combustion; (2) Lower
Heat Value (LHV) which is related to the production of the gaseous water as
the product (Sheng and Azevedo, 2005). The heating values
of biomass materials can be determined experimentally by using an adiabatic
bomb calorimeter or they can be calculated by using mathematical models based
on the chemical composition, proximate or ultimate analysis of the biomass fuel
(Sheng and Azevedo 2005; Yin, 2011).
The experimental measurement of the biomass heating value is complicated, time
consuming and involves manual laboratory work. On the other hand, ultimate,
proximate and chemical analyses using typical or modern laboratory equipments
are accomplished with ease, speed and lessexpense (Sheng
and Azevedo, 2005). Therefore, researchers who are concerned with biomass
heating values would want to determine the heating value with an easy and fast
method within an acceptable tolerance, based on the basic characteristics of
the biomass (Akkaya, 2009).
Hence, there have been many attempts to estimate the heating value of biomass
fuels based on chemical analysis (Demirbas, 2001; Sheng
and Azevedo, 2005), proximate analysis (Parikh et
al., 2005) and ultimate analysis (Huang et al.,
2009). Moreover, Sheng and Azevedo (2005) proposed
correlations to estimate the heating values of biomass fuels, based on chemical,
proximate and ultimate analyses and found that the correlations based on the
ultimate analysis were the most accurate and reliable. However, the correlations
based on the proximate analysis had low accuracy because the proximate analysis
would provide only an empirical composition of the biomass while the correlations
based on the chemical analysis were not reliable because of the variation of
the component properties. Nevertheless, there has been no attempt contributed
to oil palm fronds as a biomass fuel. Therefore, the objective of the present
study is to develop correlations between the heating values and the ultimate
analysis for oil palm fronds produced in Malaysia.
The OPF is one of the most abundant byproducts produced in Malaysia during
the harvesting and pruning time (Kawamoto et al.,
2001). The fronds consist of the petiole and leaflets and the structure
of the fronds was found comparable to that of hardwood (Shuit
et al., 2009). Mohammed et al. (2005)
found the chemical composition of the fronds to be as follow: 49.8% cellulose,
83.5% hemicellulose and 20.5% lignin. However, due to the high amount of cellulose
and low amount of lignin, OPF considered as convenient feedstock for the gasification
technology. Therefore, currently the OPF has started to be used as a feedstock
for the gasification technology (Elneel et al., 2011;
Sulaiman et al. 2011).
EXISTING MATHEMATICAL MODELS BASED ON THE ULTIMATE ANALYSIS There have been many proposed models for calculation of heating values based
on chemical analysis, proximate analysis and ultimate analysis. The ultimate
analysis based models reported by Dulong’s (1880) was the first model developed
for calculation of heating values which was intended for prediction of heating
values for coal samples (Kathirvale et al., 2003;
VargasMoreno et al., 2012).
Thereafter, many researchers proposed the variations of Dulong’s model,
including new coefficients and sometimes new expressions (VargasMoreno
et al., 2012). In 1978, Tillman (cited in Yin,
2011; VargasMoreno et al., 2012) developed
two new equations (the second derived from the first) to estimate the heating
value from the ultimate analysis for biomass fuels and suggested that the biomass
heating value has a very strong function in its carbon content (HHV = 0.4373C1.6701).
However, Table 1 shows some of the more common models that
have been used for coal and biomass. Demirbas et al.
(1997) (cited in VargasMoreno et al., 2012)
proposed two empirical equations to predict the heating values for the hydropyrolytic
oils of poplar trees and lignocellulose materials. The first one took into account
the elements C, H and O while the second equation took into account the same
elements plus N. A large quantity of biomass samples together with their higher
heating values and the basic analysis data were collected from the open literature
and used by Sheng and Azevedo (2005) to develop a series
of models, two of these models were based on the results of ultimate analysis.
Recently, proposal for a new model for a wide range of biomass materials are
published in the lierature (Friedl et al., 2005;
Thipkhunthod et al., 2005; Huang
et al., 2009; CallejonFerre et al.,
2011; Yin 2011).
Generally, it can be noted that, carbon elements were taken into account for all the published models that were based on the ultimate analysis, due to the fact that carbon has a direct impact on the heating value. Moreover, some of these models also took into account the element hydrogen and oxygen. Nitrogen and sulfur were accounted in a few models.
Table 1: 
Models based on the ultimate analysis for coal and biomass 

W: wt% of water, O*: Sum of the contents of the oxygen and
other elements (including S, N, Cl, etc.) in the organic matter 
MATERIALS AND METHODS Sample preparation: The biomass material (OPF) used for these experiments was obtained from different places in the state of Perak in Malaysia. The sample collection was carried out throughout a six month period from August 2011 to January 2012. The biomass samples were first dried naturally and then dried further in a dryer oven at 105°C for 24 h. After drying, the samples were ground and sieved to fine size of 0.212 and 0.6 mm.
Sample characterization: The characterizations of the samples included
the elemental analysis and determination of the higher heating value. The elemental
analysis to find the percentages of carbon (C), hydrogen (H), nitrogen (N) and
sulfur (S) content in the samples was carried out using LECO CHNS932 elemental
analyzer while the higher heating value was measured using the IKA Werke C5000
bomb calorimeter.
Heating value model: Biomass materials composed of the elements C, H,
O, N and S, where the first three elements represented about 9799% of the biomass
organic mass. The relationships between the higher heating values and ultimate
analysis components (C, H, O and N) are plotted in Fig. 1.
Sulfur was found in a very small amount (less than 0.17%) and therefore it was
not considered in the model. It can be observed from the same Fig.
1 that there is a linear relationship between the heating value with carbon,
hydrogen and oxygen. However, the heating value increased with an increase in
carbon and hydrogen content which agrees with the results of Sheng
and Azevedo (2005). The heating values decreased with an increase in oxygen
content and this correlates well with the fact that oxygen is not a reactive
element (Sheng and Azevedo 2005).

Fig. 1: 
Correlation between heating value of OPF and its ultimate
analysis content 
Table 2: 
Calculated and critical correlation coefficient for the biomass
components 

On the other hand, there is no trend observed between heating values and nitrogen
content. Therefore, in calculating the higher heating value of the OPF, only
three elements are selected which are carbon, hydrogen and oxygen. For the current
experimental data and confidence level of 0.95%, there are significant correlations
between the biomass heating value and the carbon, oxygen and hydrogen components;
for which the correlation coefficients (R^{2}) are 0.903, 0.817 and
0.523, respectively. On the other hand, there is no correlation between the
nitrogen components (0.026) as shown in Table 2. From the
same Table 2, it can be noted that the computed correlation
coefficients for carbon, oxygen and hydrogen are greater than the critical correlation
coefficients for the number of samples (Wheeler and Ganji,
2003).
The multiple linear regression analysis was adopted to develop an empirical equation to estimate the heating value of the OPF samples from the C, H and O percentages. The evaluations of the variables were also conducted at a significant confidence level of 0.95%. The procedure to create the equation and statistics analysis was developed by using spreadsheet software. The resulted empirical equation was next used to calculate the heating values which were then compared with those obtained experimentally. RESULTS AND DISCUSSION
Development of models: In the development of multiple linear regression
models, carbon (C), hydrogen (H) and oxygen (O) were taken as input independent
variables while the actual heating values are taken as an output dependent variable.
In this study, four empirical equations were proposed to determine the heating
values using the elements of the ultimate analysis. These equations were developed
based on the fact that, the heating value of biomass is a function of carbon,
carbon and hydrogen, carbon and oxygen and also as a function of carbon, hydrogen
and oxygen.
Table 3: 
Models based on the ultimate analysis elements 

AAE: Average absolute error, ABE: Average bias error 
However, carbon was included in the entire proposed models; this is because
of its nature as the major reactive element present in the biomass. The proposed
models are shown in Table 3.
Selection of the best model: In selecting the best model, all equations were evaluated and compared with the fitting performance criteria which evaluated the accuracy and validity of the model. In this study, the considered performance criteria were the correlation coefficient (R^{2}), Average Absolute Error (AAE) and Average Bias Error (ABE) which can be calculated as:
where, subscribes M and C denote the measured and calculated values for the
higher heating value, respectively while
is the measured average higher heating value of all the samples and n, is the
number of samples. The correlation coefficient (R^{2}) is used widely
in statistical and regression analyses to quantify the accuracy of the model.
The higher value of the correlation coefficient, indicates the better estimation
(A perfect model has an R^{2} of 1.0). The Average Absolute Error (AAE)
of the correlation indicates the correlation accuracy. The lower the AAE value,
the higher would be the accuracy of the correlation model. ABE describes the
average bias error of the correlation. A positive value of ABE means an overall
overestimation while a negative value indicates an overall underestimation
of the sample population. The smaller is the absolute value of ABE is, the smaller
would be the bias of the correlation. Table 3 shows the values
of the R^{2}, AAE and ABE for the proposed model equations. It can be
seen that the model M4 gives the maximum R^{2} and minimum AAE performance,
according to that, the best model of predicting the heating value of the OPF
by using the ultimate analysis could be:
Table 4: 
Comparison between measured and predicted heating values
of the OPFs 


Fig. 2: 
Comparison between measured and calculated heating values
of the oil palm fronds 

Fig. 3: 
Comparison of measured and predicted heating value 
Model validation: Validation of the model was carried out by comparing
the predicted heating values with those obtained experimentally. The predicted
and measured values are shown in Fig. 2 and 3.
It can be seen from the same Fig. 2 and 3
that the predicted values were quite consistent with the experimental measurements.
Figure 4 shows the scatter plot of the standard residuals
against the predicted heating values. It can be seen that the standard residuals
were randomly distributed along the predicted heating values and most of the
values lay within ±2.

Fig. 4: 
Standard residuals against the predicted heating values 
The developed model was also tested with six other samples of oil palm fronds
(not included in developing the model) to confirm the reliability of the model.
Table 4 shows the comparison between the measured and predicted
heating values for the tested samples. However, the developed model is capable
of predicting the heating value of oil palm fronds with an average absolute
error of less than 6%, indicating its good predicting capability. In addition,
the model was used to predict the heating value for the OPF sample reported
in the literature (Abdullah et al., 2010). The
measured and predicted heating values are presented in Table 4,
with an average absolute error of 6.42%, indicating its good applicability for
Malaysian oil palm fronds.
CONCLUSION
The multiple linear regression analysis was used to develop an empirical correlation
for the prediction of the heating value of oil palm fronds based on the ultimate
analysis. The predicted heating values calculated by the proposed correlation
showed good agreement with experimental values. However, the results showed
a correlation coefficient (R^{2}) of 0.92, with an Average Absolute
Error (AAE) and Average Bias Error (ABE) of 1.4 and 0.16%, respectively. Moreover,
the results showed that the predicted heating values were in agreement with
those obtained by the experimental measurements. Therefore, the developed model
may be acceptable for estimating the heating values of Malaysian oil palm fronds.
Correlations based on the proximate analysis and chemical analysis should be investigated in the future. ACKNOWLEDGMENT The authors would like to acknowledge the Universiti Teknologi PETRONAS for providing the financial support and research facilities for this study.

REFERENCES 
1: Akkaya, A.V., 2009. Proximate analysis based multiple regression models for higher heating value estimation of low rank coals. Fuel Proc. Technol., 90: 165170. CrossRef 
2: Wheeler, A.J. and A.R. Ganji, 2003. Introduction to Engineering Experimentation. 2nd Edn., Prentice Hall, New York, USA., ISBN13: 9780130658449, Pages: 464.
3: Demirbas, A., D. Gullu, A. Caglar and F. Akdeniz, 1997. Estimation of calorific values of fuels from lignocellulosics. Energy Sources, 19: 765770. CrossRef 
4: CallejonFerre, A.J., B. VelazquezMarti, J.A. LopezMartinez and F. ManzanoAgugliaro, 2011. Greenhouse crop residues: Energy potential and models for the prediction of their higher heating value. Renewable Sustainable Energy Rev., 15: 948955. CrossRef 
5: Demirbas, A., 2001. Relationships between lignin contents and heating values of biomass. Energy Conversion Manage., 42: 183188. CrossRef 
6: Friedl, A., E. Padouvas, H. Rotter and K. Varmuza, 2005. Prediction of heating values of biomass fuel from elemental composition. Anal. Chim. Acta, 544: 191198. CrossRef 
7: Gordillo, G. and K. Annamalai, 2010. Adiabatic fixed bed gasification of dairy biomass with air and steam. Fuel, 89: 384391. CrossRef 
8: Kawamoto, H., W.Z. Mohamed, N.I.M. Shukur, M.S.M. Ali, Y. Ismail and S. Oshio, 2001. Palatability, digestibility and voluntary intake of processed oil Palm fronds in cattle. Jap. Agric. Res. Quart., 35: 195200. CrossRef  Direct Link 
9: Huang, C., L. Han, Z. Yang and X. Liu, 2009. Ultimate analysis and heating value prediction of straw by near infrared spectroscopy. Waste Manage., 29: 17931797. CrossRef 
10: Kathirvale, S., M.N.M. Yunus, K. Sopian, A.H. Samsuddin and R.A. Rahman, 2003. Modeling the heating value of municipal solid waste. Fuel, 82: 11191125. CrossRef  Direct Link 
11: Li, J., Y. Yin, X. Zhang, J. Liu and R. Yan, 2009. Hydrogenrich gas production by steam gasification of palm oil wastes over supported trimetallic catalyst. Int. J. Hydrogen Energy, 34: 91089115. CrossRef 
12: Mohammed, M.A.A., A. Salmiaton, W.A.K.G. Wan Azlina, M.S.M. Amran, A. FakhrulRazi and Y.H. TaufiqYap, 2005. Hydrogen rich gas from oil palm biomass as a potential source of renewable energy in Malaysia. Renewable Sustainable Energy Rev., 15: 12581270. CrossRef 
13: Parikh, J., S.A. Channiwalab and G.K. Ghosal, 2005. A correlation for calculating HHV from proximate analysis of solid fuels. Fuel, 84: 487494. CrossRef 
14: Elneel, R., A. Shaharin and A. Bambang, 2011. Feasibility study of gasification oil palm fronds in an updraft gasifier. Proceedings of the 4th International AIP Conference, Meeting of Advances in Thermofluids, October 34, 2011, Malaka, Malaysia, pp: 972980.
15: Abdullah, S.S., S. Yusup, M.M. Ahmad, A. Ramli and L. Ismail, 2010. Thermogravimetry study on pyrolysis of various lignocellulosic biomass for potential hydrogen production. Proc. World Acad. Sci. Eng. Technol., 72: 129133. Direct Link 
16: Sulaiman, S.A., M.R.T. Ahmad and S.M. Atnaw, 2011. Prediction of biomass conversion process for oil palm fronds in a downdraft gasifier. Proceedings f the 4th International Meeting of Advances in Thermofluids, October 34, 2011, Melaka, Malaysia, pp: 10011010.
17: Sheng, C. and J.L.T. Azevedo, 2005. Estimating the higher heating value of biomass fuels from basic analysis data. Biomass Bioenergy, 28: 499507. CrossRef 
18: Shuit, S.H., K.T. Tan, K.T. Lee and A.H. Kamaruddin, 2009. Oil palm biomass as a sustainable energy source: A Malaysian case study. Energy, 34: 12251235. CrossRef  Direct Link 
19: Thipkhunthod, P., V. Meeyoo, P. Rangsunvigit, B. Kitiyanan, K. Siemanond and T. Rirksomboon, 2005. Predicting the heating value of sewage sludges in Thailand from proximate and ultimate analyses. Fuel, 84: 849857. CrossRef 
20: VargasMoreno, J.M., A.J. CallejonFerre, J. PerezAlonsoa and B. VelazquezMarti, 2012. A review of the mathematical models for predicting the heating value of biomass materials. Renewable Sustainable Energy Rev., 16: 30653083. CrossRef 
21: Yin, C.Y., 2011. Prediction of higher heating values of biomass from proximate and ultimate analyses. Fuel, 90: 11281132. CrossRef 



