TY - JOUR
T1 - Precise prediction of biogas thermodynamic properties by using ANN algorithm
AU - Farzaneh-Gord, Mahmood
AU - Mohseni-Gharyehsafa, Behnam
AU - Arabkoohsar, Ahmad
AU - Ahmadi, Mohammad Hossein
AU - Sheremet, Mikhail A.
PY - 2020/3
Y1 - 2020/3
N2 - There are technical problems related to storage and transport of biogas gas that should be addressed before practical injection of these fuels into the existing natural gas networks. In addition, their different final applications resulting in the presence of various components and in various concentrations make the problem harder. Therefore, it is indispensable for designers of the pipeline network to know exactly what the thermodynamic properties of a gas mixture are, especially its density, which would vary a lot. In this work, a MLP (Multi-layer Perceptron) neural network is used for the development of the desired biogas properties predictor model. In order to train the network, the biogas thermodynamic properties created using ISO 20765-2 (2015) (where applicable) and experimental values are employed. Results are compared with the values estimated from the GERG2008 equations of state, which are the reference equations for natural gases and experimental values. The results indicate that the developed MLP model presents a high accuracy in the calculations over a wide range of biogas mixtures and input properties ranges for all the output properties including density, compressibility factor, isochoric heat capacity, isobaric heat capacity, isentropic exponent, internal energy, enthalpy, entropy, Joule-Thomson coefficient, and speed of sound. The Root Mean Square Error (RMSE) of the mentioned properties of test data are 0.00012536, 0.00016593, 0.0025213, 0.0016208, 0.00337, 0.0096329, 0.0099837, 0.0035625, 0.01055, and 0.00039428 respectively.
AB - There are technical problems related to storage and transport of biogas gas that should be addressed before practical injection of these fuels into the existing natural gas networks. In addition, their different final applications resulting in the presence of various components and in various concentrations make the problem harder. Therefore, it is indispensable for designers of the pipeline network to know exactly what the thermodynamic properties of a gas mixture are, especially its density, which would vary a lot. In this work, a MLP (Multi-layer Perceptron) neural network is used for the development of the desired biogas properties predictor model. In order to train the network, the biogas thermodynamic properties created using ISO 20765-2 (2015) (where applicable) and experimental values are employed. Results are compared with the values estimated from the GERG2008 equations of state, which are the reference equations for natural gases and experimental values. The results indicate that the developed MLP model presents a high accuracy in the calculations over a wide range of biogas mixtures and input properties ranges for all the output properties including density, compressibility factor, isochoric heat capacity, isobaric heat capacity, isentropic exponent, internal energy, enthalpy, entropy, Joule-Thomson coefficient, and speed of sound. The Root Mean Square Error (RMSE) of the mentioned properties of test data are 0.00012536, 0.00016593, 0.0025213, 0.0016208, 0.00337, 0.0096329, 0.0099837, 0.0035625, 0.01055, and 0.00039428 respectively.
KW - Biogas
KW - Thermodynamic properties
KW - Artificial neural network
KW - Multilayer perceptron
KW - GERG2008 equation of state
UR - http://www.scopus.com/inward/record.url?scp=85071649859&partnerID=8YFLogxK
U2 - 10.1016/j.renene.2019.08.112
DO - 10.1016/j.renene.2019.08.112
M3 - Journal article
SN - 0960-1481
VL - 147
SP - 179
EP - 191
JO - Renewable Energy
JF - Renewable Energy
IS - 1
ER -