MODELING OF CO2 SOLUBILITY IN METHYLDIETHANOLAMINE + PIPERAZINE AQUEOUS SOLUTIONS USING NEURAL NETWORK MODEL

Acid gas removal processing from natural gas, synthesis gas and refinery gas
stream is very important in plant industry to prevent corrosion in the subsequent piping
and as per required by users. Generally, the gaseous feedstock stream is contacted with
alkanolamine aqueous solution countercurrently in the acid gas absorber tower. The
acid gas will partly be converted to non-volatile ionic species by the basic amine and
partly be dissolved physically in the liquid solution. Acid gas removal unit is very
important in petrochemical processing plant.
There had been great numbers of studies on the solubility of CO2 in
methyldiethanolamine (MDEA) aqueous solutions. M. Vahidi used extended Debye-
Huckle model to explain on the solubility data of CO2 absorption in MDEA and
Piperazine (PZ) at certain concentration, temperature and pressure. The average
absolute relative deviation percent (AAD %) reported were 8.11%.
Model of solubility of CO2 in MDEA aqueous solution using neural network
model is presented. In this work, neural network was used to determine solubility of
CO2 in MDEA + PZ solutions based on 3 experimental research data, M. Vahidi
(Mehdi Vahidi, 2009), H. B. Liu (Hua-Bing Liu, 1999) and B. Si Ali (B. Si. Ali, 2004).
Data prediction was conducted at different temperature and partial pressure and for
various solution concentrations. The model developed has an absolute relative deviation
of 3.047% compared to other papers; M. Vahidi with 8.11% deviation, B. Lemoine of
7.84% and H. B. Liu with 11.6% error deviation. The model also not capable to
extrapolate prediction of CO2 loading on zero promoter system due to the experimental
data at zero promoter is not included in the training process.