Prediction of CO2 equilibrium moisture content using least squares support vector machines algorithm
Ghiasi, MM, Abdi, J, Bahadori, M, Lee, M & Bahadori, A 2016, 'Prediction of CO2 equilibrium moisture content using least squares support vector machines algorithm', Petroleum and Coal, vol. 58, no. 1, pp. 27-46.
The burning of fossil fuels such as gasoline, coal, oil, natural gas in combustion reactions results in the production of carbon dioxide. The phase behavior of the carbon dioxide + water system is complex topic. Unlike methane, CO2 exhibits a minimum in the water content. These minima cannot be predicted by existing methods accurately. In this communication, two mathematicalbased procedures have been proposed for accurate computation of CO2 water content for temperatures between 273.15 and 348.15 K and the pressure range between 0.5 and 21 MPa. The first is based on least squares support vector machine (LSSVM) algorithm and the second applies multilayer perceptron (MLP) artificial neural network (ANN). Furthermore, the constants of the previously developed empirical correlation have been re-tuned. Statistical error analysis has been utilized to evaluate the adequacy and accuracy of the novel models and empirical correlation. It was found that the predictions of the presented intelligent models and the empirical correlation are in excellent agreement with reported data in the literature with the absolute average relative deviations percent (%AARDs) of generally less than 0.9 % and R2 of generally greater than 0.999. However, using the LSSVM model contributes to obtain slightly better results.