Kinetics of Slurry Phase Fischer-Tropsch Synthesis

Abstract

The overall objective of this project is to develop a comprehensive kinetic model for slurry-phase Fischer-Tropsch synthesis (FTS) employing iron-based catalysts. This model will be validated with experimental data obtained in a stirred-tank slurry reactor (STSR) over a wide range of process conditions. Three STSR tests of the Ruhrchemie LP 33/81 catalyst were conducted to collect data on catalyst activity and selectivity under 25 different sets of process conditions. The observed decrease in 1-olefin content and increase in 2-olefin and n-paraffin contents with the increase in conversion are consistent with a concept that 1-olefins participate in secondary reactions (e.g. 1-olefin hydrogenation, isomerization and readsorption), whereas 2-olefins and n-paraffins are formed in these reactions. Carbon number product distribution showed an increase in chain growth probability with increase in chain length. Vapor-liquid equilibrium calculations were made to check validity of the assumption that the gas and liquid phases are in equilibrium during FTS in the STSR. Calculated vapor phase compositions were in excellent agreement with experimental values from the STSR under reaction conditions. Discrepancies between the calculated and experimental values for the liquid-phase composition (for some of the experimental data) are ascribed to experimental errors in the amount of wax collected from the reactor, and the relative amounts of hydrocarbon wax and Durasyn 164 oil (start-up fluid) in the liquid samples. Kinetic parameters of four kinetic models (Lox and Froment, 1993b; Yang et al., 2003; Van der Laan and Beenackers, 1998, 1999; and an extended kinetic model of Van der Laan and Beenackers) were estimated from experimental data in the STSR tests. Two of these kinetic models (Lox and Froment, 1993b; Yang et al., 2003) can predict a complete product distribution (inorganic species and hydrocarbons), whereas the kinetic model of Van der Laan and Beenackers (1998, 1999) can be used only to fit product distribution of total olefins and n-paraffins. The kinetic model of Van der Laan and Beenackers was extended to account separately for formation of 1- and 2-olefins, as well as n-paraffins. A simplified form of the kinetic model of Lox and Froment (1993b) has only five parameters at isothermal conditions. Because of its relative simplicity, this model is well suited for initial studies where the main goal is to learn techniques for parameter estimation and statistical analysis of estimated values of model parameters. The same techniques and computer codes were used in the analysis of other kinetic models. The Levenberg-Marquardt (LM) method was employed for minimization of the objective function and kinetic parameter estimation. Predicted reaction rates of inorganic and hydrocarbon species were not in good agreement with experimental data. All reaction rate constants and activation energies (24 parameters) of the Yang et al. (2003) model were found to be positive, but the corresponding 95% confidence intervals were large. Agreement between predicted and experimental reaction rates has been fair to good. Light hydrocarbons were predicted fairly accurately, whereas the model predictions of higher molecular weight hydrocarbons values were lower than the experimental ones. The Van der Laan and Beenackers kinetic model (known as olefin readsorption product distribution model = ORPDM) provided a very good fit of the experimental data for hydrocarbons (total olefins and n-paraffins) up to about C{sub 20} (with the exception of experimental data that showed higher paraffin formation rates in C{sub 12}-C{sub 25} region, due to hydrocracking or other secondary reactions). Estimated values of all model parameters (true and pseudo-kinetic parameters) had high statistical significance after combining parameters related to olefin termination and readsorption into one (total of 7 model parameters). The original ORPDM was extended to account separately for formation of 1- and 2-olefins, and successfully employed to fit experimental data of three major groups of hydrocarbon products (n-paraffins, 1-olefins and 2-olefins). This model is referred to as an extended ORPDM (8 model parameters in its final form). In general, all three groups of products were fitted well, and the estimated model parameters were all positive and the corresponding 95% confidence intervals were small. Even though the extended ORPDM provided a very good fit of experimental data, it can not be used for the prediction of product distributions for a given set of process conditions. This model has several pseudo-kinetic parameters whose values vary with process conditions. Additional work is needed to expand capabilities of the model to predict molar flow rates of all inorganic species and major hydrocarbon products in terms of true kinetic (temperature dependent) constants.