The low cost of K2CO3/Al2O3 adsorbent is encouragement to use it for CO2 capture from the flue gas of fossil fuel power plants. In this study, optimization of the CO2 capture process using a K-based adsorbent in a fixed-bed reactor has been investigated. The sorbent was also characterized by different techniques such as SEM, BET, and XRD analysis before and after the reactions. Response surface methodology (RSM) combined with Box-Behnken design (BBD) was employed to evaluate the effects of the process variables (temperature, mole ratio of H2O/CO2, and vapor pretreatment time) and their interaction on the responses (CO2 capture capacity and deactivation rate constant) to achieve the optimal conditions. In addition to the experiments, the deactivation model in the noncatalytic heterogeneous reaction system was employed to evaluate the kinetic parameters (sorption rate and deactivation rate constants) using nonlinear-least-squares technique. According to the analysis of variance (ANOVA), the vapor pretreatment time and temperature were found to be the most important process variables which affect the CO2 adsorption capacity. Moreover, two quadratic semiempirical correlations were established to calculate the optimum operating conditions of the CO2 capture process. The predicted values of the correlations showed very good agreement with the experimental data. The optimum process variables obtained from the numerical optimization corresponded to 61.3 degrees C, 1 and 9 min for the adsorption temperature, mole ratio of H2O/CO2, and vapor pretreatment time, respectively. Based on the optimal condition, the highest adsorption capacity of 87.71 mg of CO2/g of sorbent in 100% CO2 removal zone (corresponding to 97.82% of theoretical adsorption capacity in the total zone) and the lowest deactivation rate constant of 0.1872 min(-1) were obtained. Furthermore, additional experiments performed in the optimal conditions resulted in 86.97 mg of CO2/g of sorbent adsorption capacity and a deactivation rate constant of 0.1874 min(-1). The results indicate that the presented models could adequately predict the responses and provide suitable information for the process scale-up.