In order to understand the effects of activated carbon pore structure and adsorbed compound property on adsorption efficiency, the adsorption characters of pure acetone gas on activated carbon and the adsorption efficiencies of activated carbons of different structure to carbonyls in smoke were determined. Langmuir model and D-R model were respectively used to fit the data of gaseous acetone adsorption on activated carbon, and these two models were compared in terms of fitting precision and adsorption heat prediction. The relationships of model parameters with adsorption efficiency, with activated carbon structure and adsorbed compound property were further analyzed. The results showed that: 1) Comparing with Langmuir model, D-R model gave higher correlation coefficient, lower average relative standard deviation and better fitting results for the data of pure acetone gas adsorption on activated carbon. 2) The theoretical adsorption heat of pure acetone gas on activated carbon calculated by 10-4-3 potential function was 17.9 kJ/mol, the smaller adsorption heat indicated that it was mainly caused by physical adsorption. The adsorption heat predicted by D-R model was 15.8 kJ/mol, which was more close to the theoretical calculated value. The adsorption heat predicted by Langmuir model was 40.7 kJ/mol, which was much higher than the theoretical calculated value. 3) The key to predicting the adsorption efficiency of activated carbon to a compound lay in the prediction of adsorption heat of that compound. The factors affecting adsorption efficiency were mainly adsorption temperature, carbon amount used, pore volume, molecular weight of compound, collision diameter and energy parameters. The analysis of adsorption heat was helpful in investigating the effects of pore size on adsorption efficiency and adsorption selectivity.