One way to reduce the costs of investment in underground natural gas storage processes is to replace a portion of the base gas with a cheap gas. Optimizing the replacement amount of the base gas is one of the most attractive issues as mixing between the replaced base gas and the working gas and exceeding the injection pressure above the formation fracturing pressure are the two important phenomena that can occur and limit the amount of base gas replacement. In the present study, the concept of smart gas is applied to alleviate the encountered limitations during base gas replacement. Smart gas is referred to as the gas, the composition of which is adjusted so as to obtain the maximum amount of the base gas replacement. In the present work, dry air is used as the candidate gas to study the base gas replacement process in a depleted gas reservoir. Due to the fact that execution of a large-scale compositional reservoir simulator incorporated with an optimization algorithm is a time-consuming process, in order to calculate the optimum composition of the candidate gases, a proxy model is applied as a computationally inexpensive alternative to full numerical simulation. The results reveal that the composition of CO 2 in air is an important parameter in controlling mixing between the candidate base gas and the working gas, while maintaining the injection pressure below the fracturing pressure of the formation. In addition, when the CO 2 composition exceeds a specific value (40.95% in the modified air), the pressure that is required for gas production at a target rate cannot be supplied. The optimization of the CO 2 composition in the candidate gas employing the proxy model shows that it is possible to replace 28.4% of the base gas and reach an enhanced gas recovery of about 18.55% in the reservoir under study using the optimized CO 2 composition of 20.08% in the modified air as the replacing gas.