Linear Programming (LP) methods have previously been proposed for the optimization of underground storage fields. As the name implies, these methods necessitate many assumptions for the linearization of the constraints and the objective function, reducing accuracy of the results. A new approach for field optimization is developed that eliminates these assumptions. Genetic Algorithm (GA) is used as the optimization tool. A population consisting of individuals that encode a set of well rates for the field is created and evolved into new generations through a stochastic though structured algorithm that models some natural phenomena. The three-dimensional finite-difference model of the field is used as the GA＇s evaluation function. Individuals that return higher field cumulative production are given higher chance to reproduce. Individuals that cause a flowing well pressure lower than the minimum allowable are penalized. It is observed that the LP method gives significantly different results than those of the GA approach because of the assumptions it bears.