The performance predictive target is always too optimistic using the LIT equation and material balance equation for the supply, multi-layer and other non-homogeneous low-permeability gas reservoir. To solve this problem, the project intends to start from the theory of pressure transient testing to establish different types of non-homogeneous gas reservoir performance prediction theoretical model, to explore optimization algorithm. The overall objective is to establish a more realistic dynamic prediction theory. It can improve the early prediction results for non-homogeneous gas reservoir, and this theory is considered non-homogeneous gas layer or regional non-homogeneous variables, but also simple and quick to predict the dynamic change of pressure and gas rate. New method is amended by numerical simulation. A single-well radial/composite model is constructed with measured rock and fluid properties. This compositional simulator is used to generate synthetic pressure transient data which are subsequently interpreted in well testing package. The impacts of test time and reservoir heterogeneity on prediction results are investigated. Through theoretical analysis and numerical simulation, the mechanism for gas flow, the pressure, the output rate variation over time is analyzed. A low permeability gas wells are analyzed by new method, thus improving long-term well performance predictions.