This paper focuses on the unknown problems of the well testing interpretation method for CO2 flooding in low permeability oil reservoir. Based on the percolation theory of multi-region composite model，an improved model of well testing interpretation is developed，which is solved by the numerical differential method. In this model，the threshold pressure gradient，pressure sensitivity effect and fluid dynamic property during CO2 flooding in the low permeability oil reservoir are considered. Meanwhile，Simultaneous Perturbation Stochastic Approximation（SPSA）optimum searching algorithm is conducted to fit pressure curve and pressure derivative curve simultaneously. Finally，automatic matching and parameter interpretation method of well testing for CO2 flooding in the low permeability oil reservoir is proposed. The variation of interpretation parameters obtained from practical well testing curves is analyzed. The results show that consideration of low permeability characteristics would have an obvious influence on the interpretation of reservoir permeability，and the interpreted permeability after optimization is increased by 37.97% compared with the initial value. On the contrary，the interpreted pressure transmitting coefficient and compressibility variation index are decreased by 16.94% and 21.97% respectively from the original interpretations，which are considered to be influenced seriously by fluid heterogeneity. The applications reveal that the proposed interpretation method is easier to operate，and has good feasibility.