Traditionally, oil well formation tests aim to characterize the reservoir permeability field from pressure transient analysis (PTA) of drawdown and build up based on isothermal flow models in porous media. With the advancement of well test instrumentation, more accurate temperature records became available and have encouraged researches based on non-isothermal models that made possible the temperature transient analysis (TTA). In addition to the characterization of reservoir parameters such as permeability and porosity by TTA, the pressure data obtained from a non-isothermal model represent better the physical phenomenon, especially when the tests are subjected to greater drawdowns. This work consists in the development of a simulator for formation test that considers non-isothermal modeling of a unidimensional radial reservoir coupled to a production well and in the use of this simulator, associated with multivariable optimization methods, to solve the inverse problem of reservoir parameters characterization. Some optimization methods were tested and the Nelder-Mead Simplex algorithm presented better efficiency. Three types of problems were established and used in three hypothetical cases, including artificially imposed noise in pressure and temperature signals used to solve the inverse problem.