Reserves are the main assets responsible for valuation of oil companies in the world market. However, to set any volume of hydrocarbon as reserves, it is necessary to develop a business plan to exploit and commercialize all oil and gas available. Therefore, the prior knowledge of the volume of hydrocarbons becomes strategic to anticipate the development and facilitate the acquisition of investments. The goal of this dissertation is to propose a methodology for reserves forecasting through improvements in the methods proposed by Knoring et al. (1999) and Barcelos (2006). The first improvement is related to optimize the fit of the proposed models for recoverable volume time series via an iterative method to provide initial solutions to the nonlinear least squares estimation method. The second improvement occurs in the process of choosing the best model avoiding the whole process needs to be revised after inserting new data to historical series each year. To analyze the effectiveness of the proposed work, the method is applied to seven recoverable volume time series and compared with the current methodology of reserves forecasting used by the company sponsoring the study.