The purpose of this paper is to forecast petroleum consumption in Brazil for the year 2000 based upon logistic models, learning models, and translog models using the technique of intensity of energy use. The models employ a time series of 30 years for projection. An investigation of the evolution of petroleum consumption profile was made based upon three characteristic effects: structural, content and scale effects. Evaluation of forecasting models presented good results, with the translog model showing the best performance in terms of accuracy. The learning and translog models indicated that GDP is the main determinant for petroleum consumption evolution in the future, defining a range of 64 000 and 109 000 thousand of tonnes of oil equivalent on two defined GDP growth scenarios.219958968