Modelos de previsão de demanda de derivados de petróleo

Abstract

Making demand forecasts is essential to help determine the necessary resources to any company including oil refineries. In this work, it has been made short term forecast analysis for oil production and consumption using time series. The goal is to test different methods to each derivate in order to determine the accuracy of each one. Among the methods, Winters turned to be the most accurate, followed by Exponential Smoothing. However the difference in accuracy between the best method and the others was often below two percentage points. The forecast for production using national data was more accurate than the ones using regional data, showing that the national production is more systematic and less random than the regional one. The forecast for consumption was very accurate in both national and regional data, showing that fuel distributors can use these methods in its planning.