Liquefied natural gas (LNG) plays an important role in the country s natural gas and electricity supply chain. In Brazil, the imported product primarily targets the supply of thermoelectric demand for this input. The thermoelectric generation model is flexible and the main characteristic of this model is the need to guarantee a flexible fuel supply for plants that must meet an equally flexible thermoelectric demand. The objective of this work is modeling and evaluate the performance of a causal dynamic regression model in long-term LNG demand forecasts, for supplying of natural gas thermoelectric plants and to compare the predictive capacity with traditional models widely discussed in the literature, such as the ARIMA and the dynamic regression model. Scenarios of the explanatory variable were established in order to evaluate how the explanatory variables influences LNG demand. The product developed in this dissertation showed the viability of causal models to be used as a tool for the simulation of LNG demand considering scenarios of the causal variables. What if analysis.