There are different models which are used in analyzing ENERGY DEMAND. Some of them are designed only' FOR analyzing ENERGY DEMAND, while some others are not designed specifically FOR this purpose, but are used in the analyses. Structural and non-structural econometric models and end use models are the main models FOR analyzing ENERGY DEMAND. Model FOR ENERGY DEMAND Evaluation (MEDEE) is one of the famous end use models which are used only FOR analyzing ENERGY DEMAND. Stun linear expenditure system, trans log DEMAND system, the constant elasticity DEMAND system, Rotterdam DEMAND system, almost ideal DEMAND system, autoregressive process, moving average process, autoregressive moving average process, autoregressive integrated moving average process, vector autoregressive model and vector error correction models; are some of the famous structural and non-structural econometric models which are used in analyzing ENERGY DEMAND.In this paper, among the above models, vector error correction model (VECM) because of its unique characteristics has been used FOR dynamic analysis of ENERGY DEMAND in Iran. In this regard three separate models FOR studying DEMAND FOR oil products, electricity and natural gas have been FORmulated. Using Eviews and Microfit software and time series of RGDP, real price of ENERGY carriers and physical value of DEMAND FOR ENERGY carriers in 1959-2004, the FORmulated vector error correction model has been estimated. The Johansen method shows that there is one co integration vector between the variables entered in the models of DEMAND FOR oil - products and electricity; but there is no dynamic long - run equilibrium relationship in the VECM of natural gas. In the VECM of oil products, all estimated coefficients are statistically significant, and the error correction term has the correct sign. In this model, the adjustment speed of deviation from the equilibrium relationship is slow. According to coefficients in the VECM of electricity, the long run price and income elasticity of DEMAND FOR electricity are - 0.86 and 1.84, respectively. It is worth mentioning that in this research by short run we mean periods less than one year, and long run refers to periods more than one year until there will be no structural change in relations among variables.