Today, the oil energy as one of nonrenewable energy sources, has allocated an important place among the world's energy resources. The price structure of this goods and modeling has always pay attention economic research and attempts to evaluate and predict the volatility has been done. Iran is one of the largest oil exporters in the world and according to experts, the country's economy is dependent on oil revenues, therefore modeling and prediction of iran's crude oil price volatility is very important. In many time series, especially financial time series, features clearly observed heteroskedasticity. In this regard, various ltr" The data used in this study is the data of heavy crude oil per day, five days a week during the first third of january 2002 to november 2011. To study the behavior of crude oil price and a comprehensive understanding of the models, GARCH, IGARCH, GARCH-M, EGARCH, GJRGARCH, APARCH and FIGARCH is used for modeling. The results indicate that the effect of oil price shocks on the volatility of crude oil prices is asymmetric and has a high degree of stability and negative shocks than positive shocks have a greater effect on the conditional variance of oil prices. o the series of heavy crude oil price returns has the leverage effect. Crude oil price return series has a long memory and volatility in the form of conditional variance models are better modeling than the form of the conditional standard deviation. The results of survey the relationship between the series returns and conditional variance series show that the changes of conditional variance series has no significant effect on the return series, but the reverse relationship is true. In this study also examines the characteristics of the distribution of return series in performance evaluation GARCH (1,1) model and four standard normal distribution, T-student distribution, generalized error distribution (GED) and skew T-student distribution are also being studied. The results of performance evaluation models using loss functions criteria and DM test indicate that there are in terms of excessive kortusis and skewness in the return series using skew T-student distribution as the distribution of the error values has better forecast accuracy compared to other distributions