摘要
In order to accurately identify the mathematical difference between the low contrast hydrocarbon reservoir and other reservoirs and precisely discover the low contrast hydrocarbon reservoir,the feature selection and classification models were introduced. They were applied to exploring the particularity of low contrast hydrocarbon reservoir.This method doesn’t require the calculated reservoir parameters to be fully equivalent to the practical reservoir characteristics.It will be workable as long as the numerical differences among different types of fluids can be depicted.Taking the productivity forecast of ultra-low permeability oil layer in the Longdong region of Ordos Basin as an example,its whole process was elaborated.The productivity model was utilized to predict the fracturing productivities of 77 new wells.The mean absolute error is 0.95 t/d and the mean relative error is 9.49%.And the accuracy rate is satisfactory.