Abstract: Tahe oilfield is a crack-cave carbonate reservoir with extremely serious heterogeneity. It is very difficult to recognize this type of reservoir only according to static behavior data. A new idea is put forward that this type of reservoir could be described by using production per-formance data or information. Artificial neural network technique is advantageous for dealing with and predicting the non-linear correlative pa-rameters. This paper, based on percolation theory and integrated with well test result, presents the relations between formation factor and pro-duction performance information, develops the structural model for reservoir parameter prediction by artificial neural network. The field study shows that this technique is of more practical value in the proper recognition of Tahe field.

Key words: Taha oilfield, neural network, carbonate rock, reservoir, description, characterization