In order to use 3D seismic data for productivity index prediction of alluvial fan coarse-grained clastic reservoirs with low porosity and low permeability, we chose the Triassic Baikouquan Formation reservoir in Y-region at the northwestern margin of Junggar Basin as a case. Based on traditional reservoir prediction such as thickness and porosity of coarse-grained clastic reservoirs, we analyzed in detail factors influencing permeability and oil-bearing properties, and introduced the average monthly production data at the early stage of development into 3D seismic inversion to predict the productivity index. The following work flow was established: `finding coarse-grained clastic reservoirs through typical curve inversion, finding high quality coarse-grained clastic reservoirs through porosity inversion, finding permeable coarse-grained clastic reservoirs through spontaneous potential inversion, finding oil-bearing coarse-grained clastic reservoirs through resistivity inversion, and predicting reservoir productivity index with the combination of several methods'. A productivity index cube was finally generated through Neural Network modeling by using the monthly productivity as hard data and wave impedance, porosity, resistivity, spontaneous potential inversion data cube and time domain structure as trai-ning samples. The result shows that there is a positive correlation (R2=0.948 7) between the predicted monthly productivity and initial average monthly production. For wells with an initial monthly average production more than 300 ton, the error of prediction is less than 10%. The data cube contains various information controlling hydrocarbon distribution, such as lithology, reservoir property, oil-bearing property and permeability. The oil production of three appraisal wells deployed based on this research in Y-region reached more than 5 ton per day, which verified the accuracy and practicability of this productivity index prediction technology.
