Fuzzy neural network-based tight sandstone reservoir inversion - A case study from the Denglouku Formation in Changling 1 gas field

Fuzzy neural network reservoir inversion can avoid the integration of data of different types and scales and well preserve the data integrity, making it possible to reveal both the overall variation tendency and the local detailed features. In view of the rapid facies change and poor lateral continuity of the Denglouku Formation sandstone in the Changling1 gas field, we integrated various data including seismic, well logging, drilling and well testing and performed inversion of the sandstone thickness and porosity characters by using the fuzzy neural network theory. The following results were obtained. (1) The areal distributions of the Denglouku Formation sandstones in the Changling1 gas field varies greatly, with the thicker sandtone occurring mainly in areas along the Changshen 103, Changshen 1-3, Changshen 1-1 wells and the Changshen 2 well. The thickness of sandstone decreases progressively in northeast direction. The thickness of D3 layer near in the Changshen 103 well is up to 33.2 m. (2) The major D3 and D4 pay zones in the Denglouku Formation have a relatively low porosity, averaging at only about 5%. The sandbodies with relatively high porosity (>9%) only occur in area around the Changshen 1 well. The Denglouku Formation is tight as a whole. (3) Comparison of the inversion results with the actual test data shows that the error of sandstone thickness is within 2.5 m and the error of the porosity is under 0.49%, indicating a high reliability of the inversion results. It is concluded that fuzzy neural network-based inversion of tight sandstone reservoirs can well reveal sandstone distribution and reservoir properties.