Hydraulic flow units (HFUs) prediction using integrated probabilistic neural network (PNN) and seismic attributes: A case study in Ekofisk Formation, North Sea

This study presents an algorithm of PNN that is used to predict the distribution of hydraulic flow units within a hydrocarbon field using seismic dataset, well data, and core data. Ekofisk formation, a promising hydrocarbon potential in the North Sea, is selected to apply and analyze the research methodology to predict the distribution hydraulic flow units through the available dataset. Flow zone indicator (FZI), a good parameter to describe the hydraulic flow units (HFUs), was analyzed using well logs Formation and showed the existence of at least four HFUs in Ekofisk. Based on FZI calculations, HFU#1 lies between 0.1 and 0.57, while HFU#2, HFU#3, and HFU#4 reports the following corresponding ranges 0.57 - 1.05, 1.05 - 1.52 and 1.52 - 2. HFU#3 and HFU#4 appear as the efficient productive conduits. Seismic inversion is utilized to provide good stratigraphic interpretation to Ekofisk Formation through the entire seismic cube and best define its boundary. A multi-attribute analysis finds out that average frequency, integrate, apparent polarity, and absolute integrate amplitude are the best seismic attribute to predicts flow zone indicators. PNN) is created a mathematical relationship to predict the distribution of various flow zone indicators throughout the seismic cube utilizing these four seismic attributes. Well F03-3 data, not used in training the PNN, is used to evaluate the performance of HFU prediction and showed high correlation between HFUs calculated at well site using log data and the corresponding predicted HFUs derived from seismic attributes. The developed model showed the detailed volumetric distribution of various HFUs in the Ekofisk Formation and other important geologic features can be investigated.