Neural Network Analysis of 3-D Seismic Data Delineates Slope- to Basin-Floor Fan Bodies in a High-Frequency Falling-Stage Systems Tract, Matzen Field, Austria

Abstract

Abstract Automatic trace shape recognition via neural network analysis was utilized on a 70-km2 3-D seismic data volume from the giant Matzen field, Vienna Basin, Austria, to identify depositional facies in slope and basin-floor settings. Seismic wave shape at key wells was used for establishing links between seismic response and facies. Continuity analysis, proportional horizon slicing, voxel-body analysis, and highfrequency stratigraphic analysis involving an additional 50+ wells were integrated to the neural network results. This methodology can rapidly isolate and highlight reservoirquality facies and clarify facies interrelationships. When carried out in a high-frequency stratigraphic framework, facies style and resulting internal heterogeneity are more accurately predicted. Utilizing the wave-shape analysis with different windows above the zone of interest made it possible to discriminate between basin-floor fan sandstone, shelf-edge delta/slope fan sandstone, and slope and basin mudstone in the falling-stage systems tract of the overlying depositional cycle. Introduction Rapid, semiautomated seismic interpretation techniques such as attribute and continuity analysis have significantly decreased interpretation time and improved the accuracy of subsurface prediction, resulting in more accurate risk analysis and greater drilling success. An additional tool recently made available to interpreters is artificial neural network analysis, an outgrowth of investments in military technology. Neural networks applied to 3-D seismic data have the potential to quickly assess stratal architecture and depositional facies in a large subsurface volume. To evaluate this technique, we applied this technology to the detection of sand distribution, depositional geometries, and facies within major oil reservoirs of the Matzen field, Vienna Basin, Austria, the largest oil and gas field in Central Europe. The analysis was conducted within a high-frequency stratigraphic and depositional framework established using more than 1,400 wells and 70 km2 of high-quality modern 3-D seismic data. Neural network results were compared with more traditional interpretation techniques including attribute and continuity analysis, proportional horizon slicing, and voxel-body analysis. The research presented here focuses on slope-and basin-floor fan deposits in a middle Miocene high-frequency falling-stage unit that represents future exploration potential. Geologic Setting and Stratigraphy Vienna Basin is a shallow Miocene to Holocene pull-apart basin underlain by the thrust belt of the Cretaceous Alps. The basin was formed by lateral strike-slip motion within the convergence zone parallel to the southeastern margin of the Bohemian Massif (Fig. 1). The giant Matzen field is situated on a large horst block near the center of the basin and consists of a gentle NE-SW-trending anticline bounded on the NW and SE by normal faults with apparent left-lateral displacement (Fig. 2). To date more than 70 million tons of oil and 28 trillion cubic meters of gas have been produced over 50 years. The current estimate of recoverable hydrocarbons is 76 million tons of oil and 40 trillion cubic meters of gas. The production in Matzen field comprises 400 production units distributed among 70 horizons.