APPLICATION OF NEURAL NETWORK TECHNOLOGY OPTIMIZED BY MULTIPLE SEISMIC ATTRIBUTES TO PREDICTION OF C81 SANDSTONE RESERVOIRS IN COAL SEAM AREA IN L BLOCK OF ORDOS BASIN

Seismic attribute values are usually dependant on some geological parameters, such as sedimentary characteristics, lithology and pore structures. However, as the seismic reflection is a kind of integrated reflection of a geological body, a single attribute parameter may have multiple solutions. On the basis of individual geological body, the Neural network technique optimized with multi-seismic attributes is used by the authors to predict the reservoir in JHER well area. Our results show that this technique may avoid the disadvantages of multiple solution for single attribute, enhance the resolution of multi-attribute discrimination, and thus improve the accuracy and efficiency of seismic reservoir prediction.