Optimization Of Seismic Attributes And Neural Networks Applied In Reservoir Prediction

The study of seismic attributes started in 1960's, and now have been applied widely after several developed steps. We can obtain more than ten seismic attribute parameters from the seismic information, but the more parameters don't mean the good results during the reservoir prediction. The invalid parameters not only add the workload and waste the limited resources, but also bring the dimension disaster. So we must optimize the parameters for effectively making use of these seismic attributes in order to predict reservoir parameters. The paper introduces some optimization methods of the seismic attributes, and uses the optimized results in reservoir prediction by analyzing principal components analysis , locally linear embedding and K-L transform optimization method.As one of the jumped-up nonlinear science, neural networks has powerful ability in nonlinear mapped. Theory researches and practicality applications of neural networks have been developed greatly for the recent years, and have applied widely in many fields and promoted neural networks. At the same time, these fields advances the high demands for the neural networks. The reservoir parameters prediction with wavelet neural networks is applied widely for its virtues, such as the strong anti-jamming , good fault-tolerant, relaxed request for parameters independency and the ability of studying vast seismic character parameters at one time. The paper introduces the wavelet neural networks theory in detail, and improves the study algorithm of the networks.The paper has a reservoir parameters prediction experiment with optimized seismic attributes and log curves by using back propagation networks and wavelet neural networks. The method obtains a nice effect.