关键词: 贝叶斯网络, 信息融合, 聚类分析, 神经网络, 油藏预测

Abstract: A Bayesian fusion model is put foreword with a corresponding algorithm designed to forecast the oil reservoir distribution A clustering analysis algorithm, k-mean algorithm, is expanded and optimized to avoid errors due to data classifying in various professional fields, such as geology, logging and shaft drilling. Eventually, a conclusion is drawn the way the fusion center processes the output from Bayesian network as objective probability knowledge in combination with the subjective knowledge provided by experts in different fields. Experimental results showed that the method has successfully solved many problems which the conventional single-neuron network-modeling method was hard to resolve, such as difficult to design, long training cycles, low forecasting speed and inexact classifying results, thus meeting the requirements for such a forecast.