碳酸盐岩储层受沉积、构造及成岩作用影响大,测井曲线不能完全反映岩性性质.分析常规交会图法在碳酸盐岩储层岩性识别中整体识别率偏低的原因.通过对比决策树、人工神经网络、支持向量机和贝叶斯网络等数据挖掘方法,发现决策树具有较高的识别正确率.采用该方法分析常规测井数据,通过构造新参数Rd/AC和Rd/Rs进行碳酸盐岩岩性识别,准确率可达到89.32％.实际应用表明,地质、测井知识和数据挖掘方法相结合,能有效识别常规方法无法准确判断的岩性,通过改进不同的优选参数,提高岩性识别的准确率,为储层沉积相解释提供更准确的地质信息.%Well logging curves can not reflect carbonate rock reservoir properties completely for great influences from deposition, construction and diagenesis. Analyzed is the cause of low recognition rate of traditional cross plot method in carbonatite reservoir lithology recognition. By comparing data mining methods such as decision tree, artificial neural network, support vector machine and Bayes network, it is found that the decision tree method has superior recognition rate. With the decision tree method, identification accuracy achieves 89. 32% with constructing new parameters Rd/AC and Rd/Rs. Practical application shows that combining data mining with geology and well logging can not only effectively identify the lithology that couldn't be identified by normal methods, but also increase the accuracy by improving optimum parameters, which provides more accurate geological information for reservoir sedimentary phase assessment and explanation.
展开▼
