基于机器学习和测井数据的碳酸盐岩孔隙度与渗透率预测

准确预测碳酸盐岩储层孔隙度和渗透率对于碳酸盐岩油气藏储层评价具有重要意义。碳酸盐岩储层裂缝与溶孔广泛发育，基于经验公式从测井曲线预测储层孔隙度和渗透率具有较大误差。以中东某碳酸盐岩油藏为研究对象，选取914块取心井岩心，测定孔隙度与渗透率，利用随机森林（RF）、K-近邻（KNN）、支持向量机（SVM）和长短期记忆网络（LSTM）4种不同机器学习方法，通过测井数据进行孔隙度与渗透率预测，优化机器学习参数，筛选出适用于碳酸盐岩油藏的测井孔隙度与渗透率预测方法。研究结果表明：4种机器学习方法预测储层孔隙度结果差异不大，通过调整输入参数种类，可进一步提高孔隙度与渗透率预测效果，当以补偿中子（NPHI）、岩性密度（RHOB）和声波时差（DT）3种测井参数数据作为输入时，基于LSTM的储层孔隙度预测精度最高，孔隙度预测结果均方根误差（RMSE）为4.536 2；由于碳酸盐岩储层的强非均质性，基于机器学习的测井储层渗透率预测效果较差，相对而言，仅以NPHI作为机器学习输入参数时，基于RF的储层渗透率预测精度最高，渗透率预测结果的RMSE为45.882 3。

Prediction of Carbonate Porosity and Permeability Basedon Machine Learning and Logging Data

Accurate prediction of porosity and permeability of carbonate reservoirs is of great significance to the evaluation of carbonate reservoirs. Fractures and dissolved pores are widely developed in carbonate reservoirs, and the prediction of reservoir porosity and permeability from logging curves based on the empirical formulas has large errors. Taking a carbonate reservoir in the Middle East as the research object, we selected 914 core wells to determine porosity and permeability. By using random forest (RF), K-nearest neighbor (KNN), support vector machine (SVM), and long short-term memory  (LSTM) machine learning methods and logging data, we predicted porosity and permeability, optimized machine learning parameters, and screened out logging porosity and permeability prediction methods suitable for carbonate reservoirs. The research results show that there is little difference between the four machine learning methods in predicting reservoir porosity. By adjusting the types of input parameters, the porosity and permeability prediction results can be further improved. When using neutron (NPHI), lithological density (RHOB) ) , and acoustic time difference (DT) logging parameter data as input, the prediction accuracy based on LSTM is the highest, the root mean square error (RMSE) of the porosity prediction result is 4.536 2, and the permeability prediction is poor due to the strong heterogeneity of carbonate reservoirs when using only NPHI as the machine learning input parameter, the RF-based reservoir permeability prediction has the highest accuracy, and its RMSE  is 45.882 3.