Reservoir Identification and Classification Based on Principal Component Analysis and Supervised Neural Network in Carbonate Rocks

Carbonates always have complex reservoir space and diverse pore structure, so it is difficult to accurately identify and classify the carbonate reservoirs. Therefore, a fusion method combining principal component analysis with supervised neural network is proposed to identify and classify of the reservoir spaces in carbonate rocks. On the one hand, principal component analysis is applied to reconstruct the conventional identification indexes such as three porosity ratio, resistivity ratio, etc. into a comprehensive index as the standard for reservoir identification. On the other hand, according to the automatic identification results marked by the comprehensive index, the sensitivity parameters such as conventional logging curves, fractured-vuggy porosity of electric imaging logging, anisotropy of array acoustic logging, etc. as machine learning objects are optimized to input the supervised neural network. The automatic classification method is implemented by establishing self-organizing mapping relationships from sensitivity parameters to reservoir types. The reservoir identification based on principal component analysis can make up for the deficiency that it may be inaccurate to identify the reservoirs through only one or two logging methods. It can be used as a reliable replacement, if a reservoir has limited and insufficient data of image loggings. The reservoir classification based on the supervised neural network is beneficial to indicate the distribution of reservoir spaces and improve the horizontal comparability of several wells. This fusion method combining principal component analysis with supervised neural network is first applied to the carbonate reservoirs located in the Pre-Caspian Basin. The results show that the novel method can not only can achieve the high coincidence with core and image logging data, but also provide algorithm support for the fine evaluation of complex carbonate reservoirs.