Search

Lithology identification is one of the keys to understand the nature of hydrocarbon reservoir. Deep learning has become a popular and reliable method for image classification and in other fields. Instead of using ordinary neural networks and conventional logging curves, this paper developed deep learning methods and showed that it is possible to identify lithology, using results from borehole image logs. In this work, a Convolutional Neural Network (CNN), which consists of two convolutional layers, two pooling layers and one fully-connected layer, is employed to identify lithology. Training is performed through back-propagation using the stochastic gradient descent algorithm with Nesterov Momentum. The trained CNN can be applied to new wells and provide accurate output (about 95%) of lithology types.

The original distribution of formation fluid around the well will be changed while mud filtrate invades permeability formations. Based on the water-oil phase flow theory and convective diffusion theory, the method ofnumerical simulation is applied to build the formation model for the study of the characteristics of mud invasion in low permeability reservoir, researching the distribution characteristics of salinity, water saturation, resistivity and remaining oil around the well during mud invasion. The numerical simulation results of mud invasion in different permeability formation show that mud invasion depth increases as permeability increases in the same invasion time in low permeability reservoir, and the invasion profile of salinity , water saturation ,resistivity and remaining oil change obviously as invasion time increases. Combined with practical well numerical simulation analysis found that the characteristics and resistivity separation of array induction logging response influenced by mud invasion depth and invasion time.The simulated results show that the distribution of remaining oil have obvious relationship with reservoir fluid property. Correctly understanding the characteristics of mud invasion and the distribution of remaining oil around the well can improve the accuracy in electric logging reservoir evaluation.

A new digital focusing method with different bucking coefficients is introduced, with no change in the conventional dual laterolog device, compared to the hardware focusing method. The conclusions can be achieve that using the new digital focusing method, we can get array measurement similar to the result obtained by array laterolog devices, on the other hand ,the error in hardware focusing method badly affects the measurement. This new method can be considered to be applied in the laterolog, with simple device which can save economy cost and be operated more easily.