Exploration
Mahdi Bajolvand; Ahmad Ramezanzadeh; Amin Hekmatnejad; Mohammad Mehrad; Shadfar Davoodi; Mohammad Teimuri
Abstract
Shear Wave Slowness Log (DTSM) is one of the most important petrophysical logs applicable for studying reservoirs, especially geomechanical studying of the oil and gas fields. However, lack of this parameter in wellbore logging can import great sources of uncertainty into geomechanical studies. This ...
Read More
Shear Wave Slowness Log (DTSM) is one of the most important petrophysical logs applicable for studying reservoirs, especially geomechanical studying of the oil and gas fields. However, lack of this parameter in wellbore logging can import great sources of uncertainty into geomechanical studies. This study aims to provide solutions for decreasing the uncertainty of geomechanical models with estimation of the DTSM log using the high accurate deep machine learning models. The main idea is using data from offset fields for extending the range of training data and improving the estimation ability and generalizability of machine learning models. For this purpose, petrophysical data from 8 wells of 4 Iranian oil fields were collected. In the first stage, data preprocessing was performed for reducing the effects of wrong data, missing value, noises, and outliers. Then, machine learning (regression learning-based and deep neural network-based) and analytical models implemented for estimating DTSM. The results indicated that the Gated Recurrent Unit (GRU) model with the values of 1.9 and 2.14 for RMSE and 0.99 for R-square had the most exact answers, for training and test data, respectively. Meanwhile, evaluation of the accuracy of the models on the validation well data indicated that GRU model with the values of 2.43 and 0.93 had been the most accurate model for RMSE and R-square, respectively. Accordingly, using a multi-field comprehensive data bank and applying machine learning methods are strongly recommended to estimate the DTSM, for the cases where limited offset data is available.
Y. Asgari Nezhad; A. Moradzadeh
Abstract
One of the most essential factors involved in unconventional gas reserves for drilling and production is a suitable quality facies determination. The direct core and geochemical analyses are the most common methods used for studying this quality. Due to the lack of this data and the high cost, the researchers ...
Read More
One of the most essential factors involved in unconventional gas reserves for drilling and production is a suitable quality facies determination. The direct core and geochemical analyses are the most common methods used for studying this quality. Due to the lack of this data and the high cost, the researchers have recently resorted to the indirect methods that use the common data of the reservoir (including petro-physical logs and seismic data). One of the major problems in using these methods is that the complexities of these reproducible repositories cannot be accurately modeled. In this work, the quality of facies in shale gas is zoned using the deep learning technique. The applied method is long short-term memory (LSTM) neural network. In this scheme, the features required for zoning are automatically extracted and used to model the reservoir complexities properly. The results of this work show that zoning is done with an appropriate accuracy (86%) using the LSTM neural network, while it is 78% for a conventional intelligent MLP network. This specifies the superior accuracy of the deep learning method.
M. Abedini; M. Ziaii; Y. Negahdarzadeh; J. Ghiasi-Freez
Abstract
The porosity within a reservoir rock is a basic parameter for the reservoir characterization. The present paper introduces two intelligent models for identification of the porosity types using image analysis. For this aim, firstly, thirteen geometrical parameters of pores of each image were extracted ...
Read More
The porosity within a reservoir rock is a basic parameter for the reservoir characterization. The present paper introduces two intelligent models for identification of the porosity types using image analysis. For this aim, firstly, thirteen geometrical parameters of pores of each image were extracted using the image analysis techniques. The extracted features and their corresponding pore types of 682 pores were used for training two intelligent models, BPN (back-propagation network) and SAE (stacked autoencoder). The trained models take the geometrical properties of pores to classify the type of six porosity types including intra-particle, inter-particle, vuggy, moldic, biomoldic, and fracture. The MSE values for the BPN and SAE models were found to be 0.0042 and 0.0038, respectively. The precision, recall, and accuracy of the intelligent models for classifying the types of pores were calculated. The BPN model was able to correctly recognize 193 intra-particle pores out of 197 ones, 45 inter-particle pores out of 50 ones, 7 vuggy pores out of 9 ones, 10 moldic pores out of 12 ones, 2 biomoldic pores out of 3 ones, and 6 fractures out of 7 ones. Also the SAE model was able to correctly classify 193 intra-particle pores out of 197 ones, 46 inter-particle pores out of 50 ones, 8 vuggy pores out of 9 ones, 10 moldic pores out of 12 ones, 3 biomoldic pores out of 3 ones, and 7 fractures out of 7 ones. The results obtained showed that the SAE model carried out a bit more accuracy for classification of the inter-particle, vuggy, biomoldic, and fracture pores.
