Porosity prediction and application with multiwell-logging curves based on deep neural network

Zhang

et al. 2022

Energies

Porosity is a vital parameter in reservoir research. In the process of oil exploration, reservoir research is very important for oil and gas exploration. Because it is necessary to take cores for indoor test in order to accurately obtain the porosity value of cores, this process consumes significant manpower and material resources. Therefore, this paper introduces the method of machine learning to predict the porosity by using logging curves. This paper creatively develops a WOA (whale optimization algorithm) optimized Elman neural network model to predict porosity through logging parameters PE, DEN, M2R1, AC, GR, R25, R4 and CNL. Porosity measurement is constructed by taking cores for indoor experiments. It contains a total of 328 sample points. The data is divided into training set and test set. The logging parameters are used as the input parameters of the prediction model, and the porosity measured in the laboratory are used as the output parameter. In order to evaluate the performance of the model, RMSE, R2, MAE and VAF evaluation indexes are introduced to evaluate. This paper also introduces the non-optimized Elman neural network and BP neural network to compare with this optimization model. The research shows that the WOA algorithm optimizes the super parameters of the Elman neural network, so that the performance of the WOA–Elman model is better than the Elman neural network model and the BP neural network model.

Section: Analysis Of Prediction Resultsmentioning

confidence: 69%

Section: Analysis Of Prediction Resultsmentioning

confidence: 99%

WOA (Whale Optimization Algorithm) Optimizes Elman Neural Network Model to Predict Porosity Value in Well Logging Curve

Zhang

et al. 2022

Energies

“…The statistical results showed that in terms of calculation speed, the GPR prediction model predicated reservoir porosity and permeability faster than BPNN, GRNN (generalized regression neural networks), and RBFNN (radial basis function neural networks) methods. The deep learning model was effective and practical in the field of porosity prediction, and it also used multiple logging parameters such as compensated neutron, acoustic moveout, natural gamma, and compensated density to establish nonlinear relationships with porosity (An, 2018) . Urang (2020) established two independent multilayer perceptron neural networks using the Bayesian framework and established a mathematical model to predict reservoir physical properties, and the results showed that these models were simple, economical, and effective reservoir property estimation methods.…”

Section: Introductionmentioning

confidence: 99%

Bidirectional Long Short-term Neural Network Based on the Attention Mechanism of the Residual Neural Network (ResNet–BiLSTM–Attention) Predicts Porosity through Well Logging Parameters

Zhang

et al. 2023

ACS Omega

Porosity is an integral part of reservoir evaluation, but in the field of reservoir prediction, due to the complex nonlinear relationship between logging parameters and porosity, linear models cannot accurately predict porosity. Therefore, this paper uses machine learning methods that can better handle the relationship between nonlinear logging parameters and porosity to predict porosity. In this paper, logging data from Tarim Oilfield are selected for model testing, and there is a nonlinear relationship between these parameters and porosity. First, the data features of the logging parameters are extracted by the residual network, which uses the "hop connections" method to transform the original data closer to the target variable. In addition, the residual blocks inside the residual network use jump connections, which alleviates the gradient vanishing problem caused by increasing depth in deep neural networks. The dynamic nature of data would necessitate LSTM in the first place. Then, a bidirectional long short-term network (BiLSTM) is used to predict the porosity of the extracted logging data features. Among them, the BiLSTM is composed of two independent reverse LSTMs, which can better solve the nonlinear prediction problem. In order to further improve the accuracy of the model, this paper introduces an attention mechanism to learn by weighting each of the inputs in proportion to their impact on the porosity. The experimental results also show that the data features extracted by the residual neural network can be better used as the input of the BiLSTM model.

“…For example, machine learning is used for estimating the depth of buried magnetic steel drums (Salem et al, 2000), first break picking of seismic traces (McCormack et al, 1993), locating subsurface targets using electromagnetic data (Poulton et al, 1992), and many others. In the last 3 years, the geophysics community has found a big surge of interest in machine learning, especially in seismic interpretation (Araya‐Polo et al, 2017; Di et al, 2018; X. Wu et al, 2018), well‐logging (An et al, 2018; Bestagini et al, 2017; P. Y. Wu et al, 2018), and lithofacies classification (Bhattacharya, 2016; L. Liu et al, 2017; Pires de Lima et al, 2019). However, the use of machine learning for the purpose of predicting physical parameters from magnetic data maps, such as magnetization directions, has not been attempted yet.…”

Section: Introductionmentioning

confidence: 99%

Predicting Magnetization Directions Using Convolutional Neural Networks

Nurindrawati

2020

JGR Solid Earth

Magnetic data have been widely used for understanding basin structures, mineral deposit systems, the formation history of various geological systems, and many others. Proper interpretation of magnetic data requires accurate knowledge of total magnetization directions of the source bodies in an area of study. Existing approaches for estimating magnetization directions involve either unstable data processing steps or computationally intensive processes such as 3-D inversions. In this study, we developed a new method of automatically predicting the magnetization direction of a magnetic source body using convolutional neural networks (CNNs). CNNs have achieved great success in many other applications such as computer vision and seismic image interpretation but have not been used to extract parameters from magnetic data. We simulated many magnetic data maps with different magnetization directions from a synthetic source body, all subject to the same background field. Two CNNs were trained separately, one for predicting the inclination and the other for predicting declination. We determined the optimal CNN architectures for predicting inclinations and declinations by systematically comparing 13 different CNN architectures. In addition, we investigated the effect of having different parameters such as magnetization magnitude, source body shape, location, and depth on the performance of our predictive models. We also tested the method using field data from Black Hill Norite, Australia, and Yeshan region, China, for which prior research results are available for comparison. Our study shows that machine learning provides an effective means of automatically predicting magnetization directions based on magnetic data maps.