CN111058840A - Organic carbon content (TOC) evaluation method based on high-order neural network - Google Patents
Organic carbon content (TOC) evaluation method based on high-order neural network Download PDFInfo
- Publication number
- CN111058840A CN111058840A CN201911310645.4A CN201911310645A CN111058840A CN 111058840 A CN111058840 A CN 111058840A CN 201911310645 A CN201911310645 A CN 201911310645A CN 111058840 A CN111058840 A CN 111058840A
- Authority
- CN
- China
- Prior art keywords
- toc
- neural network
- sample
- carbon content
- organic carbon
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH DRILLING; MINING
- E21B—EARTH DRILLING, e.g. DEEP DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
- E21B49/00—Testing the nature of borehole walls; Formation testing; Methods or apparatus for obtaining samples of soil or well fluids, specially adapted to earth drilling or wells
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH DRILLING; MINING
- E21B—EARTH DRILLING, e.g. DEEP DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
- E21B47/00—Survey of boreholes or wells
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention discloses an organic carbon content (TOC) evaluation method based on a high-order neural network, which is characterized in that a correlation relation is established between the TOC and a corresponding conventional logging curve to determine the input of a proper logging curve, and the logging curve with strong correlation, namely Density (DEN), Resistivity (RT), Gamma Ray (GR), neutron (CNL) and sound wave time difference (AC), is selected, on the basis of improving the calculation speed, the high-order neural network method is adopted, deep learning is specially carried out on the weights of the logging curves, the accuracy and the timeliness of the evaluation of the organic carbon content (TOC) are effectively improved, and the defects in the prior art are overcome.
Description
Technical Field
The invention relates to an evaluation method of organic carbon content (TOC) of a reservoir in the field of petroleum and natural gas reservoir geology, and particularly relates to an organic carbon content (TOC) calculation method based on an artificial neural network, which can effectively improve the accuracy of evaluation of the organic carbon content (TOC) of the reservoir in an evaluation stage.
Background
With the explosive development of north american shale exploration, chinese continental basin shale oil and gas has received extensive attention. Shale oil and gas refers to oil and gas that is concentrated in black shale formations rich in organic matter and has not undergone a normal migration process. In recent years, shale gas has become a research hotspot for the exploration and development of unconventional natural gas in the world, and related research work of shale oil is emerging continuously. Many shale reservoirs have been discovered, such as Sichuan, Bohai Bay, Songliao, Jianghan, Tarim basins. The TOC is a basic index for evaluating the abundance of organic matters and reflects the hydrocarbon generation potential of the shale, so that accurate and continuous prediction is urgently needed. However, direct laboratory methods are time consuming and expensive. In recent years, the relationship between geochemical parameters of source rock and well log information has been studied.
The log data associated with organic matter includes Density (DEN), Resistivity (RT), Gamma Ray (GR), neutron (CNL), and acoustic time difference (AC). Previous studies have shown that better prediction accuracy is only possible when DEN, AC and TOC are correlated. Kadhodaie-ilkhchiy et al (2009) have used neural networks to select appropriate input parameters and found that selecting AC, DEN, RT and CNL as input parameters results in the smallest Mean Square Error (MSE). Wang et al (2018) indicate that a log selected based on the Mean Impact Value (MIV) will improve prediction accuracy. There are three established techniques for quantitative evaluation of TOC in hydrocarbon source rocks, the Passey method (Passey et al, 1990) and based on Schmoker density logging (Schmoker,1979,1980) and Artificial Neural Networks (ANN) (here we refer to back-propagation artificial neural networks). In practical application, due to the existence of a complex nonlinear relation between the TOC and the well logging curve, the calculation of the TOC by the neural network is more reliable and advanced. Therefore, it has been applied to many petroleum engineering fields, particularly unconventional resource evaluation. However, ANN also has some disadvantages during training. The neural network is easy to fall into local optimum, the overfitting problem exists, and the prediction precision is greatly reduced. Salaheddin (2019) applied an adaptive differential evolution (SaDE) optimization method to determine the optimal combination of Artificial Neural Network (ANN) parameters (number of hidden layers, number of neurons per layer, training function, transfer function, training-to-test ratio). On the basis of the optimized SaDE-ANN model, a new empirical correlation is established for estimating TOC by using a well logging curve.
The higher-order neural network is an extension of the multi-layer perceptron neural network. It adds auxiliary elements on the basic model of the sensor to change the input vector into the N values of the sensor (for example, the input parameter x in the multi-layer sensor neural network1、x2In the higher-order neural network, it becomes
x1x2、x1、x2). The network output corresponds to the input high-order correlation function, the calculation complexity is reduced, and even if the data volume is large, the convergence speed of the network output is obviously superior. And the high-order neural network does not contain a hidden layer, so that a higher training speed can be obtained, local minimum values are not easy to appear, and the problems of the number of the hidden layers and the number of nodes are solved.
Logging data based on the high cost, long duration of core analysis and high longitudinal resolution can provide continuous organic carbon content evaluation along well profiles. The prediction accuracy of the Passey method is low, and the BP neural network is set according to experience because the number of layers and the number of nodes of a hidden layer are large, so that the convergence speed is slow when the data volume is large, and the problem that the BP neural network is easy to fall into a local minimum value exists. In order to reasonably solve the problem, the invention provides a high-order neural network-based organic carbon content prediction method, which is expected to further improve the accuracy and timeliness of organic carbon content evaluation.
Disclosure of Invention
The invention aims to provide a high-order neural network-based organic carbon content (TOC) evaluation method, which is based on a nonlinear pattern recognition technology of an artificial neural network, and utilizes five logging curves of Density (DEN), Resistivity (RT), Gamma Ray (GR), neutron (CNL) and sound wave time difference (AC) to carry out deep learning specially aiming at the weight of the logging curves influencing the organic carbon content, thereby effectively improving the accuracy of the comprehensive evaluation of the organic carbon content (TOC) in the reservoir evaluation stage and overcoming the defects in the prior art.
In order to achieve the above technical objects, the present invention provides the following technical solutions.
A high-order neural network-based organic carbon content (TOC) evaluation method comprises the steps of inputting a proper logging curve, selecting a logging curve with strong correlation, namely Density (DEN), Resistivity (RT), Gamma Ray (GR), neutron (CNL) and sound wave time difference (AC), and improving calculation speed
The core of the method is to determine the appropriate logging curve input and the calculation of the weight of each logging curve. The method sequentially comprises the following steps:
(1) establishing a correlation between the TOC and the logging curve to determine the input of a proper logging curve, selecting the logging curve with strong correlation, namely Density (DEN), Resistivity (RT), Gamma Ray (GR), neutron (CNL) and sound wave time difference (AC), and improving the calculation speed;
(2) carrying out normalization processing on the data to determine a training sample;
(3) setting the order of a neural network, converting an input sample, and inputting;
(4) randomly setting an initial connection right;
(5) calculating actual output;
(6) updating the connection weight according to the error between the expected output and the actual output, and continuously performing network training;
(7) and when all the samples are trained and the network precision requirement is met, determining the weight of each logging curve. And obtaining organic carbon content prediction models of all samples.
Drawings
FIG. 1 is a schematic diagram of a high-order neural network model of a high-order neural network-based organic carbon content (TOC) evaluation method provided by the invention;
fig. 2 is a schematic flow chart of a method for evaluating organic carbon content (TOC) based on a high-order neural network provided by the invention.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The embodiment of the present invention is calculated step by step according to the flowchart shown in fig. 2, and is described and illustrated below by taking a specific example as an example. The content is to explain the invention and not to limit the scope of protection of the invention.
The organic carbon content evaluation method based on the high-order neural network in the reservoir evaluation stage sequentially comprises the following detailed steps of:
the method comprises the following steps: a correlation relationship is established between the TOC and the logging curves to determine a proper logging curve input, and the logging curves with strong correlation, namely Density (DEN), Resistivity (RT), Gamma Ray (GR), neutron (CNL) and sound wave time difference (AC), are selected to improve the calculation speed.
Step two: and reading corresponding DEN, RT, GR, CNL and AC data for each test TOC, and performing statistics and normalization processing to obtain a plurality of input samples. From the first sample X1(equation 1) the operation was started.
X1＝(x1,x2,x3,x4,x5) (1)
Wherein x is1＝DEN，x2＝RT，x3＝GR，x4＝CNL，x5＝AC。
Step three: the conversion order of the high-order neural network to the input parameter is set to be 2, that is, the input sample is changed into (equation 2).
The square terms of the 5 input parameters have 5 terms, the quadratic terms of the pairwise product of the 5 input parameters have 10 terms, the primary terms of the 5 input parameters are 5 terms, the last term constant 1 is added, and the total term number is 21, namely, the 5 input parameters are converted into 21 parameters containing high-order independent variables.
Step four: let the expected output corresponding to the first sample be, set the initial weight vector between and, the middle element can be randomly set to any value between 0-1 (equation 3) (fig. 1).
W1＝(w1,w2,···,w21)T(3)
Wherein, w1,w2,···,w21The weight coefficient of each item in the third step.
Step five: calculating the actual output Z of the first output node1(formula 4).
Z1＝f(W1 TX1 *) (4)
Where f is an excitation function representing the functional relationship between the input samples and the output. Common stimulus functions include: sigmoid function, tanh function, ReLU function, etc. The Sigmoid function is chosen here as the excitation function.
Step six: output O according to expectation1And the actual output Z1Error of (3), updating the connection weight vector W1Is W2(formula 5), W2I.e. the initial weight vector for the second sample.
W2＝W1+η(O1-Z1)X*(5)
In the formula, η is a weight coefficient update step, η gradually decreases as the number of iterations increases, and generally the initial value is randomly set to a smaller positive value, where η is set as the initial value of the step1，η1The value is 0.1, the product is set to be 0.1 multiplied by 100 rounds per iteration, namely the step length is η after all samples are repeatedly fed back and trained for 100 rounds2，η2＝0.1η10.1 × 0.1 — 0.01, and so on.
Step seven: and (5) assuming that the total number of the samples is n, circulating the step five and the step six, and sequentially obtaining an initial weight vector of each sample.
Step eight: the average error E (equation 6) is calculated for all samples.
Wherein, OjAnd ZjThe expected output and the actual output of the jth sample, respectively.
Step nine: when the average error E is less than or equal to epsilon (set
) The weights of the logs can then be determined. And obtaining organic carbon content prediction models of all samples. Otherwise, returning to the step five.
The flow chart corresponding to the above steps is shown in fig. 2 below.
The foregoing illustrates and describes the principles, general features, and advantages of the present invention. Those skilled in the art will appreciate that the above-described embodiments are not intended to limit the invention in any way. All the technical solutions obtained by means of equivalent substitution or equivalent transformation fall within the protection scope of the present invention.
Claims (5)
1. A method for evaluating organic carbon content (TOC) based on a high-order neural network is characterized by sequentially comprising the following detailed steps of:
(1) establishing a correlation between the TOC and the logging curve to determine the input of a proper logging curve, selecting the logging curve with strong correlation, namely Density (DEN), Resistivity (RT), Gamma Ray (GR), neutron (CNL) and sound wave time difference (AC), and improving the calculation speed;
(2) reading corresponding DEN, RT, GR, CNL and AC data for each TOC tested in the step (1) for statistics and normalization processing to obtain a plurality of input samples; starting the run from the first sample;
(3) setting the order of the high-order neural network, and converting the order of the input sample in the step (2);
(4) setting an initial weight vector of the converted first sample;
(5) outputting the converted first sample;
(6) updating the connection weight vector of the first sample into an initial weight vector of the second sample according to the error of the expected output and the actual output of the first sample;
(7) circulating the step (5) and the step (6), and sequentially obtaining an initial weight vector of each sample;
(8) and (5) calculating the average error of all samples, determining the weight of each logging curve when the average error meets the accurate reading requirement, calculating a prediction model of the organic carbon content of all samples, and otherwise, returning to the step (5).
2. The method for evaluating organic carbon content (TOC) based on a higher-order neural network according to claim 1, wherein the method for updating the connection weight vector of the first sample in step (6) is as follows:
W2＝W1+η(O1-Z1)X*
in the formula, W1To connect weight vectors, W2For updated connection weight vectors, O1To the desired output, Z1For actual output η is the weight coefficient update step size, and η is gradually reduced as the number of iterations increases.
3. The method of claim 2, wherein the weight coefficient update step is set by randomly setting an initial value to a small positive value, wherein the initial value of the weight coefficient update step is η1，η1The value is 0.1, the multiplication is set to be 0.1 by 100 rounds per iteration, namely the step length becomes η after all samples repeat the feedback training for 100 rounds2，η2＝0.1η10.1 × 0.1 — 0.01, and so on.
4. The method for evaluating the organic carbon content (TOC) based on the high-order neural network as claimed in claim 1, wherein the method for calculating the average error of all samples in the step (8) comprises the following steps:
wherein E is the average error of all samples; o isjAnd ZjThe expected output and the actual output of the jth sample, respectively.
5. The method for evaluating the organic carbon content (TOC) based on the high-order neural network as claimed in claim 1, wherein the average error precision requirement in the step (8) is 3 per thousand.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911310645.4A CN111058840A (en) | 2019-12-18 | 2019-12-18 | Organic carbon content (TOC) evaluation method based on high-order neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911310645.4A CN111058840A (en) | 2019-12-18 | 2019-12-18 | Organic carbon content (TOC) evaluation method based on high-order neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111058840A true CN111058840A (en) | 2020-04-24 |
Family
ID=70302275
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911310645.4A Pending CN111058840A (en) | 2019-12-18 | 2019-12-18 | Organic carbon content (TOC) evaluation method based on high-order neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111058840A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111749677A (en) * | 2020-06-30 | 2020-10-09 | 宋立才 | Logging information processing method and device, electronic equipment and storage medium |
| CN113837501A (en) * | 2021-11-25 | 2021-12-24 | 中国科学院地质与地球物理研究所 | Shale total organic carbon prediction method, system and equipment based on graph neural network |
| CN113835138A (en) * | 2021-11-25 | 2021-12-24 | 中国科学院地质与地球物理研究所 | Method for predicting total organic carbon content of shale based on deep coding decoding network |
| CN115032361A (en) * | 2021-03-03 | 2022-09-09 | 中国石油化工股份有限公司 | Method for evaluating organic carbon content of shale oil reservoir based on genetic optimization neural network algorithm |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573344A (en) * | 2014-12-25 | 2015-04-29 | 中国海洋石油总公司 | Method for acquiring gas content of shale reservoir through well logging data |
| CN105804740A (en) * | 2016-03-17 | 2016-07-27 | 成都创源油气技术开发有限公司 | Evaluation method of shale gas reservoir organic carbon content and logging parameter relationship and application |
| WO2017024530A1 (en) * | 2015-08-11 | 2017-02-16 | 深圳朝伟达科技有限公司 | Method for calculating content of organic carbon in hydrocarbon source rock |
| CN106950347A (en) * | 2017-03-03 | 2017-07-14 | 中国石油大学(华东) | A kind of method for evaluating mud shale each group partial volume |
| CN108843312A (en) * | 2018-06-04 | 2018-11-20 | 西南石油大学 | Oilfield reservoir in-layer heterogeneity integrated evaluating method |
| CN110442951A (en) * | 2019-07-30 | 2019-11-12 | 西北大学 | A kind of hydrocarbon source rock total content of organic carbon prediction technique considering density factor |
-
2019
- 2019-12-18 CN CN201911310645.4A patent/CN111058840A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573344A (en) * | 2014-12-25 | 2015-04-29 | 中国海洋石油总公司 | Method for acquiring gas content of shale reservoir through well logging data |
| WO2017024530A1 (en) * | 2015-08-11 | 2017-02-16 | 深圳朝伟达科技有限公司 | Method for calculating content of organic carbon in hydrocarbon source rock |
| CN105804740A (en) * | 2016-03-17 | 2016-07-27 | 成都创源油气技术开发有限公司 | Evaluation method of shale gas reservoir organic carbon content and logging parameter relationship and application |
| CN106950347A (en) * | 2017-03-03 | 2017-07-14 | 中国石油大学(华东) | A kind of method for evaluating mud shale each group partial volume |
| CN108843312A (en) * | 2018-06-04 | 2018-11-20 | 西南石油大学 | Oilfield reservoir in-layer heterogeneity integrated evaluating method |
| CN110442951A (en) * | 2019-07-30 | 2019-11-12 | 西北大学 | A kind of hydrocarbon source rock total content of organic carbon prediction technique considering density factor |
Non-Patent Citations (1)
| Title |
|---|
| 赵兴齐等: "BP神经网络在西湖凹陷烃源岩评价中的应用", 《测井技术》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111749677A (en) * | 2020-06-30 | 2020-10-09 | 宋立才 | Logging information processing method and device, electronic equipment and storage medium |
| CN115032361A (en) * | 2021-03-03 | 2022-09-09 | 中国石油化工股份有限公司 | Method for evaluating organic carbon content of shale oil reservoir based on genetic optimization neural network algorithm |
| CN113837501A (en) * | 2021-11-25 | 2021-12-24 | 中国科学院地质与地球物理研究所 | Shale total organic carbon prediction method, system and equipment based on graph neural network |
| CN113835138A (en) * | 2021-11-25 | 2021-12-24 | 中国科学院地质与地球物理研究所 | Method for predicting total organic carbon content of shale based on deep coding decoding network |
| CN113835138B (en) * | 2021-11-25 | 2022-02-25 | 中国科学院地质与地球物理研究所 | Method for predicting total organic carbon content of shale based on deep coding decoding network |
| CN113837501B (en) * | 2021-11-25 | 2022-03-08 | 中国科学院地质与地球物理研究所 | Shale total organic carbon prediction method, system and equipment based on graph neural network |
| US20230162052A1 (en) * | 2021-11-25 | 2023-05-25 | Institute Of Geology And Geophysics, Chinese Academy Of Sciences | Graph neural network (gnn)-based prediction system for total organic carbon (toc) in shale |
| US11880776B2 (en) * | 2021-11-25 | 2024-01-23 | Institute Of Geology And Geophysics, Chinese Academy Of Sciences | Graph neural network (GNN)-based prediction system for total organic carbon (TOC) in shale |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111058840A (en) | Organic carbon content (TOC) evaluation method based on high-order neural network | |
| CN112989708B (en) | Well logging lithology identification method and system based on LSTM neural network | |
| CN111048163B (en) | Shale oil hydrocarbon retention amount (S1) evaluation method based on high-order neural network | |
| Huang et al. | An integrated neural-fuzzy-genetic-algorithm using hyper-surface membership functions to predict permeability in petroleum reservoirs | |
| CN111324990A (en) | Porosity prediction method based on multilayer long-short term memory neural network model | |
| CN111027882A (en) | Method for evaluating brittleness index by utilizing conventional logging data based on high-order neural network | |
| CN110083125B (en) | Machine tool thermal error modeling method based on deep learning | |
| CN113052371A (en) | Residual oil distribution prediction method and device based on deep convolutional neural network | |
| WO2005098647A2 (en) | Genetic algorithm based selection of neural network ensemble for processing well logging data | |
| CN112017289A (en) | Well-seismic combined initial lithology model construction method based on deep learning | |
| CN113359212B (en) | Reservoir characteristic prediction method and model based on deep learning | |
| CN110909931A (en) | Logging curve prediction method based on modal decomposition reconstruction and depth LSTM-RNN model | |
| CN110515123B (en) | Convolution neural network seismic logging joint inversion method based on small sample learning | |
| CN113610945B (en) | Ground stress curve prediction method based on hybrid neural network | |
| Zhan et al. | Subsurface sedimentary structure identification using deep learning: A review | |
| CN111766635B (en) | Sand body communication degree analysis method and system | |
| CN109165737B (en) | Porosity prediction method based on conditional random field and BP neural network | |
| CN114723095A (en) | Missing well logging curve prediction method and device | |
| Gasmi et al. | Physics informed deep learning for flow and transport in porous media | |
| CN111723949A (en) | Porosity prediction method based on selective ensemble learning | |
| CN116107279A (en) | Flow industrial energy consumption multi-objective optimization method based on attention depth neural network | |
| CN113935489A (en) | Variational quantum model TFQ-VQA based on quantum neural network and two-stage optimization method thereof | |
| CN116047583A (en) | Adaptive wave impedance inversion method and system based on depth convolution neural network | |
| CN110633868A (en) | Method for predicting properties of oil and gas in exploration well oil testing layer by optimizing neural network through genetic algorithm | |
| CN107909202B (en) | Time sequence-based oil well liquid production integrated prediction method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200424 |
|
| WD01 | Invention patent application deemed withdrawn after publication |