Deep learning based reservoir monitoring index early warning method

Abstract

The application provides a kind of oil reservoir monitoring index early warning method based on deep learning, which comprises the following steps: step 1, collect historical data; step 2, establish a convolutional neural network model, with bottom hole flowing pressure value as input value, corresponding bottom hole temperature maximum and minimum as output value; step 3, use the label data corresponding to the maximum and minimum of bottom hole temperature to train two neural networks respectively to obtain two prediction models; step 4, according to the predicted maximum and minimum of bottom hole temperature, the uncertainty interval is quantified. The oil reservoir monitoring index early warning method based on deep learning provides reliable data-driven basis for oil reservoir dynamic early warning. Meanwhile, the deep convolutional neural network structure used for predicting the upper and lower bounds of bottom hole temperature in the application has the advantages of good noise resistance and high accuracy compared with the fully connected neural network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oilfield development, and particularly relates to an oil reservoir monitoring index early warning method based on deep learning. BACKGROUND

[0002] Dynamic monitoring technology has gradually formed technologies in aspects of basic experiment, monitoring scheme design, monitoring data collection, data interpretation and comprehensive application through years of development, and becomes another important means for understanding and evaluating oil reservoir development dynamics and effects besides oil reservoir engineering and oil reservoir numerical simulation. The traditional oil reservoir engineering method based on analytical solution of percolation mechanics equation needs to greatly simplify the oil reservoir model, and cannot fully consider the influence of reservoir heterogeneity and complex flow mechanism. Oil reservoir numerical simulation is a conventional method for oil reservoir dynamic analysis and prediction. Its advantages are that the correlation between parameters is clear based on mechanism, the method is mature through years of development, and can be realized based on commercial software. Its main limitations are that it depends on accurate geological model and percolation mechanism, the error is large for blocks with unclear geological model or percolation mechanism, and only part of the monitoring data can be considered. The data-driven method is to directly fit the nonlinear relationship between data by using machine learning and deep learning methods, and to carry out corresponding analysis and prediction, without clear percolation mechanism or geological model, but the quality requirement of collected data is high.

[0003] The data analysis based on machine learning mainly includes correlation analysis, denoising and dimensionality reduction, classification and regression, prediction, inversion and control, and has formed algorithms such as k-nearest neighbor method, support vector machine, logistic regression and neural network. How to combine machine learning algorithms with specific engineering practice and establish corresponding data-driven analysis technology is the focus of industrial big data research. For oil reservoir monitoring data, establishing corresponding data analysis algorithms based on machine learning is beneficial to fully excavate the characteristics of monitoring data and comprehensively consider the relationship between each dimension of data, so as to make monitoring data play a greater role in guiding efficient development of oilfields, and has important significance for intelligent oil and gas field construction. The existing oil reservoir monitoring index early warning mainly depends on subjective experience and percolation mechanism, and still lacks a data-driven early warning method.

[0004] In the Chinese patent application No. CN202011615557.8, a method and system for oilfield production dynamic analysis based on big data are disclosed. The method includes obtaining basic data of a target well and its nearest neighbor wells, which includes static data, dynamic data, and associated data. The obtained data is preprocessed and fused to obtain index data. A distributed storage data warehouse is established, and the index data is stored in the data warehouse through a data interface. The index data in the data warehouse is used as modeling parameters to construct an oilfield production diagnosis and early warning model. The model is visualized using data visualization technology, and early warning is performed based on the model.

[0005] In the Chinese patent application No. CN202011608819.8, an intelligent fault diagnosis method and system for offshore oil and gas wells are disclosed. The method includes data cleaning, data analysis, model establishment, and model evaluation. The method is based on offshore oil and gas wells and includes data cleaning, feature analysis, fault diagnosis, and fault prediction. It is beneficial to improve collaboration efficiency and reduce communication costs. The invention designs a development well history data integration and intelligent early warning system that applies statistical analysis techniques to monitor oil and gas well production processes in real time and provide early warnings for abnormal trends. It also identifies possible faults and risks in different oil and gas fields to help production managers take timely measures, thereby improving production efficiency and reducing maintenance costs.

[0006] In the Chinese patent application No. CN202211297047.X, a method for oil well multi-parameter anomaly diagnosis and early warning based on machine learning is disclosed. The method includes data preprocessing of oil well production data, threshold-based abnormal data processing for abnormal data, analysis of production parameters, long-term trend anomaly identification for single parameters in oil wells, establishment of a fault classification model for multi-parameter anomaly identification, development of fault identification rules and fault type identification rules, suggestion of disposal methods, formation of a complete oil well fault case library through deep identification and manual inspection, and early warning of possible oil well anomalies based on long-term operation condition index. The invention can collect oil well data in real time, accurately diagnose possible fault types, and provide effective deep early warning, thereby effectively and long-term improving oil well production efficiency.

[0007] In the Chinese patent application with application number: CN202111029072.5, a deep learning-based pumping unit well timing indicator diagram prediction method is involved. The method comprises: (1) constructing a training data set; (2) normalization and interpolation processing; (3) first stage: indicator diagram working condition trend discrimination; (4) second stage: indicator diagram prediction; (5) third stage: diagnosis and early warning; the invention constructs a training data set by slicing, normalizing and interpolating historical indicator diagrams, applies a multi-channel convolutional neural network (CNN) to discriminate the timing trend working conditions, such as waxing, insufficient liquid supply, gas lock and pump valve leakage, and applies an LSTM sequence-to-sequence model to capture the change trend characteristics and change speed characteristics of the current and past indicator diagrams, making quantitative prediction of future indicator diagrams, so that the working condition warning information is more intuitive and has stronger interpretability.

[0008] The above prior art is quite different from the present application and cannot solve the technical problems we want to solve. Therefore, we have invented a new deep learning-based oil reservoir monitoring index early warning method. SUMMARY

[0009] The purpose of the present application is to provide a deep learning-based oil reservoir monitoring index early warning method with good anti-noise effect and high accuracy.

[0010] The purpose of the present application can be achieved by the following technical measures: a deep learning-based oil reservoir monitoring index early warning method, comprising:

[0011] Step 1, collect historical data;

[0012] Step 2, establish a convolutional neural network model, with the bottom hole flowing pressure value as the input value, and the corresponding maximum and minimum bottom hole temperature values as the output value;

[0013] Step 3, train two neural networks using the label data corresponding to the maximum and minimum bottom hole temperature values to obtain two prediction models;

[0014] Step 4, obtain the predicted maximum and minimum bottom hole temperature values according to the prediction model, thereby quantifying the uncertainty interval.

[0015] The purpose of the present application can also be achieved by the following technical measures:

[0016] In step 1, the daily oil well bottom hole flowing pressure value and the corresponding maximum and minimum bottom hole temperature values of the bottom hole flowing pressure value are collected.

[0017] In step 1, daily oil well bottom flow pressure values and all possible bottom hole temperature values corresponding to the bottom flow pressure values are collected, and three sets of data, i.e. bottom flow pressure and corresponding bottom hole temperature maximum and minimum values, are obtained by taking the maximum value as the label data for neural network training.

[0018] In step 2, the convolutional neural network comprises multiple modules and is finally output to the required dimension by the fully connected layer after dimension transformation.

[0019] In step 2, each module is composed of a convolutional layer, a normalization layer, a ReLU activation function layer and a maximum pooling layer arranged in sequence.

[0020] In step 2, the convolutional layer is used to extract data features through convolution operation; the normalization layer is used to normalize network parameters to improve training stability and convergence; the maximum pooling layer is used to reduce dimension to improve training efficiency; and the matrix after the maximum pooling layer is deformed into one dimension and output to the dimension of the bottom hole temperature through a linear layer.

[0021] In step 3, the Adam optimizer is used for training, and two neural networks are trained respectively using the label data corresponding to the maximum and minimum values of the bottom hole temperature to obtain two prediction models.

[0022] In step 3, the obtained prediction model is:

[0023]

[0024]

[0025]

[0026] wherein, p wf =p wf (t) is the bottom flow pressure value, is the approximate value of the maximum bottom hole temperature, is the approximate value of the minimum bottom hole temperature,

[0027] f is a neural network operator mapping from the bottom flow pressure value p wf to the approximate value of the maximum bottom hole temperature .

[0028] g is a neural network operator mapping from the bottom flow pressure value p wf to the approximate value of the minimum bottom hole temperature .

[0029] The bottom hole temperature corresponding to the bottom flow pressure value varies in a range.

[0030] At step 4, the measured well bottom flow pressure value is input into two prediction models at the prediction time to obtain the predicted maximum and minimum well bottom temperature values, so as to quantify the uncertainty interval thereof.

[0031] At step 4, a warning is given if the measured well bottom temperature value is continuously outside the uncertainty interval.

[0032] The deep learning-based reservoir monitoring index warning method in the application collects daily well bottom flow pressure values and corresponding maximum and minimum well bottom temperature values of the well bottom flow pressure values; then a convolutional neural network model is established, specifically taking the well bottom flow pressure value as the input value, and the corresponding maximum and minimum well bottom temperature values as the output value, and the Adam optimizer is used for training to obtain a neural network model; finally, the measured well bottom flow pressure value is input into the neural network model to obtain the well bottom temperature value, and a warning is given if the well bottom temperature value is continuously outside the well bottom temperature variation interval.

[0033] The application adopts a data-driven algorithm and does not rely on subjective experience and seepage mechanism, and can automatically learn the well bottom temperature uncertainty interval corresponding to different well bottom flow pressures based on historical data, thereby providing a reliable data-driven basis for reservoir dynamic warning; meanwhile, the deep convolutional neural network structure for predicting the upper and lower bounds of the well bottom temperature in the application has the advantages of good noise resistance and high accuracy compared with the fully connected neural network. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 FIG. 1 is a schematic diagram of fitting of the maximum well bottom temperature historical data based on the convolutional neural network in an embodiment of the application;

[0035] Figure 2 FIG. 2 is a schematic diagram of fitting of the minimum well bottom temperature historical data based on the convolutional neural network in an embodiment of the application;

[0036] Figure 3 FIG. 3 is a schematic diagram of prediction of the upper bound of the well bottom temperature uncertainty interval in an embodiment of the application;

[0037] Figure 4 FIG. 4 is a schematic diagram of prediction of the lower bound of the well bottom temperature uncertainty interval in an embodiment of the application;

[0038] Figure 5 FIG. 5 is a structural schematic diagram of the neural network model in an embodiment of the application;

[0039] Figure 6 FIG. 6 is a flowchart of a specific embodiment of the deep learning-based reservoir monitoring index warning method of the application. DETAILED DESCRIPTION

[0040] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0041] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments consistent with the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0042] The deep learning-based reservoir monitoring index early warning method of the present application comprises:

[0043] Step 1, collect historical data;

[0044] Step 2, establish a convolutional neural network model, with the bottom hole flowing pressure value as the input value, and the corresponding maximum and minimum bottom hole temperature values as the output value;

[0045] Step 3, use the label data corresponding to the maximum and minimum bottom hole temperature values to train two neural networks to obtain two prediction models;

[0046] Step 4, obtain the predicted maximum and minimum bottom hole temperature values according to the prediction model, thereby quantifying the uncertainty interval.

[0047] The following is a specific embodiment of the application

[0048] In a specific embodiment 1 of the application, as shown in Figure 6 , the deep learning-based reservoir monitoring index early warning method of the present application comprises: Figure 6

[0049] S101, collect historical data

[0050] Collect daily oil well bottom hole flowing pressure values and all possible bottom hole temperature values corresponding to the bottom hole flowing pressure values, and obtain three groups of data by taking the maximum value, i.e. bottom hole flowing pressure and corresponding maximum and minimum bottom hole temperature values of the bottom hole flowing pressure, as label data for neural network training;

[0051] ​S102, a convolutional neural network model is established, taking the bottom-hole flowing pressure value as the input value, and the corresponding bottom-hole temperature maximum and minimum value as the output value. The convolutional neural network comprises a plurality of modules, and finally outputs to the required dimension through the fully connected layer after dimension transformation. Each module is composed of a convolutional layer, a normalization layer, a ReLU activation function layer and a maximum pooling layer arranged in sequence. The role of the convolutional layer is to extract data features through convolution operation; the role of the normalization layer is to normalize the network parameters to improve the training stability and convergence; the role of the maximum pooling layer is to reduce dimension to improve the training efficiency. The matrix after the maximum pooling layer is deformed into one dimension, and is output to the dimension of the bottom-hole temperature through the linear layer.

[0052] S103, an Adam optimizer is used for training, and two neural networks are trained respectively by using the label data corresponding to the maximum and minimum values of the bottom-hole temperature to obtain two prediction models:

[0053]

[0054]

[0055]

[0056] Wherein, p wf =p wf (t) is the bottom-hole flowing pressure value, is the approximate value of the maximum bottom-hole temperature, is the approximate value of the minimum bottom-hole temperature,

[0057] f is a neural network operator mapping from the bottom-hole flowing pressure value p wf to the approximate value of the maximum bottom-hole temperature ,

[0058] g is a neural network operator mapping from the bottom-hole flowing pressure value p wf to the approximate value of the minimum bottom-hole temperature ;

[0059] The bottom-hole flowing pressure value corresponds to the bottom-hole temperature variation interval;

[0060] S104, in the prediction, the measured value of the bottom-hole flowing pressure is input into the two neural network models, the predicted maximum and minimum values of the bottom-hole temperature are obtained, and the uncertainty interval thereof is quantified, and if the measured bottom-hole temperature value continuously exceeds the uncertainty interval, a warning is given.

[0061] In a specific embodiment 2 of the application, the deep learning-based reservoir monitoring index early warning method comprises the following steps:

[0062] S1, collecting historical data

[0063] collecting daily oil well bottom flow pressure values and maximum and minimum well bottom temperature values corresponding to the well bottom flow pressure values;

[0064] S2, a convolutional neural network model is established, taking the well bottom flow pressure value as an input value, and the corresponding maximum and minimum well bottom temperature values as output values

[0065] S3, an Adam optimizer is used for training, to obtain two neural network prediction models (as shown in Figure 4

[0066]

[0067]

[0068]

[0069] wherein, p wf = p wf (t) is the well bottom flow pressure value, is an approximation of the maximum well bottom temperature value, as shown in Figure 1 is an approximation of the minimum well bottom temperature value, as shown in Figure 2

[0070] f is a neural network operator mapping from the well bottom flow pressure value p wf to the maximum well bottom temperature value approximation

[0071] g is a neural network operator mapping from the well bottom flow pressure value p wf to the minimum well bottom temperature value approximation

[0072] the well bottom temperature variation interval corresponding to the well bottom flow pressure value, as shown in Figure 3

[0073] S4, the measured well bottom flow pressure value is input into the two neural network models, to obtain the maximum and minimum well bottom temperature values, and if the well bottom temperature value continuously exceeds the well bottom temperature uncertainty interval (i.e., the upper and lower bounds in the figure), a warning is given, as shown in Figure 4

[0074] In a specific embodiment 3 of the application, as shown in Figure 5 ​​​​​​​As shown, the convolutional neural network comprises a plurality of modules. Each module is composed of a convolutional layer, a normalization layer, a ReLU activation function layer and a maximum pooling layer arranged in sequence. The role of the convolutional layer is to extract data features through convolution operation; the role of the normalization layer is to normalize network parameters to improve training stability and convergence; the role of the maximum pooling layer is to reduce dimension to improve training efficiency. The matrix after the maximum pooling layer is deformed into one dimension and output to the dimension of the bottom hole temperature through a linear layer. The neural network input is the bottom hole flow pressure time series data, and the output is the bottom hole temperature maximum or minimum value time series data. The training method is supervised training based on labeled data, and the Adam optimizer and mean square error function are used.

[0075] Finally, it should be noted that: the above only for the preferred embodiments of the present application, and not for limiting the present application, although the foregoing detailed description of the present application is made with reference to the foregoing embodiments, for those skilled in the art, it still can be modified, or part of the technical features of the equivalent replacement. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0076] In addition to the technical features described in the specification, they are known to those skilled in the art.

Claims

1. A deep learning-based reservoir monitoring index early warning method, characterized in that, The deep learning-based oil reservoir monitoring index early warning method comprises: Step 1, collecting historical data; Step 2, establishing a convolutional neural network model, taking the bottom hole flowing pressure value as the input value, and the corresponding maximum and minimum bottom hole temperature values as the output value; Step 3, training two neural networks respectively using the label data corresponding to the maximum and minimum bottom hole temperature values to obtain two prediction models; Step 4, obtaining the predicted maximum and minimum bottom hole temperature values according to the prediction model, thereby quantifying the uncertainty interval thereof; In step 1, the daily oil well bottom hole flowing pressure value and the corresponding maximum and minimum bottom hole temperature values of the bottom hole flowing pressure value are collected; In step 3, the obtained prediction model is: ； ； ； wherein, is the bottom hole flowing pressure value, is the approximate value of the maximum bottom hole temperature, is the approximate value of the minimum bottom hole temperature, f is a function of the bottom hole pressure value to the bottom hole temperature maximum approximation a mapped neural network operator, g is the value of the flowing bottom hole pressure to the approximation of the minimum value of the bottom hole temperature a mapped neural network operator; a bottom hole temperature variation interval corresponding to the bottom hole flow pressure value. 2.The deep learning-based reservoir monitoring index early warning method according to claim 1, characterized in that, In step 1, the daily oil well bottom hole flowing pressure value and all possible bottom hole temperature values corresponding to the bottom hole flowing pressure value are collected, and three groups of data are obtained by taking the maximum value, i.e. the bottom hole flowing pressure and the maximum and minimum bottom hole temperature values corresponding to the bottom hole flowing pressure, as the label data for neural network training. 3.The deep learning-based reservoir monitoring index early warning method according to claim 1, characterized in that, In step 2, the convolutional neural network comprises multiple modules and is finally output to the required dimension by the fully connected layer after dimension transformation.

4. The deep learning-based reservoir monitoring index early warning method according to claim 3, characterized in that, In step 2, each module is composed of a convolution layer, a normalization layer, a ReLU activation function layer and a maximum pooling layer arranged in sequence.

5. The deep learning-based reservoir monitoring index early warning method according to claim 4, characterized in that, In step 2, the convolution layer is used to extract data features through convolution operation; the normalization layer is used to normalize network parameters to improve training stability and convergence; the maximum pooling layer is used to reduce dimension to improve training efficiency; and the matrix after the maximum pooling layer is transformed into one dimension and output to the dimension of the bottom hole temperature through a linear layer. 6.The deep learning based reservoir monitoring index early warning method according to claim 1, characterized in that, In step 3, the Adam optimizer is used for training, and two prediction models are obtained by training two neural networks respectively using the label data corresponding to the maximum and minimum bottom hole temperature values. 7.The deep learning based reservoir monitoring index early warning method according to claim 1, characterized in that, In step 4, during prediction, the measured bottom hole flowing pressure value is input into the two prediction models to obtain the predicted maximum and minimum bottom hole temperature values, thereby quantifying the uncertainty interval thereof. 8.The deep learning-based reservoir monitoring index early warning method according to claim 7, characterized in that, In step 4, if the measured bottom hole temperature value is continuously outside the uncertainty interval, an early warning is performed.

Images