CN106338778B - A kind of shale petrofacies continuous prediction method based on well logging information - Google Patents

Abstract

The invention discloses a method for continuously predicting shale lithofacies based on well logging information. The method for continuously predicting shale lithofacies based on well logging information is to first use the analysis and testing of core wells and thin-section data to analyze the shale lithofacies. Determine the type of shale, find out the logging response characteristics that can reflect the type of shale lithofacies, and use artificial neural network technology to predict the continuity of shale lithofacies types by using logging curves. The invention overcomes the problem that the shale lithofacies is easily lost based on rock sample sampling and analysis, and the continuity attribute and its difference of the shale are easily lost, resulting in differences in understanding of shale attributes; the method of predicting shale lithofacies by using logging information is A low-cost, fast and effective method.

Description

A continuous prediction method for shale lithofacies based on logging information

technical field

The invention belongs to the technical field of shale gas development, in particular to a method for continuously predicting shale lithofacies based on logging information.

Background technique

The division of shale lithofacies is mostly done from the perspective of genesis, which is not applicable to intervals that are favorable for the development of shale gas.

Rock sample sampling analysis to establish shale lithofacies is easy to lose the continuity attributes and differences of shale, resulting in differences in the understanding of shale attributes. There are also many judgment errors and differences in interpretation results by different interpreters when using logging interpretation methods .

Contents of the invention

The purpose of the present invention is to provide a continuous prediction method for shale lithofacies based on well logging information, aiming at solving the problem of establishing shale lithofacies through sampling analysis of rock samples, which is easy to lose the continuity attribute and its difference of shale, resulting in shale attribute The problem of recognizing differences.

The present invention is realized in this way, a method for continuously predicting shale lithofacies based on well logging information. The method for continuously predicting shale lithofacies based on well logging information is to utilize the well logging information and adopt the analytical tests and The type of shale lithofacies is judged by thin section data, and then according to the relationship between shale lithofacies and logging response characteristics, artificial neural network technology is used to establish a model for predicting shale lithofacies types by using logging parameters; Continuous prediction of lithofacies types.

Further, the method for continuously predicting shale lithofacies based on logging information includes the following steps:

Step 1: Determine the type of shale lithofacies using the analytical tests and thin-section data of the core well, and divide the shale lithofacies into organic-rich siliceous shale, organic-rich carbonate shale, and organic-rich clay shale. There are seven types of rock, organic-poor siliceous shale, organic-poor carbonate shale, organic-poor clay shale, and organic-poor limestone, which are represented by A, B, C, D, E, F, and G. There are 7 types, which are quantified and represented by 7-dimensional row vectors. If a sampling point is determined to be one of the categories, the value of this category is 1, while the value of other categories is 0 , for example, if a certain sampling point is judged to belong to class A, it will be represented by vector [1 0 0 0 0 0 0], and if it belongs to class B, it will be represented by vector [0 1 0 0 0 0 0]; then the log curve and Compare the shale lithofacies types at corresponding depths, and find several conventional logging curves that can reflect the shale lithofacies types, as the logging response characteristics for predicting shale lithofacies types;

For each type of shale lithofacies, find several corresponding well logging sampling points as learning samples for establishing the prediction model;

Using the shale lithofacies types of all sampling points and the logging data corresponding to the depth, P logging parameters [X ₁ ,X ₂ ,…,X _P ] ^T are used as input, and the corresponding shale lithofacies types are used as Output, establish artificial neural network model to predict shale lithofacies type;

Step 2: Use artificial neural network technology to predict the continuity of spatial shale lithofacies, draw the shale lithofacies profile of the well section according to the prediction results, and then determine the favorable development intervals of shale gas;

Step 3, draw the shale lithofacies profile of the well section according to the prediction results, and then determine the favorable intervals for shale gas development.

Further, the model (taking Class A as an example) is:

In the formula:

For m logging sampling points, use X _Ai = [X _Ai1 , X _Ai2 ,..., X _Aip ] ^T to represent the P logging parameters of the i-th sampling point;

i = mode number;

m = total number of training patterns;

X _Ai = the i-th training pattern of type A;

σ = smoothing parameter;

P = dimensionality of the metric space;

X = parameter at a certain point of the type to be predicted.

The continuous prediction method of shale lithofacies based on well logging information provided by the present invention overcomes the problem that the shale lithofacies is established solely based on rock sample sampling and analysis, which easily loses the continuity attribute and its difference of shale, which leads to differences in understanding of shale attributes question. The continuity prediction of shale lithofacies in the present invention is of great significance for optimizing favorable intervals and exploration blocks of shale gas, for example, intervals rich in organic matter and blocks rich in organic matter are favorable intervals and blocks. The traditional identification of shale lithofacies is by core analysis and thin section identification. Because there are few core wells, the cost of coring is too high, and there are many analytical and laboratory items, thin section identification is time-consuming, so it is not conducive to the longitudinal identification of shale lithofacies. For prediction on the upper and lateral (plane), the present invention considers the method of using logging information to predict shale lithofacies, which is a low-cost, fast and effective method. For example, a well coring plus analysis and testing costs at least More than 1 million yuan, while the cost of logging a well is only about 100,000 yuan. According to the relationship between shale lithofacies and logging response characteristics, the present invention uses artificial neural network technology to establish a shale lithofacies prediction model using logging parameters; to predict the continuity of spatial shale lithofacies, not only greatly Efficiency is improved, and cost is greatly saved.

Description of drawings

Fig. 1 is a flow chart of the continuous prediction method of shale lithofacies based on well logging information provided by the implementation of the present invention.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the examples. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

In the present invention, the two parameters of organic matter content and mineral composition are selected as the basis for the division of shale lithofacies. Firstly, organic matter-rich and organic matter-poor shale is divided according to whether the organic matter content exceeds 2%; How much salt mineral content to establish the classification standard of 7 kinds of shale lithofacies types, the present invention utilizes the relationship between logging information and shale lithofacies identified through rock analysis test data, and uses artificial neural network technology to establish a prediction model , to carry out the continuity prediction of spatial shale lithofacies. Overcoming the establishment of shale lithofacies based on rock sample sampling and analysis, it is easy to lose the continuity attributes and differences of shale, which leads to differences in the understanding of shale attributes.

The present invention will be further described below in conjunction with FIG. 1 .

The method for continuously predicting shale lithofacies based on well logging information in the embodiment of the present invention includes the following steps:

The shale lithofacies types were judged by the analytical tests and thin section data of cored wells, and the shale lithofacies were divided into organic-rich siliceous shale, organic-rich carbonate shale, organic-rich clay shale, and organic-poor clay shale. There are seven types of organic siliceous shale, organic-poor carbonate shale, organic-poor clay shale, and organic-poor limestone. For the convenience of description, A, B, C, D, E, F, and G are used below Represent these 7 types, quantify them and express them with 7-dimensional row vectors. If a certain sampling point is determined to be one of the categories, the value of this category is 1, while the value of other categories is is 0, for example, if a certain sampling point is judged to belong to class A, it will be represented by vector [1 0 0 0 0 0 0], and if it belongs to class B, it will be represented by vector [0 1 0 0 0 0 0], ..., with And so on.

Then compare the logging curve with the shale lithofacies type at the corresponding depth to find several conventional logging curves (parameters) that can reflect the shale lithofacies type, such as acoustic wave AC, gamma HCGR, density DEN, resistivity P logging parameters such as RT... are used as logging response features (parameters) for predicting shale lithofacies types. For the convenience of description, the vector X=[X ₁ ,X ₂ ,...,X _P ] ^T is used to represent the P logging parameters.

For each shale lithofacies type, find out several corresponding well logging sampling points as learning samples for establishing a prediction model. If there are m logging sampling points in type A, use X _Ai = [X _Ai1 , X _Ai2 ,...,X _Aip ] ^T represents the P logging parameters of the i-th sampling point of type A. Then, using the shale lithofacies types of all sampling points and the logging data corresponding to the depth, P logging parameters [X ₁ ,X ₂ ,…,X _P ] ^T are used as input, and the corresponding shale lithofacies types As an output, it is used as a learning sample for establishing an artificial neural network model for predicting shale lithofacies types.

According to the logging response characteristics and distribution ranges corresponding to different shale lithofacies types, use the learning samples corresponding to 7 shale lithofacies types to establish a prediction model of shale lithofacies types with the method of probability neural network, such as type A The prediction model for is:

In the formula:

i = mode number.

m = total number of training patterns.

X _Ai =i-th training pattern (sampling point) of type A.

σ = smoothing parameter.

P = dimensionality of the metric space.

X = parameter at a certain point of the type to be predicted.

The prediction model for class B is:

The prediction model of class G is:

Substituting the M logging parameters of the well section to predict the type of shale lithofacies into the above model one by one in order of depth, calculate f _A (X), f _B (X), ..., f _G (X), and find the maximum value 1 , for example, if f _A (x)=1, then the point is classified into class A, and so on.

According to the prediction results, the shale lithofacies profile of the well section is drawn, and then the favorable intervals for shale gas development are determined.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. within range.

Claims (2)

1. A continuous prediction method for shale lithofacies based on well logging information, characterized in that, the continuous prediction method for shale lithofacies based on well logging information is to first use the analysis and testing of core wells and thin section data to analyze the shale Determine the type of lithofacies, and then establish a model to predict the type of shale lithofacies based on the logging response characteristics; use artificial neural network technology to continuously predict the type of lithofacies in space; The method for continuously predicting shale lithofacies based on logging information includes the following steps: Step 1: Determine the type of shale lithofacies by using the analytical tests and thin-section data of cored wells, and divide the shale lithofacies into organic-rich siliceous shale, organic-rich carbonate shale, and organic-rich clay shale. rock, organic-poor siliceous shale, organic-poor carbonate shale, organic-poor clay shale, and organic-poor limestone. After quantifying it, it is represented by a 7-dimensional row vector. If a certain sampling point is determined to be one of the categories, the value of this category is 1, while the value of other categories is 0. When a certain sampling point is determined to belong to category A, it is represented by vector [1 0 0 0 0 0 0], and if it belongs to category B, it is represented by vector [0 1 0 0 0 0 0]; Compare the depth of shale lithofacies types, and find several conventional logging curves that can reflect the shale lithofacies types, as the logging response characteristics for predicting shale lithofacies types; Step 2, for each type of shale lithofacies, find a number of corresponding logging sampling points as learning samples for establishing a prediction model; Using the shale lithofacies types of all sampling points and the logging data corresponding to the depth, P logging parameters [X ₁ ,X ₂ ,…,X _P ] ^T are used as input, and the corresponding shale lithofacies types are used as Output, establish artificial neural network model to predict shale lithofacies type; Step 3: Use artificial neural network technology to predict the continuity of spatial shale lithofacies, draw the shale lithofacies profile of the well section according to the prediction results, and then determine the favorable intervals for shale gas development. 2. the shale lithofacies continuous prediction method based on logging information as claimed in claim 1, is characterized in that, the model of described A class is: <mrow><msub><mi>f</mi><mi>A</mi></msub><mrow><mo>(</mo><mi>X</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><msup><mrow><mo>(</mo><mn>2</mn><mi>&amp;pi;</mi><mo>)</mo></mrow><mrow><mi>p</mi><mo>/</mo><mn>2</mn></mrow></msup><msup><mi>&amp;sigma;</mi><mi>p</mi></msup></mrow></mfrac><mfrac><mn>1</mn><mi>m</mi></mfrac><msubsup><mi>&amp;Sigma;</mi><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>m</mi></msubsup><mi>exp</mi><mo>&amp;lsqb;</mo><mo>-</mo><mfrac><mrow><msup><mrow><mo>(</mo><mi>X</mi><mo>-</mo><msub><mi>X</mi><mrow><mi>A</mi><mi>i</mi></mrow></msub><mo>)</mo></mrow><mi>T</mi></msup><mrow><mo>(</mo><mi>X</mi><mo>-</mo><msub><mi>X</mi><mrow><mi>A</mi><mi>i</mi></mrow></msub><mo>)</mo></mrow></mrow><mrow><mn>2</mn><msup><mi>&amp;sigma;</mi><mn>2</mn></msup></mrow></mfrac><mo>&amp;rsqb;</mo><mo>;</mo></mrow> In the formula: M is the logging sampling point, and X _Ai = [X _Ai1 , X _Ai2 ,..., X _Aip ] ^T represents the P logging parameters of the i-th sampling point; i = mode number; m = total number of training patterns; X _Ai = the i-th training pattern of type A; σ = smoothing parameter; P = dimensionality of the metric space; X = the parameter of a certain point of the type to be predicted; The model of class B is: <mrow><msub><mi>f</mi><mi>B</mi></msub><mrow><mo>(</mo><mi>X</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><msup><mrow><mo>(</mo><mn>2</mn><mi>&amp;pi;</mi><mo>)</mo></mrow><mrow><mi>p</mi><mo>/</mo><mn>2</mn></mrow></msup><msup><mi>&amp;sigma;</mi><mi>p</mi></msup></mrow></mfrac><mfrac><mn>1</mn><mi>m</mi></mfrac><msubsup><mi>&amp;Sigma;</mi><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>m</mi></msubsup><mi>exp</mi><mo>&amp;lsqb;</mo><mo>-</mo><mfrac><mrow><msup><mrow><mo>(</mo><mi>X</mi><mo>-</mo><msub><mi>X</mi><mrow><mi>B</mi><mi>i</mi></mrow></msub><mo>)</mo></mrow><mi>T</mi></msup><mrow><mo>(</mo><mi>X</mi><mo>-</mo><msub><mi>X</mi><mrow><mi>B</mi><mi>i</mi></mrow></msub><mo>)</mo></mrow></mrow><mrow><mn>2</mn><msup><mi>&amp;sigma;</mi><mn>2</mn></msup></mrow></mfrac><mo>&amp;rsqb;</mo><mo>;</mo></mrow> The model of class G is: <mrow><msub><mi>f</mi><mi>G</mi></msub><mrow><mo>(</mo><mi>X</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><msup><mrow><mo>(</mo><mn>2</mn><mi>&amp;pi;</mi><mo>)</mo></mrow><mrow><mi>p</mi><mo>/</mo><mn>2</mn></mrow></msup><msup><mi>&amp;sigma;</mi><mi>p</mi></msup></mrow></mfrac><mfrac><mn>1</mn><mi>m</mi></mfrac><msubsup><mi>&amp;Sigma;</mi><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>m</mi></msubsup><mi>exp</mi><mo>&amp;lsqb;</mo><mo>-</mo><mfrac><mrow><msup><mrow><mo>(</mo><mi>X</mi><mo>-</mo><msub><mi>X</mi><mrow><mi>G</mi><mi>i</mi></mrow></msub><mo>)</mo></mrow><mi>T</mi></msup><mrow><mo>(</mo><mi>X</mi><mo>-</mo><msub><mi>X</mi><mrow><mi>G</mi><mi>i</mi></mrow></msub><mo>)</mo></mrow></mrow><mrow><mn>2</mn><msup><mi>&amp;sigma;</mi><mn>2</mn></msup></mrow></mfrac><mo>&amp;rsqb;</mo><mo>.</mo></mrow> 2