CN106338778A - Shale lithofacies continuity prediction method based on logging information - Google Patents

Abstract

The invention discloses a shale lithofacies continuity prediction method based on logging information. First, the shale lithofacies type is determined based on analysis test of a cored well and slice data, logging response characteristics reflecting the shale lithofacies type are found, and the continuity of shale lithofacies type is predicted by use of an artificial neural network technology based on a logging curve. The problem that the continuity attributes of shale and the difference thereof are easy to lose and difference in shale attribute understanding is caused when the shale lithofacies is established based on rock sample analysis only is overcome. The method of predicting shale lithofacies based on logging information is a low-cost, quick and effective method.

Description

A kind of shale petrofacies continuous prediction method based on well logging information

Technical field

The invention belongs to shale gas development technique field, more particularly, to a kind of shale petrofacies based on well logging information are continuously pre- Survey method.

Background technology

How shale petrofacies divide from origin cause of formation angular divisions, and exploitation interval favourable to shale gas is inapplicable.

Rock sample sample analysis set up shale petrofacies, are easily lost continuity attribute and its difference of shale, lead to page Rock attribute cognitive diversity, be there is also with log interpretation method and judges more, the different explanation personnel's explanation results difference of error Problem.

Content of the invention

It is an object of the invention to provide a kind of shale petrofacies continuous prediction method based on well logging information is it is intended to solve rock Shale petrofacies are set up in the analysis of stone sample, are easily lost continuity attribute and its difference of shale, lead to shale attribute to recognize The problem of difference.

The present invention is achieved in that a kind of shale petrofacies continuous prediction method based on well logging information, described based on survey The shale petrofacies continuous prediction method of well information is to utilize well logging information, using coring well analytical test and thin slice data to page Rock Lithofacies Types are judged, then according to the relation between shale petrofacies and logging response character, with artificial neural network Technology, sets up the model predicting shale Lithofacies Types using log parameter；Space shale Lithofacies Types are entered with Line Continuity prediction.

Further, described comprised the following steps based on the shale petrofacies continuous prediction method of well logging information:

Step 1, analytical test and thin slice data using coring well judge to shale Lithofacies Types, by shale petrofacies It is divided into rich organic matter siliceous shale, rich organic matter carbonate matter shale, rich organic matter clay matter shale, the siliceous page of organic-lean 7 types such as rock, organic-lean's carbonate matter shale, organic-lean's clay matter shale, organic-lean's limestone, with a, b, c, d, e, F, g represent this 7 type, will represent, such as certain sample point is judged to wherein a certain class with 7 dimension row vectors after its quantification, Then the value in this class is 1, and the value in other classes is 0, and for example a certain sample point through judging to belong to a class, is then used Vectorial [1 00000 0] represent, belong to b class, then use vectorial [0 10000 0] to represent；Then by log with The shale Lithofacies Types of corresponding depth are contrasted, and find out several Logging Curves that can reflect shale Lithofacies Types, as The logging response character of prediction shale Lithofacies Types；

Every kind of shale Lithofacies Types are found out with several well logging sample points corresponding, as set up forecast model Practise sample；

Using shale Lithofacies Types and the well-log information corresponding with depth of all sample points, with p log parameter [x₁,x₂,…,x_p]^tAs input, using corresponding shale Lithofacies Types as output, set up the artificial of prediction shale Lithofacies Types Neural network model；

Step 2, with artificial neural network technology, predicts to space shale petrofacies continuity, according to predicting the outcome, draws The shale lithofacies successions figure of this well section, and then determine the favourable exploitation interval of shale gas；

Step 3, according to predicting the outcome, draws the shale lithofacies successions figure of this well section, and then determines the favourable of shale gas Exploitation interval.

Further, described model (taking a class as a example) is:

$f_a\left(x\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{p/2}{\mathrm{\σ}}^p}\frac{1}{m}{\mathrm{\σ}}_{i=1}^m\exp \[-\frac{{(x-x_{ai})}^t(x-x_{ai})}{2{\mathrm{\σ}}^2}\];$

In formula:

M well logging sample point, uses x_ai=[x_ai1,x_ai2,…,x_aip]^tRepresent p log parameter of i-th sample point；

I=mode number；

M=training mode sum；

x_aiI-th training mode of=type a；

σ=smoothing parameter；

The dimension of p=metric space；

X=wants the parameter of certain point of type of prediction.

The shale petrofacies continuous prediction method based on well logging information that the present invention provides, overcomes only according to rock sample sampling Shale petrofacies are set up in analysis, are easily lost continuity attribute and its difference of shale, lead to the problem of shale attribute cognitive diversity. For preferred shale gas, favourable interval and Favorable exploration area have important meaning for the continuity prediction of the shale petrofacies of the present invention, For example it is rich in organic interval and be only favourable interval and block rich in organic block.The knowledge of traditional shale petrofacies It is not the method using core analysis and thin section identification, because coring well is less and coring cost is too high and analytical test item The more thin section identification of mesh again time-consuming thus be unfavorable for shale petrofacies in the vertical and laterally in (plane) prediction, the present invention examines Consider the method that shale petrofacies are predicted using well logging information, be a kind of low cost and fast and effectively method, such as a bite well coring At least want more than 1,000,000 yuan plus analytical test expense, and a bite borehole logging tool expense only needs 100,000 yuan about.The present invention is according to page Relation between rock petrofacies and logging response character, with artificial neural network technology, sets up and predicts shale using log parameter Lithographic model；Space shale petrofacies are entered with Line Continuity prediction, not only greatly improves efficiency, and greatly saved one-tenth This.

Brief description

Fig. 1 is the flow chart of the shale petrofacies continuous prediction method based on well logging information that the present invention implements to provide.

Specific embodiment

In order that the objects, technical solutions and advantages of the present invention become more apparent, with reference to embodiments, to the present invention It is further elaborated.It should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not used to Limit the present invention.

The present invention selects the foundation that the content of organic matter and the big parameter of mineral composition two divide for shale petrofacies, first presses organic matter Content, whether more than 2%, marks off rich organic and organic-lean's shale；Press clay, quartz and feldspar content, carbonate mine again Thing content number set up the criteria for classifying of 7 kinds of shale Lithofacies Types, the present invention utilize well logging information with by petrographical analysis Relation between the shale petrofacies of certification of registered capital material identification, with artificial neural network technology, sets up forecast model, carries out space shale The continuity prediction of petrofacies.Overcome and set up shale petrofacies only according to rock sample sample analysis, be easily lost the continuity of shale Attribute and its difference, thus lead to the difference in shale attribute understanding.

Below in conjunction with Fig. 1, the present invention is further illustrated.

Being comprised the following steps based on the shale petrofacies continuous prediction method of well logging information of the embodiment of the present invention:

Analytical test and thin slice data using coring well judge to shale Lithofacies Types, and shale petrofacies are divided into Rich organic matter siliceous shale, rich organic matter carbonate matter shale, rich organic matter clay matter shale, organic-lean's siliceous shale, lean 7 types such as organic carbonate matter shale, organic-lean's clay matter shale, organic-lean's limestone, in order to describe conveniently, below Represent this 7 type with a, b, c, d, e, f, g, will represent, such as certain sample point is judged to 7 dimension row vectors after its quantification Wherein a certain class, then the value in this class is 1, and the value in other classes is 0, and for example a certain sample point is through judging to belong to In a class, then use vectorial [1 00000 0] to represent, belong to b class, then use vectorial [0 10000 0] to represent ... ..., with This analogizes.

Then log is contrasted with the shale Lithofacies Types of corresponding depth, find out and several can reflect shale petrofacies The Logging Curves (parameter) of type, such as sound wave ac, gamma hcgr, density d en, resistivity rt ... wait p log parameter, make For predicting the logging response character (parameter) of shale Lithofacies Types.Below in order to describe convenience vector x=[x₁,x₂,…,x_p]^t Represent p log parameter.

Every kind of shale Lithofacies Types are found out with several well logging sample points corresponding, as set up forecast model Practise sample, such as a class has m well logging sample point, then uses x_ai=[x_ai1,x_ai2,…,x_aip]^tRepresent the p of i-th sample point of a class Individual log parameter.Then utilize the shale Lithofacies Types of all sample points and well-log information corresponding with depth, logged well with p Parameter [x₁,x₂,…,x_p]^tAs input, using corresponding shale Lithofacies Types as output, predict shale petrofacies class as setting up The learning sample of the artificial nerve network model of type.

Logging response character according to corresponding to different shale Lithofacies Types and distribution, using 7 kinds of shale rocks Learning sample corresponding to facies type sets up the forecast model of shale Lithofacies Types using probabilistic neural network method, as a class Forecast model is:

$f_a\left(x\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{p/2}{\mathrm{\σ}}^p}\frac{1}{m}{\mathrm{\σ}}_{i=1}^m\exp \[-\frac{{(x-x_{ai})}^t(x-x_{ai})}{2{\mathrm{\σ}}^2}\]$

In formula:

I=mode number.

M=training mode sum.

x_aiI-th training mode (sample point) of=type a.

σ=smoothing parameter.

The dimension of p=metric space.

X=wants the parameter of certain point of type of prediction.

The forecast model of b class is:

$f_b\left(x\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{p/2}{\mathrm{\σ}}^p}\frac{1}{m}{\mathrm{\σ}}_{i=1}^m\exp \[-\frac{{(x-x_{bi})}^t(x-x_{bi})}{2{\mathrm{\σ}}^2}\];$

The forecast model of g class is:

$f_g\left(x\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{p/2}{\mathrm{\σ}}^p}\frac{1}{m}{\mathrm{\σ}}_{i=1}^m\exp \[-\frac{{(x-x_{gi})}^t(x-x_{gi})}{2{\mathrm{\σ}}^2}\]$

Will predict that m log parameter of the well section of shale Lithofacies Types substitutes into above-mentioned model one by one by depth order, calculate Go out f_a(x)、f_b(x)、…、f_gX (), finds out maximum 1, if such as f_aX ()=1, then be included into a class by this point, analogized with secondary.

According to predicting the outcome, draw the shale lithofacies successions figure of this well section, and then determine the favourable development layer of shale gas Section.

The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.

Claims (3)

1. a kind of shale petrofacies continuous prediction method based on well logging information is it is characterised in that the described page based on well logging information Rock petrofacies continuous prediction method is first using the analytical test and thin slice data of coring well, shale Lithofacies Types to be judged, so Set up the model predicting shale Lithofacies Types according to logging response character afterwards；With artificial neural network technology, to space shale Lithofacies Types enter Line Continuity prediction.

2. the shale petrofacies continuous prediction method based on well logging information as claimed in claim 1 is it is characterised in that described be based on The shale petrofacies continuous prediction method of well logging information comprises the following steps:

Step one, analytical test and thin slice data using coring well judge to shale Lithofacies Types, and shale petrofacies are drawn It is divided into rich organic matter siliceous shale, rich organic matter carbonate matter shale, rich organic matter clay matter shale, the siliceous page of organic-lean 7 types such as rock, organic-lean's carbonate matter shale, organic-lean's clay matter shale, organic-lean's limestone, with a, b, c, d, e, F, g represent this 7 type, will represent, such as certain sample point is judged to wherein a certain class with 7 dimension row vectors after its quantification, Then the value in this class is 1, and the value in other classes is 0, and for example a certain sample point through judging to belong to a class, is then used Vectorial [1 00000 0] represent, belong to b class, then use vectorial [0 10000 0] to represent；Then by log with The shale Lithofacies Types of corresponding depth are contrasted, and find out several Logging Curves that can reflect shale Lithofacies Types, as The logging response character of prediction shale Lithofacies Types；

Every kind of shale Lithofacies Types are found out corresponding several and are logged well sample points, as setting up forecast model by step 2 Learning sample；

Using shale Lithofacies Types and the well-log information corresponding with depth of all sample points, with p log parameter [x₁, x₂,…,x_p]^tAs input, using corresponding shale Lithofacies Types as output, set up the artificial neuron of prediction shale Lithofacies Types Network model；

Step 3, with artificial neural network technology, predicts to space shale petrofacies continuity, according to predicting the outcome, draws this The shale lithofacies successions figure of well section, and then determine the favourable exploitation interval of shale gas.

3. the shale petrofacies continuous prediction method based on well logging information as claimed in claim 1 is it is characterised in that described a class Model is:

$f_a\left(x\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{p/2}{\mathrm{\σ}}^p}\frac{1}{m}{\mathrm{\σ}}_{i=1}^m\exp \[-\frac{{(x-x_{ai})}^t(x-x_{ai})}{2{\mathrm{\σ}}^2}\];$

In formula:

M well logging sample point, uses x_ai=[x_ai1,x_ai2,…,x_aip]^tRepresent p log parameter of i-th sample point；

I=mode number；

M=training mode sum；

x_aiI-th training mode of=type a；

σ=smoothing parameter；

The dimension of p=metric space；

X=wants the parameter of certain point of type of prediction；

The forecast model of b class is:

$f_b\left(x\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{p/2}{\mathrm{\σ}}^p}\frac{1}{m}{\mathrm{\σ}}_{i=1}^m\exp \[-\frac{{(x-x_{bi})}^t(x-x_{bir})}{2{\mathrm{\σ}}^2}\];$

The forecast model of g class is:

$f_g\left(x\right)=\frac{1}{{\left(2\mathrm{\π}\right)}^{p/2}{\mathrm{\σ}}^p}\frac{1}{m}{\mathrm{\σ}}_{i=1}^m\exp \[-\frac{{(x-x_{gi})}^t(x-x_{gi})}{2{\mathrm{\σ}}^2}\].$