CN104880737A - Multivariate Logistic method using logging information to identify type of underground fluid - Google Patents

Abstract

The invention relates to a technology of identifying the type of a fluid in an oil and gas field reservoir, particularly to a method which uses a Logistic model discriminant of logging information as fingerprint features to perform identification of the type of a reservoir fluid, thereby accurately and rapidly dividing fluid units, predicting spatial distribution of a favorable reservoir, and further guiding exploration and development of oil gas. The invention aims to overcome the defects that in the prior art, the search for offlap foreset reflection is carried out only by using seismic data so as to achieve the search for a descent system tract, and the seismic data with a relatively low resolution ratio cannot be identified easily, rapidly and accurately, and to provide a descent system tract identification technology which is accurate, simple and convenient, practical, and easy to popularize. An analysis of a seismic section is made with combination of a new method and a classical means, and division of marine strata and predication of oil-gas reservoir strata can be carried out more accurately and more rapidly.

Description

The polynary Logistic method of well-log information identification underground fluid type

Technical field:

The present invention relates to a kind of oil gas field fluid type of reservoir through recognition technology, in particular to the discriminating utilizing the Logistic Model checking formula of well-log information to carry out fluid type of reservoir through as fingerprint characteristic, thus divide fluid Dan Yuan quickly and accurately, the spatial of prediction Favorable Reservoir, and then instruct a kind of method of the exploration and development of oil gas.

Background technology:

In recent years, the scientific-technical progress of petroleum exploration and development is constantly expanded at theoretical and technical depth and broadness with development, the specific type complex hydrocarbon such as complex lithology, low porosity and low permeability, the low resistitvity reservoir status ensconced in petroleum vapour exploration becomes and becomes more and more important, and the complicated reservoirs flow net model problem that attention rate is in recent years higher especially.All the time, the profit identification of the complicated reservoirs such as complex lithology, low hole, hypotonic, low-resistance reservoir is the important research direction of well logging interpretation.For complicated reservoirs, the response of reservoir fluid in above-mentioned logging method is subject to the impact of formation lithology, pore texture and intrusion etc. factor, response characteristic is obvious not, and the basis thus differentiating profit method at conventional logging being applied the difference that multiple mathematical measure deeply excavates complicated reservoirs oil-water-layer becomes important development trend.

Logistic curvilinear equation is derived when studying population and increasing in 1838 by mathematician, biologist P.F.Verhulst, R.Pearl and L.T.Reed rediscovered and obtained broader applications the twenties in 20th century.Current Logistic curve is widely used in social science and natural science field.Logistic regression model has following advantage: when carrying out Logistic and returning discriminatory analysis, and require not tight to the normality of data and the hypothesis of equal covariance matrix, the result obtained is but very stable; Logistic regression class is similar to regretional analysis, has direct statistical test, comprises non-linear effect and diagnosis on a large scale; Logistic returns the not special requirement of independent variable, and independent variable can be discrete variable also can be continuous variable; The dependent variable that Logistic returns is classified variable, predict that the result obtained is the probability that event occurs, and precision of prediction is high.

Traditional crossplot method of identification has very large limitation, and plate identification can only utilize two independents variable to carry out intersection at every turn, and same plate is difficult to identify and distinguish all dependent variable types.Meanwhile, the chart method that crosses requires that independent variable is continuous variable, and some discrete argument data, as lithology data, then at the Taxonomic discussion of having in chart method that crosses.The more important thing is, chart method has very large artificial subjectivity.This is invented with factor of porosity, permeability, water saturation, shale index and lithology etc. as independent variable, with reservoir fluid classification for dependent variable, set up polynary Logistic model, be expected to realize easier with identify efficiently, improve the accuracy rate of identification simultaneously, have no granted patent at home and abroad.

Summary of the invention:

The object of the invention is to overcome and existingly utilize that crossplot method can only consider two kinds of parameters, same plate is difficult to distinguish multiple dependent variable type, independent variable must be continuous variable and the inferior position such as artificial subjective factor is larger simultaneously, make up the deficiency be difficult to complicated reservoirs fluid type identifications such as complex lithology, low hole, hypotonic, low-resistance reservoir, provide a kind of accurate, easy, practical and be easy to the fluid type of reservoir through recognition technology promoted.

Method of the present invention changes traditional utilize crossplot method and the normal linear Return Law thus identify the method for the oil gas water type of complicated reservoirs.First new method carries out pre-service to logging trace, and core Location is corrected, and sets up various parameter model, and solves various parameter.Take field data as dependent variable again, with above-mentioned parameters for independent variable, carry out Logistic regretional analysis, set up the fingerprint recognition discriminant of oil-gas-water layer, as shown in Figure 1.

Gordian technique main points comprise:

(1) polynary many-valued Logistic analyzes

In the reservoir of complex geologic conditions, be difficult to a possible outcome for one or two kind of well logging interpretation parameter supposition fluid type of reservoir through.More then be by multiple log parameter, as factor of porosity, permeability, lithology, shale index and water saturation etc. are all included in unified supposition discrimination model, greatly improve precision and the accuracy of well logging interpretation.Such analysis is multivariate analysis.Analyze polynary Logistic, the simplest one is that polynary two-value Logistic analyzes.Suppose that dependent variable has two kinds of values (0 and 1), an existing n independent variable, is respectively x1, x2, x3...xn, and probability when so independent variable value is 1 is:

$p=p(y=1|x1,x2,x3...,\mathrm{xn})=\frac{e(\mathrm{\α}+\mathrm{\β}1*x1+\mathrm{\β}2*x2+\mathrm{\β}3*x3...+\mathrm{\βn}*\mathrm{xn})}{1+e(\mathrm{\α}+\mathrm{\β}1*x1+\mathrm{\β}2*x2+\mathrm{\β}3*x3...+\mathrm{\βn}*\mathrm{xn})}$

Its another form:

$\log \mathrm{it}\left(p\right)=\ln \left(\frac{p1}{1-p1}\right)=\mathrm{\α}+\mathrm{\β}1*x1+\mathrm{\β}2*x2+\mathrm{\β}3*x3...+\mathrm{\β}n*\mathrm{xn}$

In formula, c is constant term, and β i (i=1,2...n) is partial regression coefficient.

In fact, the fluid type in reservoir is generally more than two kinds, and namely the value of dependent variable is many-valued.Dependent variable is had to the n meta-model of k kind value, then has:

$\log \mathrm{it}\left(p1\right)=\ln \left(\frac{p1}{1-p1}\right)=\mathrm{\α}1+\mathrm{\β}11*x1+\mathrm{\β}12*x2+\mathrm{\β}13*x3...+\mathrm{\β}1n*\mathrm{xn}$

$\log \mathrm{it}\left(p2\right)=\ln \left(\frac{p2}{1-p2}\right)=\mathrm{\α}2+\mathrm{\β}12*x1+\mathrm{\β}22*x2+\mathrm{\β}32*x3...+\mathrm{\βn}2*\mathrm{xn}$

...

$\log \mathrm{it}\left(\mathrm{pk}\right)=\ln \left(\frac{\mathrm{pk}}{1-\mathrm{pk}}\right)=\mathrm{\αk}+\mathrm{\β}1k*x1+\mathrm{\β}2k*x2+\mathrm{\β}3k*x3...+\mathrm{\βnk}*\mathrm{xn}$

In formula, cj (j=1,2...k) is constant term matrix, β ij (i=1,2...n; J=1,2...k) be partial regression coefficient matrix.

(2) multicollinearity is avoided

Multicollinearity (Multicollinearity) refers to and makes model distortion estimator owing to there is accurate correlationship or height correlation relation between the explanatory variable in linear regression model (LRM) or be difficult to estimate accurately.Choosing of independent variable should be avoided occurring multicollinearity, as namely natural gamma and shale index should not appear at (function that the latter is the former) in same model.Also collinearity inspection should be carried out in time to the model drawn.

Accompanying drawing illustrates:

The polynary Logistic method flow diagram of Fig. 1 well-log information identification underground fluid type

Embodiment:

One, select rational logging trace data, formation testing or the means of production, log data and core test data: the kind of logging trace is complete as much as possible, the requirement that the parameter models such as shale index, water saturation, porosity and permeability are set up can be met.

Two, pre-service (as inclined shaft alignment, curve smoothing process, environmental correction etc.) is carried out to logging trace, and core Location is corrected, set up the corresponding relation with logging trace, and then determine the empirical constant solving the parameter models such as shale index, water saturation, porosity and permeability (as A Erqi interpretation model) etc., draw the parameters such as the shale index of objective interval, water saturation, porosity and permeability.

Three, with the oil-gas-water layer type in formation testing or the means of production for dependent variable, with parameters such as shale index, water saturation, porosity and permeabilities for independent variable, carry out Logistic regretional analysis, set up the fingerprint recognition discriminant of oil-gas-water layer; Repeatedly verify with formation testing or production data and improve discriminant.

Claims (3)

1. the present invention relates to the recognition technology of fluid type in a kind of oil gas field reservoir well logging interpretation field, Multivariate Logistic Regression is carried out by variablees such as factor of porosity, permeability, shale index and water saturation, set up the fingerprint characteristic of fluid type of reservoir through, and then quick and precisely differentiate the method for oil gas water:

Step one, select rational logging trace data, formation testing or the means of production, log data and core test data: the kind of logging trace is complete as much as possible, the requirement that the parameter models such as shale index, water saturation, porosity and permeability are set up can be met.

Step 2, pre-service (as inclined shaft alignment, curve smoothing process, environmental correction etc.) is carried out to logging trace, and core Location is corrected, set up the corresponding relation with logging trace, and then determine the empirical constant solving the parameter models such as shale index, water saturation, porosity and permeability (as A Erqi interpretation model) etc., draw the parameters such as the shale index of objective interval, water saturation, porosity and permeability.

Step 3, with the oil-gas-water layer type in formation testing or the means of production for dependent variable, with parameters such as shale index, water saturation, porosity and permeabilities for independent variable, carries out Logistic regretional analysis, sets up the fingerprint recognition discriminant of fluid type of reservoir through; Repeatedly verify with formation testing or production data and improve discriminant.

2. method for judging fluid type of reservoir through according to claim 1, is characterized in that described fluid type of reservoir through refers to that oil reservoir in subsurface reservoir, water layer, gas-bearing formation, dried layer, oil-containing water layer, gassiness water layer, oil-water common-layer and air water are with types such as layers.

3. oil-gas-water layer method of discrimination according to claim 1, it is characterized in that judging that each parameter of fluid type of reservoir through can integrated use, the type being difficult to distinguish in low dimensional space is positioned over high-dimensional space, and then more easily distinguishes and set up fingerprint characteristic discriminant.