CN102455438A - Method for predicting volume of carbonate rock fractured cave type reservoir - Google Patents

Abstract

The invention relates to a method for predicting the volume of a carbonate rock fractured cave type reservoir. The method comprises the following steps of: acquiring the seismic attribute volume of a tested fractured cave through seismic wave; and predicting the volume of the tested fractured cave according to the product of the seismic attribute volume and a correction coefficient. When the volume predicting method is put into industrial application, the recovery rate of the carbonate rock reservoir is improved, and the yield is quickly increased.

Description

Carbonatite seam hole type reservoir volume Forecasting Methodology

Technical field

The present invention relates to the oil geology exploitation, be specifically related to a kind of carbonatite seam hole type reservoir volume Forecasting Methodology.

Background technology

The main body of system in Tahe Oilfield is an ORDOVICIAN CARBONATE seam hole type oil reservoir, and the The main reservoir type is seam hole type reservoir, and reservoir bodies receives karst, Crack Control; Complex shape; Strong to nonuniformity in length and breadth, to bury deeply, seismic reflection signals resolution is low; And receive the influence of many factors, its reservoir prediction difficulty is very big.Along with the discovery and the expansion of system in Tahe Oilfield, RESERVOIR RECOGNITION and forecasting techniques have also been passed through simple, the single method by morning, to complicated, multi-disciplinary reservoir technical method series of later stage.Main method for predicting reservoir has: amplitude change rate technology, meticulous coherent technique, wave impedance inversion technique, waveform analysis techniques, forward modeling technology etc.; Some other method is like: three-dimensional multiscale analysis technology, palaeogeomorphology palaeodrainage pattern analytical technology, frequency splitting technology, multiparameter cluster analysis technology or the like, and the type oil reservoir has obtained using preferably in ORDOVICIAN CARBONATE seam hole.But this a series of technology mainly is based on plane reservoir prediction qualitatively, and just to the prediction in type reservoir development zone, carbonatite seam hole, stitches the quantitative notion of the big small-scale of hole body and lack.

On the other hand; Stitch the calculation of reserves of hole type oil reservoir at present for carbonatite; Go back neither one method very accurately and effectively both at home and abroad, still adopt the volumetric method of effective reservoir flattening-out to calculate basically at present, this itself does not meet the characteristics of nonuniformity carbonatite seam hole type oil reservoir.

Summary of the invention

The technical issues that need to address of the present invention are, how a kind of carbonatite seam hole type reservoir volume Forecasting Methodology is provided, can quantitative forecast seam hole type reservoir volume, the foundation of an amount is provided for the calculating of petroleum-in-place.

Technical matters of the present invention solves like this: make up a kind of carbonatite seam hole type reservoir volume Forecasting Methodology, may further comprise the steps:

1.1) obtain the earthquake attribute volume volume that the hole body is stitched in test through seismic event;

1.2) multiply by correction coefficient according to said earthquake attribute volume volume and predict said test seam hole body volume.

According to volume Forecasting Methodology provided by the invention, said correction coefficient equals width correction coefficient Ccw and height correction coefficient Cch is long-pending.

According to volume Forecasting Methodology provided by the invention, said correction coefficient can be the constant normal correction coefficient of relative one type of seismic properties.

According to volume Forecasting Methodology provided by the invention, said correction coefficient is obtained like this:

2.1) just drilling to blow out through the numerical simulation of a large amount of seam hole modellings and wave equation and align the data of drilling and carry out overlap-add procedure, STACK DATA is carried out attributes extraction;

2.2) through the volume and the statistics of earthquake attribute volume volume of a large amount of seams hole phantom type, obtain the earthquake attribute volume volume and stitch between the body volume of hole than positive coefficient.

According to volume Forecasting Methodology provided by the invention; Said correction coefficient can be the change correction coefficient that changes with earthquake attribute abnormal apparent volume, and sets: said change correction coefficient becomes linear dependence (using linear function fit) or non-linear dependence (using the nonlinear function match) with said apparent volume.

Carbonatite provided by the invention stitches hole type reservoir volume Forecasting Methodology, is applied to main force's block 2～8 districts of system in Tahe Oilfield, has obtained effect preferably.137 seams unit, hole, 2～8 districts has been carried out the reflection strength body engraving of three-dimensional visualization; Through " carbonatite seam hole type reservoir model is just being drilled and the volume correction technology " obtain than positive coefficient; The reflection strength body is proofreaied and correct; Obtain the volume of effective reservoir bodies of 137 seams unit, hole, 2～8 districts, the volumetric parameter of reservoir bodies is provided for the later stage calculation of reserves.

Description of drawings

Fig. 1 is the unusual apparent volume correction coefficient of a seismic properties of the present invention constant current journey really synoptic diagram;

Fig. 2 is with linear function fit correction coefficient synoptic diagram;

Fig. 3 is the correction coefficient template curve synoptic diagram that probabilistic neural network obtains.

Embodiment

At first, ultimate principle of the present invention is described:

Since Ordovician of Tahe oil main force payzone buried depth big (greater than 5000m), after seismic event arrives zone of interest, frequencies go lower, the resolution characteristic of seismic event is lower, and the system in Tahe Oilfield drilling well discloses the cave height generally between 1-10m, less than the distinguishable 50m of seismic event; But investigate from being generally less than the width of seismic event resolution 96m from modern karst.Therefore can't on time and road number, differentiate seam hole body; And in resolution; Reflected energy is approximated to direct ratio with the size of seam hole body; Therefore adopt the calculating of the unusual apparent volume correction coefficient of seismic properties, seismic properties is meant the seam hole unit exception body of confirming (prediction) from earthquake attribute volume unusually, and earthquake attribute volume can be the attribute volume that extracts from real seismic record or theoretical model carried out the attribute volume that composite traces that the earthquake forward simulation obtains extracts.

According to definition; It is meant the coefficient that the equivalent width of anomalous body on the seismic properties is proofreaied and correct the unusual width correction coefficient of seismic properties (Ccw); For the corresponding seismic properties of the theogram of theory seam hole model, the ratio of the width that is meant theoretical seam hole model and the width of the anomalous body of the seismic properties of corresponding theogram.In like manner; It is meant the coefficient that the equivalent height of anomalous body on the seismic properties is proofreaied and correct the unusual height correction coefficient of seismic properties (Cch); For the corresponding seismic properties of the theogram of theory seam hole model, the ratio of the height that is meant theoretical seam hole model and the height of the anomalous body of the seismic properties of corresponding theogram.Different seismic properties types has different widths and height correction coefficient; The seismic properties abnormal space volume that different volumes is corresponding different with difform seam hole model also promptly has different correction coefficient, therefore; Must design a series of seams hole theoretical model; Calculate a large amount of correction coefficient, therefrom statistics is tried to achieve and the unusual corresponding correction coefficient template of different seismic properties, i.e. height and width correction coefficient template.Correction coefficient equals the width correction coefficient and the height correction coefficient is long-pending.

In second step, it is crucial to specify the present invention:

As shown in Figure 1, through a large amount of seam hole Model Design, just drilling through the numerical simulation seismic wave field of wave equation and to blow out; Align the data of drilling and carry out overlap-add procedure; STACK DATA is carried out seismic properties extract,, obtain the correction relationship between seismic properties and the seam hole body volume through the volume of a large amount of seams hole phantom type and the statistics of earthquake attribute volume volume; With this correction relationship; Be applied in the actual seismic data, just can calculate the active volume of actual seismic data seam hole body, for calculation of reserves provides basic data.

In the 3rd step, practical implementation of the present invention is described:

1. adopt the method for man-machine interaction to set up seismogeology model (desirable solution cavity model and set up seam hole reservoir model two type according to crossing the well seismic section) based on the purpose of the method for building up of seam hole model and modelling; Seam hole model to being designed carries out grid discretization; Adopt then near recording geometry and the acquisition parameter of actual field acquisition and carry out Wave equation forward modeling, the big gun collection is write down carry out conventional processing and pre-stack time migration obtains migrated section at last.

2. extract seismic properties such as seismic reflection information incoherentness, amplitude change rate, time-frequency information, instantaneous amplitude, wave impedance

3. adopt mode identification method to carry out the seismic properties reservoir prediction,, therefrom extract and four responsive seismic properties of seam hole quantification: reflection strength, wave impedance, RMS amplitude, amplitude change rate through seismic properties optimization.

4. for the unit, seam hole of a regular shape, the space spread shape of the attribute volume in the corresponding theogram is not a rule body, but an obscurity boundary and a bigger irregular body of metamorphosis.After seam unit center position, hole is confirmed; Must confirm the equivalent height (time orientation) and the equivalent width (road or line direction) of seam unit, hole; Adopt border search method to confirm the border of seam unit, hole, and then confirm the equivalent height and the equivalent width of seam unit, hole, concrete steps are following.

(1) with seam hole unit center time corresponding and road (t _m, x _m) be the center, at [x _m-Δ X, x _m+ Δ X] in the road window ranges, at [t _m-Δ T, t _m+ Δ T] time window scope in, new property value is carried out two dimension median filter, eliminate because the shake of the property value that random disturbance causes;

(2) based on the result of medium filtering, with seam hole unit center time corresponding and road (t _m, x _m) be the center, at [t _m-Δ T, x _m-Δ X]～[t _m+ Δ T, x _m+ Δ X] in the two-dimentional window ranges, apparent dip step delta p increases progressively with the time, and its unit is a millisecond/rice, and the mode of in the inclination angle scope that the user sets, carrying out linear search is calculated p to should the inclination angle time _jFrontier point, also promptly begin from central point along this inclination direction, judge the roads of property value to two by the road greater than the attribute threshold value, calculate road number greater than the attribute threshold value

And corresponding number of samples

To all inclination angles, calculate maximum road number and time number of samples, then equivalent width and highly being respectively:

$\begin{array}{c}{w}_a=\underset{j}{\max }\left\{{\stackrel{\‾}{x}}_j\right\}\×\mathrm{dx}-\mathrm{dx}\\ h_a=\left[\underset{j}{\max \left\{{\stackrel{\‾}{t}}_j\right\}-1}\right]\×\mathrm{dt}\×V\×0.5\end{array}$

Wherein, V is the speed of seam unit, hole, and dx is a track pitch, and dt is the time sampling interval.

5. for certain type of seismic properties, optimum seismic properties threshold value can be obtained,, the correction coefficient of the seismic properties unusual apparent volume corresponding can be obtained with a certain model based on this seismic properties threshold value through above-mentioned statistical analysis technique; For a plurality of seismic properties, can obtain a plurality of optimum seismic properties threshold values, based on these seismic properties threshold values, just can obtain the correction coefficient of the unusual apparent volume of seismic properties under a series of corresponding seismic properties threshold values.How comprehensively these correction coefficient improve the precision that the seam hole quantizes, and are the main tasks that optimum apparent volume correction coefficient is confirmed.For this reason; We adopt three kinds of methods to come the compute optimal correction coefficient; Also promptly adopt normal correction coefficient (correction coefficient average), become correction coefficient the match relation of the corresponding unusual volume of seismic properties (correction coefficient with) and the definite correction coefficient of nonlinear mapping technique to come earthquake attribute abnormal volume is proofreaied and correct, utilize seismic properties to stitch the result of hole quantification thereby reach.

(1) normal correction coefficient is proofreaied and correct

Normal correction coefficient is meant that the unusual volume to certain type of all seismic properties adopts same correction coefficient to proofread and correct; This correction coefficient should be to utilize corresponding optimum seismic properties threshold value of each type seismic properties that foregoing statistical analysis technique obtains and the correction coefficient average corresponding with this seismic properties threshold value; Different seismic properties just has different correction coefficient averages.At first we proofread and correct the apparent volume Scij that obtains near true model to the unusual apparent volume of this seismic properties at the correction coefficient average of utilizing a certain seismic properties, and wherein i representes i theoretical model being designed, and j representes j seismic properties; The corresponding attribute abnormal volume of all seismic properties to optimizing carries out same processing, obtains the m corresponding with a certain theoretical model the apparent volume near true model, and wherein m is the attribute number; The apparent volume that each attribute calculated is carried out weighting summation obtain comprehensive apparent volume Scai, promptly near true model

${\mathrm{Sca}}_i=\underset{j=1}{\overset{m}{\mathrm{\Σ}}}{w}_j\×{\mathrm{Sc}}_{\mathrm{ij}}$

In the formula, w _jBe seismic properties weights coefficient, and have

The back chapters and sections will be introduced the calculating of these weights; Seismic properties to all models are corresponding is carried out above-mentioned same processing, obtains n the approaching comprehensive apparent volume of true model separately, and wherein n is the model number; Compare with the volume of separately true model, calculate the apparent volume relative error of all models, carry out the precision evaluation analysis, for reality seam hole quantizes to provide fiducial interval based on this error.

Adopt normal correction coefficient the corresponding seismic properties of actual earthquake 3-D data volume to be proofreaied and correct unusually the apparent volume that can obtain corresponding to the unusual volume of all seismic properties corresponding property that preferentially goes out after calibrated; The apparent volume that each attribute calculated is carried out the comprehensive apparent volume after weighting summation obtains proofreading and correct, and this volume is the result of seam hole quantificational description.

(2) becoming correction coefficient proofreaies and correct

The constant coefficient bearing calibration is simple, but since normal correction coefficient obtain through statistics, can only guarantee that most of seismic properties abnormal results is proofreaied and correct after, its volume approaches the volume of true model, after a part of earthquake attribute was calibrated, its error maybe be bigger.Therefore, adopt the variable coefficient bearing calibration that changes with earthquake attribute abnormal apparent volume.

The change correction coefficient is meant proofreaies and correct with the correction coefficient of this volume change the unusual volume employing of certain type of seismic properties; Utilize foregoing statistical analysis technique can obtain the unusual definite optimum seismic properties threshold value of corresponding each type seismic properties of a series of models; For certain model; Just can obtain the unusual correction coefficient of seismic properties under this seismic properties threshold value; Also promptly one wherein i representes i theoretical model being designed to correction coefficient Ccij that should the unusual volume of seismic properties, j representes j seismic properties; The corresponding seismic properties of all models (n) is carried out above-mentioned same processing, obtain the correction coefficient Ccij of n corresponding different earthquake attribute abnormal volume, it is right with corresponding correction coefficient data also promptly to obtain n the unusual volume of seismic properties; Adopt least-square fitting approach; Obtain the corresponding a certain function (like linear function) of these data of best-fit, Fig. 2 demonstrates with the resulting linear function expression formula of the unusual correction coefficient of linear function fit wave impedance seismic properties Y=5.88E-006*X+0.073.

Based on fitting function, can carry out the correction volume test to a series of seam hole theoretical models that design, calculate the apparent volume relative error of all models, be used for estimating seam hole quantified precision, for reality seam hole quantizes to provide fiducial interval; Simultaneously, can obtain each earthquake attribute abnormal volume weighting factor after optimum the correction.At first the unusual apparent volume of corresponding seismic properties to the theogram of some theoretical models calculates; Based on fitting function; Obtain the corresponding correction coefficient of this apparent volume; And then obtain apparent volume Scij near true model, and wherein i representes i theoretical model being designed, j representes j seismic properties; The corresponding attribute abnormal volume of all seismic properties to preferentially going out carries out same processing, obtains the m corresponding with a certain theoretical model the apparent volume near true model, and wherein m is the attribute number; The apparent volume that each attribute calculated is carried out weighting summation obtain comprehensive apparent volume Scai near true model; Seismic properties to all models are corresponding is carried out above-mentioned same processing, obtains n the approaching comprehensive apparent volume of true model separately, and wherein n is the model number; Compare with the volume of separately true model, calculate the apparent volume relative error of all models, carry out error analysis.

Adopt fitting function the corresponding seismic properties of actual earthquake 3-D data volume to be proofreaied and correct unusually the apparent volume that can obtain corresponding to the unusual volume of all seismic properties corresponding property that preferentially goes out after calibrated again; The apparent volume that each attribute calculated is carried out the comprehensive apparent volume after weighting summation obtains proofreading and correct, and this volume is the result of seam hole quantificational description.

(3) the non-linear of apparent volume correction coefficient confirmed

The seismic response of the reservoir model of being made up of seam hole or hole and actual bores meets the seam hole or the corresponding well side seismologic record of hole type reservoir can be known; Receive the factor affecting such as space spread characteristic of stuff character and reservoir in the degree of scatter, hole of height, width, the form of reservoir bodies, inner hole in the wave field characteristics of seam hole reservoir performance on the 3-D seismics record that the relatively more single Ordovician system inside story of lithology exists; Therefore; The unusual corresponding relation more complicated with seam hole volume of seismic properties; Be difficult to represent with a mathematical function; Therefore, we can utilize the volume of theoretical model that the unusual apparent volume of the seismic properties of its synthetic seismic data is carried out " training " based on artificial neural network technology; Obtain the Nonlinear Mapping relation of the unusual apparent volume of seismic properties and the volume of theoretical model, also being the unusual apparent volume of seismic properties concerns with the Nonlinear Mapping of corresponding correction coefficient; Utilize then and should concern; Seismic properties to actual earthquake 3-D data volume is corresponding is proofreaied and correct unusually; Reach and utilize seismic properties to stitch the purpose of hole quantificational description, wherein the seismic properties type is through the seismic properties that is to say opposite joint hole volume or seam hole spatial shape sensitivity after optimizing.

This project will adopt the nonlinear probability neural network structure from the unusual volume of the earthquake attribute volume that optimizes and the volume prediction earthquake attribute abnormal volume correction factor of theory of correspondences model.

Probabilistic neural network (The Probabilistic Neural Network writes a Chinese character in simplified form into PNN) is actually a kind of mathematics interpolation means, and it realizes having used just the structure of neural network.This is a kind of potential advantage, because through the research mathematical formulae, we can understand this method better than multilayer feedforward neural network (Multi-Layer Feedforward Neural Network writes a Chinese character in simplified form into MLFN).

The data of probabilistic neural network training usefulness are the same with the data of multilayer feedforward neural network training.It has comprised a series of training " sample set ", and the unusual apparent volume of seismic properties of the theogram of corresponding all models of each sample can be write as following vector form:

(A ₁₁，A ₂₁，…A _m1，Cc ₁)、(A ₁₂，A ₁₂，…A _m2，Cc ₂)、...、(A _1n，A _1n，…A _mn，Cc _n)

N sample and m seismic properties are arranged here.Cc _iValue can be through the correction coefficient of each training sample value with the apparent volume acquisition of corresponding theoretical model.

Given training data, the linear combination that each output calibration coefficient value of probabilistic neural network hypothesis can be write as the unusual apparent volume of seismic properties in the training data.Seismic properties collection for a new samples (sequence number is j) can be write as vector

$\stackrel{\→}{x}=(A_{1j},A_{2j},\·\·\·A_{\mathrm{mj}})$

Then new correction coefficient value can be estimated as:

$\begin{array}{c}\hat{C}c\left(x\right)=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{Cc}}_i\mathrm{Exp}(-D(x,x_i))}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\mathrm{Exp}(-D(x,x_i))}\\ D(x,x_i)=\underset{j=1}{\overset{m}{\mathrm{\Σ}}}{\left(\frac{x_j-x_{\mathrm{Ij}}}{{\mathrm{\σ}}_j}\right)}^2\end{array}$ Wherein:

Unknown quantity D (x, x _i) be input point and each training sampling point x _iBetween " distance ".This distance is to measure in the hyperspace that occupies of seismic properties, is through unknown quantity σ _jDemarcate, it maybe be different with each seismic properties value.

Can find out that from the formula of above-mentioned probabilistic neural network network training comprises confirms smoothing parameter σ _jOptimal.The rule of confirming these parameters is to make it to have minimum verify error.

The assay that defines j target sampling point is following:

$\hat{C}{c}_j\left(x_j\right)=\frac{\underset{i\≠j}{\overset{n}{\mathrm{\Σ}}}{\mathrm{Cc}}_i\exp (-D(x_j,x_i))}{\underset{i\≠j}{\overset{n}{\mathrm{\Σ}}}\exp (-D(x_j,x_i))}$

When sampling point was not within training data, it was exactly the predicted value of j target sampling point.Because the value of our known sampling point, just can calculate the predicated error of sampling point.To each this process of this repetition of training appearance collection, we just can define total predicated error of training data:

$E_V({\mathrm{\σ}}_1,{\mathrm{\σ}}_2,{\mathrm{\σ}}_3)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{Cc}}_i-\hat{C}{c}_i)}^2$

Notice that predicated error depends on parameter σ _jSelection.This unknown quantity is to come minimized through using non-linear conjugated gradient algorithm.Train later network and just had the minimum characteristic of verify error.

The correction coefficient template curve synoptic diagram that Fig. 3 obtains for probabilistic neural network.The figure shows the X plot of a correction coefficient value merchandiser seismic properties.Conventional linear regression method makes mean square prediction error calculate minimum calculating with fitting a straight line, also can come match with luminance curve, and be to use neural network to obtain the relation between the variable here.Can find out that from this figure probabilistic neural network has the good characteristics of fitting data, this is the same with multilayer feedforward neural network.But it has more stability in the seismic properties scope.The maximum problem of probabilistic neural network is that each output sample compares with each training sample, so computing time is slow because it is carried out in around all training datas.

6. the unusual volume bearing calibration of seismic properties realizes

At the three dimensions of trying to achieve with the regular shape of the unusual volume of irregular seismic properties equivalence; After also promptly trying to achieve the length with the equivalence of the space spread of erose earthquake attribute volume; We just can proofread and correct according to different scale factors respectively with width the height of earthquake attribute abnormal respectively, thereby have realized a discontinuous three-dimensional seismic properties anomalous body in the irregular border of form is carried out volume correction.

Suppose that Wc is the cross-level coefficient, Hc is the correction coefficient on the time orientation, also is vertical correction coefficient.Wa is the equivalent width along the attribute abnormal on the seismic properties section of CDP direction, and Ha is the height of the attribute abnormal on vertically, also is the spatial dimension of beading seismic reflection attribute volume.

After correction coefficient longitudinally and laterally obtains; We consider that spatial shape to earthquake attribute abnormal body is according to carrying out autoscan and compression to correction coefficient in length and breadth; Make seismic properties really can reflect position, size and the form of stitching the hole unusually, to realize reflexing to the quantification of seam unit, hole by beading.

At last, invention usage range of the present invention and application prospect are described:

This method has practicality widely for nonuniformity carbonatite seam hole type oil reservoir; Stitch the calculation of reserves of hole type oil reservoir at present for carbonatite; Go back neither one method very accurately and effectively both at home and abroad; At present still adopt the volumetric method of effective reservoir flattening-out to calculate basically; This itself does not meet the characteristics of nonuniformity carbonatite seam hole type oil reservoir, and carbonatite seam hole type reservoir model is just being drilled and the volume correction technology has then been considered carbonatite seam hole type oil reservoir nonuniformity, and therefore being with a wide range of applications also can be further perfect.

Claims (7)

1. a carbonatite seam hole type reservoir volume Forecasting Methodology is characterized in that, may further comprise the steps:

1.1) obtain the earthquake attribute volume volume that the hole body is stitched in test through seismic event;

1.2) multiply by correction coefficient according to said earthquake attribute volume volume and predict said test seam hole body volume.

2. according to the said volume Forecasting Methodology of claim 1, it is characterized in that said correction coefficient equals the width correction coefficient and the height correction coefficient is long-pending.

3. according to the said volume Forecasting Methodology of claim 2, it is characterized in that said width correction coefficient and height correction coefficient are the constant normal correction coefficient of relative one type of seismic properties.

4. according to the said volume Forecasting Methodology of claim 3, it is characterized in that said normal correction coefficient is obtained through statistical study, specifically comprises:

2.1) just drilling to blow out through the numerical simulation of a large amount of seam hole modellings and wave equation and align the data of drilling and carry out overlap-add procedure, STACK DATA is carried out seismic properties extract;

2.2) through the volume of a large amount of seams hole phantom type and the statistics of earthquake attribute volume volume, obtain the correction coefficient between earthquake attribute volume volume and the seam hole body volume.

5. according to the said volume Forecasting Methodology of claim 2, it is characterized in that said width correction coefficient and height correction coefficient are the change correction coefficient that changes with earthquake attribute abnormal apparent volume.

6. according to the said volume Forecasting Methodology of claim 5, it is characterized in that said change correction coefficient is obtained with linear function fit on the statistical study basis.

7. according to the said volume Forecasting Methodology of claim 5, it is characterized in that said change correction coefficient is obtained with the probabilistic neural network match on the statistical study basis.

Images