CN108562950B - Method for intelligently dividing stratum horizon based on logging information - Google Patents
Method for intelligently dividing stratum horizon based on logging information Download PDFInfo
- Publication number
- CN108562950B CN108562950B CN201711310475.0A CN201711310475A CN108562950B CN 108562950 B CN108562950 B CN 108562950B CN 201711310475 A CN201711310475 A CN 201711310475A CN 108562950 B CN108562950 B CN 108562950B
- Authority
- CN
- China
- Prior art keywords
- curve
- depth
- points
- layering
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000005259 measurement Methods 0.000 claims abstract description 15
- 238000009499 grossing Methods 0.000 claims abstract description 11
- 238000013517 stratification Methods 0.000 claims description 68
- 239000013598 vector Substances 0.000 claims description 45
- 238000004364 calculation method Methods 0.000 claims description 28
- 239000011159 matrix material Substances 0.000 claims description 19
- 238000012360 testing method Methods 0.000 claims description 14
- 238000001514 detection method Methods 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 7
- FGUUSXIOTUKUDN-IBGZPJMESA-N C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 Chemical compound C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 FGUUSXIOTUKUDN-IBGZPJMESA-N 0.000 claims description 6
- GNFTZDOKVXKIBK-UHFFFAOYSA-N 3-(2-methoxyethoxy)benzohydrazide Chemical compound COCCOC1=CC=CC(C(=O)NN)=C1 GNFTZDOKVXKIBK-UHFFFAOYSA-N 0.000 claims description 4
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 230000032798 delamination Effects 0.000 claims description 2
- 230000015572 biosynthetic process Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 6
- 238000006073 displacement reaction Methods 0.000 description 6
- 230000001186 cumulative effect Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000009191 jumping Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V9/00—Prospecting or detecting by methods not provided for in groups G01V1/00 - G01V8/00
Landscapes
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Investigation Of Foundation Soil And Reinforcement Of Foundation Soil By Compacting Or Drainage (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a method for intelligently dividing stratum positions based on logging information, which comprises the following steps: a. finding the wave crest and the wave trough of the curve; b. calculating the relative depth difference and the measurement coordinate difference between the two wells; c. smoothing the input curve; d. detecting the similarity of the single curves; e. detecting the similarity of multiple curves; f. establishing a stratigraphic layering model based on a single curve; g. and establishing a multi-curve-based stratigraphic layering model. The method firstly provides the method for dividing the stratum by using the mathematical algorithm according to the logging information so as to overcome the two problems that the human influence factor of the horizon division is too large and the layering experience is difficult to transmit, and improve the speed and the precision of the stratum division.
Description
Technical Field
The invention relates to the technical field of logging interpretation and evaluation, in particular to a method for intelligently dividing stratum horizons based on logging information, which is used for dividing geological horizons for logging interpretation and evaluation.
Background
In the well logging interpretation and evaluation process chain, geological horizon division is a more critical step. At present, stratum horizons are divided, a common method is to compare and research a logging curve of a target well which is not subjected to horizon division with a logging curve of a standard well, and the horizons of the target well are manually divided under the condition that the horizon division of the standard well is known, which often needs the experience of professionals.
However, the existing method has the following defects: the manual horizon division has the influence of human factors, and the experience accumulation needs longer working time. Meanwhile, the horizon division experience is difficult to transfer on one hand; on the other hand, as exploration and development are deep and geological knowledge changes, horizon division can be changed according to new knowledge.
From the viewpoint of digital signal processing analysis technology, any one log is a varying waveform. The fundamental reason for comparing waveforms of two adjacent wells in the same layer is that the waveforms are similar in change form, but because of different thicknesses, the lengths of the waveforms are different, and the mathematical problem can be expressed by solving the shortest distance between two groups of unequal length logging information. The nature of the mathematical problem is the same as that of the modern big data analysis voice recognition technology, different people say the same sentence with different speeds, and the curve length of the sound wave received by the instrument is different, but the form is similar. It can be seen that the horizon automatic partitioning and speech recognition have good analogy on mathematical models, but the horizon partitioning has two differences: firstly, the horizon division needs to utilize a multi-dimensional changing wave curve; the second is that the horizon division is a dynamic depth window.
Disclosure of Invention
The invention aims to provide a method for intelligently dividing stratum positions based on logging information aiming at the defects and shortcomings of the prior art.
The invention is realized by adopting the following technical scheme:
a method for intelligently dividing stratum horizon based on logging information is characterized by comprising the following steps:
a. finding the wave crest and the wave trough of the curve;
b. calculating the relative depth difference and the measurement coordinate difference between the two wells;
c. smoothing the input curve;
d. detecting the similarity of the single curves;
e. detecting the similarity of multiple curves;
f. establishing a stratigraphic layering model based on a single curve;
g. and establishing a multi-curve-based stratigraphic layering model.
In the step a, the calculation process of finding the wave crest and the wave trough of the curve is as follows:
given that the curve X ═ { X1, X2, X3, … xn }, X1 ═ { X2, X3, X4, … xn, xn +1}, let X2 ═ X1-X, the difference between neighboring points is obtained, and the points at the peaks meet one of the following two characteristics:
(1) x2 is negative at position j and positive at j-1, then the j position is the peak;
(2) and if the position j of the X2 is negative and the position j-1 is 0, continuing to search from the position j-1 until the position X2(j-m) is not equal to 0, and if the position X2(j-m) is positive, the positions m +1 from the position j-m to the position j have the same value and are all located in a flat peak, and taking the midpoint as a peak point, namely (j + j-m)/2 and then rounding.
The trough calculation is the same as the peak calculation, and only the sign of the input curve needs to be reversed.
In the step B, in calculating the relative depth difference and the measurement coordinate difference between the two wells, because the two wells AB have a macroscopic similarity relationship, but the two wells start to measure different depths, the depth difference (meter, the relative displacement distance of B relative to a in the earth center direction) and the measurement coordinate difference (no unit number, the relative displacement coordinate of B relative to a in the earth center direction) between the two logging curves need to be calculated.
In step c, the input curve is smoothed by the following method:
and (3) setting an input curve A, a smooth width w and a smooth back curve B, wherein the input curve A is not smooth for the first 1:1-w of the input curve A, and is not smooth for the last N-w: N of the input curve A, and only the smooth w: N-w part is as follows:
B(1:1-w)＝A(1:1-w)
B(N-w:N)＝A(N-w:N)
for any i range between w: N-w, there is
B (i) ═ mean (a (i-w: i + w)), mean is an average value.
In the above, the curve is once smoothed, and if the curve needs to be smoothed for many times, the curve after once smoothing is smoothed again, and so on.
In the step d, the single curve similarity detection method comprises the following steps:
assuming a standard reference template R, which is an M-dimensional vector, i.e., R ═ { R (1), R (2), … …, R (M), … …, R (M) }, each component may be a number or a smaller vector; there is a template T to be tested which is an N-dimensional vector, i.e. T ═ { T (1), T (2), … …, T (N), … …, T (N) } again each component may be a number or a smaller vector, note that M is not necessarily equal to N, but the dimensions of each component should be the same;
since M is not necessarily equal to N, to calculate the similarity between R and T, the distance between each component of R and each component of T is first calculated to form a matrix of M × N. (for convenience, the number of rows is defined as the dimension M of the standard template and the number of columns is defined as the dimension N of the template to be tested).
The specific calculation method is as follows:
assuming that the reference template R is 6 letters ABCDEF in total, the test template T is 4 letters 1234, and the distance between each element in R and T is given in advance; as shown in fig. 3.
d (i, j) represents the values of i rows and j columns with the bottom left point as the initial coordinate origin, namely the Euclidean distance between the ith point of the original sequence R and the jth point (which is considered as a one-dimensional point at first) of the test template T;
d(i,j)＝(Ri-Tj)2
since T is template matching, the sequence matching order of each component has been determined, although not in a one-to-one correspondence; because the lengths of the 2 templates are different, the corresponding matching relationships thereof are various, and the matching path with the shortest distance needs to be found, it is assumed that the topics satisfy the following constraints:
a square (i, j), which may arrive from three directions, (i-1, j-1), (i-1, j) and (i, j-1), respectively, the distance values g (i, j) at (i, j) arriving from different directions are different, respectively defined as follows:
if the signal arrives from (i-1, j-1), g (i, j) ═ g (i-1, j-1) +2d (i, j)
If it arrives from (i-1, j), g (i, j) ═ g (i-1, j) + d (i, j)
If the result is from (i, j-1), g (i, j) ═ g (i, j-1) +2d (i, j)
In the three directions, only the smallest one can be selected, and the strategy is as follows:
where g (i, j) indicates that 2 templates are all matched one after another starting from the starting component, i component in M and j component in T have been reached, and the matching to this step is the local cumulative minimum distance between the 2 templates, and is all d (i, j) or 2d (i, j) is added to the result of the previous matching, and then the minimum value is taken.
In step e, the multi-curve similarity detection method comprises the following steps:
(1) definition of multidimensional points:
ri is a component of the multidimensional point (a1, a2, …, an), and n components are provided;
tj is a component (b1, b2, …, bn) of the multidimensional point, and n components are provided;
wherein ai, bi (i ═ 1., n) belong to the same parameter; such as all sonic profiles;
(2) normalization of multidimensional points:
because the distances between different parameters are different due to different units of each curve, in order to reduce the influence caused by different units under the condition of unifying a multi-dimensional point, each parameter needs to be normalized, and thus all parameters of the multi-dimensional point are in the range of 0-1;
for each dimension of the parametric sequence, e.g., c ═ ci }, (e.g., a1 above, b1 is one of ci), the following normalization is performed:
3) calculation of distances between multidimensional points
d (i, j) represents the i row and j column values starting from the bottom left point, i.e. the Euclidean distance between the ith point of the original sequence R and the jth point (possibly a multidimensional point) of the test template T, having
The rest methods are completely the same as the single curve similarity detection algorithm.
In the step f, a stratigraphic layering model based on a single curve is established:
known (1) A log: a ═ ai ═ a1, a2, … a8, where ai is the column vector; define ci as above, column vectors with a1 depth, etc.
Known (2) B log: b ═ bi, where bi are all column vectors; define ci as above, column vectors as b1 depth, and so on.
As is known, (3) geological and artificial layering of A LA＝{LA1,LA2,LA3,LA4,LA5In which L isAiAre column vectors, representing respectively: geological stratification name, geological stratification depth, geological stratification thickness, artificial stratification depth and artificial stratification thickness;
geological stratification of B, L, is known (4)B＝{LB1,LB2,B3In which L isBiAre column vectors, representing respectively: geological stratification name, geological stratification depth and geological stratification thickness;
for any j layers of A, by searching LARow j, the hierarchical data of the layer is obtained:
LA(j, 1) geological stratification name, LA(j，2) Depth of geological stratification, LA(j, 3) geological stratification thickness, LA(j, 4) depth of Artificial layering, LA(j, 5) artificial delamination thickness;
according to LA(j, 4) artificial stratification depth, and a1(namely the depth sequence of the well A, the N multiplied by 1 matrix, the line number represents the serial number, the data represents the depth in the unit of meter, N is the total number of logging data points), and the serial number k of the measurement line of the jth layer in the well A is found;
taking k-w as step, k + w curve discrete point on A as a matching template pattern A;
find L and B onA(j, 1) the geological stratification number m (i.e. the mth layer on the B or the mth row of the B matrix) of the same geological stratification name, and further finding the geological stratification result of the B: depth LS(m, 2) and thickness LS(m，2)
According to LS(m, 2) geological stratification depth, taking depth L on BS(m，2)-wB1To LS(m，2)+wB2The section of the logging curve x is used as a range for finding the best matching;
the lower limit of the curve for x should not be lower than the depth of the m-1 th layer, i.e.
LS(m，2)-wB1>LS(m-1, 2); also, the upper limit of the curve for x should not be greater than the depth of the m +1 th layer, i.e., LS(m，2)+wB2<LS(m+1，2)；wB1And wB2Taking 50 meters generally;
after the depth range of the B well search area x is obtained, the depth range can be further searched according to B1(i.e., depth series for B-well) the number range for X is obtained.
From the curve of x, traversing from the beginning to the end of length (x) points, taking p as the center and the depth of each offset wa at the ith point p as the xi position (manually defining an initialized depth parameter when the algorithm is realized, generally taking 4 meters), jumping and taking points by step to obtain a matching template pattenB on B, then calculating the similar distance between pattenA and pattenB, and obtaining the result that
di(i＝1，2，3，...，length(x))；
Theoretically, all d are locatediIn dkIs equal to min ({ d)iH), then, the depth value corresponding to the kth sequence number of the sub-curve x is the obtained layering point on B, and the depth value corresponds to the jth layering point on a one-to-one;
in practice, however, the theoretical minimum distance index k point may not be a true best match due to the presence of noise.
Setting y points of the minimum distance curve, and taking the minimum distance points which are per% before the points to form a new sequence z (z is a matrix of two columns, the first column is a serial number, and the second column is a distance value); after z is sequenced according to sequence numbers, the continuous condition of the first column (namely the sequence number column) of z on a mathematical integer is observed, a continuous sequence number with multiple sequence numbers has higher possibility of containing real layering points, if the taken per% is larger, the number of points is more, the number of groups of points is more possible, if K reference layering points are given, the average point sequence number of the groups with the continuous number of points arranged at the first K is taken, after the sequence number is obtained, for example, the sequence number is n, the automatic layering depth of the computer of the well B is B1(n), namely the number corresponding to the nth row of the well B depth matrix, and the unit is meter; by this, automatic layering is completed.
In the step g, a multi-curve-based stratigraphic layering model is established: consistent with the establishment of a single-curve stratum layered model, the difference is that a matching curve (a known layered curve A) and a matched curve (unknown and requiring a layered curve B) are not single curves any more but curve groups, and therefore, multi-dimensional calculation is adopted for the similarity distance calculation of pattenA and pattenA.
Compared with the prior art, the invention has the following beneficial effects:
firstly, the invention at least achieves two purposes, and firstly, the invention overcomes the problems that the human influence factor of horizon division is too large and the layering experience is difficult to transfer; and secondly, the speed and the precision of stratigraphic division are improved.
Secondly, aiming at the problems of multi-dimensional logging information and a dynamic depth window, the invention develops a set of brand-new shortest distance solving optimization algorithm of two groups of unequal length logging information, and the algorithm is from forward modeling to inversion, from high signal to low signal, from local optimization to global optimization. The whole method and the whole flow are easy to realize in computer languages.
By utilizing the method and the device, the speed and the precision of stratum division in the interpretation and evaluation of the petroleum logging information are improved, so that stratum position information is obtained more accurately, and compared with the traditional manual method, the method and the device have wider universal adaptability.
Drawings
The invention will be described in further detail with reference to the following description taken in conjunction with the accompanying drawings and detailed description, in which:
FIG. 1 is a time series plot prototype reference vector theoretical vector feature vector versus noise plot;
FIG. 2 is a diagram of a shortest distance optimizing path between a feature vector and a reference vector;
FIG. 3 is a diagram of the time-optimal correspondence of feature vectors to reference vectors;
FIG. 4 is a diagram of the results of automatic identification of stratigraphic horizons;
FIG. 5 is a graph of the distance between elements of the test template T and the standard template R;
FIG. 6 is a diagram showing the steps of matching the test template T with the standard template R;
FIG. 7 is a diagram of the transverse path of the matching of the test template T and the standard template R;
FIG. 8 is a diagram of the vertical up-path of the match between the test template T and the standard template R;
FIG. 9 is a graph of similar shortest distance between patternA and patternB;
fig. 10 is a similar shortest distance curve multiple valley recognition graph.
Detailed Description
Example 1
The invention firstly provides a concept of intelligently dividing the stratum horizon based on the logging information, finds a corresponding calculation method and achieves the purpose of improving the dividing precision of the stratum horizon.
a. Finding the wave crest and the wave trough of the curve;
b. calculating the relative depth difference and the measurement coordinate difference between the two wells;
c. smoothing the input curve;
d. detecting the similarity of the single curves;
e. detecting the similarity of multiple curves;
f. establishing a stratigraphic layering model based on a single curve;
g. and establishing a multi-curve-based stratigraphic layering model.
In the step a, the calculation process of finding the wave crest and the wave trough of the curve is as follows:
given that the curve X ═ { X1, X2, X3, … xn }, X1 ═ { X2, X3, X4, … xn, xn +1}, let X2 ═ X1-X, the difference between neighboring points is obtained, and the points at the peaks meet one of the following two characteristics:
(1) x2 is negative at position j and positive at j-1, then the j position is the peak;
(2) and if the position j of the X2 is negative and the position j-1 is 0, continuing to search from the position j-1 until the position X2(j-m) is not equal to 0, and if the position X2(j-m) is positive, the positions m +1 from the position j-m to the position j have the same value and are all located in a flat peak, and taking the midpoint as a peak point, namely (j + j-m)/2 and then rounding.
The trough calculation is the same as the peak calculation, and only the sign of the input curve needs to be reversed.
In the step B, in calculating the relative depth difference and the measurement coordinate difference between the two wells, because the two wells AB have a macroscopic similarity relationship, but the two wells start to measure different depths, the depth difference (meter, the relative displacement distance of B relative to a in the earth center direction) and the measurement coordinate difference (no unit number, the relative displacement coordinate of B relative to a in the earth center direction) between the two logging curves need to be calculated.
In step c, the input curve is smoothed by the following method:
let the input curve A, smooth width w, and curve B after smoothing. The first 1:1-w of A is not smooth, the last N-w: N of A is not smooth, only the w: N-w portion is smooth, then:
B(1:1-w)＝A(1:1-w)
B(N-w:N)＝A(N-w:N)
for any i range between w: N-w, there is
B (i) ═ mean (a (i-w: i + w)), mean is an average value.
In the above, the curve is once smoothed, and if the curve needs to be smoothed for many times, the curve after once smoothing is smoothed again, and so on.
In the step d, the single curve similarity detection method comprises the following steps:
assuming a standard reference template R, which is an M-dimensional vector, i.e., R ═ { R (1), R (2), … …, R (M), … …, R (M) }, each component may be a number or a smaller vector. There is a template T to be tested which is an N-dimensional vector, i.e. T ═ { T (1), T (2), … …, T (N), … …, T (N) } again each component may be a number or a smaller vector, note that M is not necessarily equal to N, but the dimensions of each component should be the same.
Since M is not necessarily equal to N, to calculate the similarity between R and T, the distance between each component of R and each component of T is first calculated to form a matrix of M × N. (for convenience, the number of rows is defined as the dimension M of the standard template and the number of columns is defined as the dimension N of the template to be tested).
The specific calculation method is as follows:
assume the standard template R is the letter ABCDEF (6) and the test template T is 1234 (4). The distances between the elements in R and T have been given as shown in fig. 3.
d (i, j) represents the i row and j column values starting from the bottom left point, i.e. the euclidean distance between the ith point of the original sequence R and the jth point of the test template T (here first considered to be a one-dimensional point).
d(i,j)＝(Ri-Tj)2
Since T is a template match, the order of the matches of the components has been determined, although not in a one-to-one correspondence. Because the lengths of the 2 templates are different, the corresponding matching relations are various, and the matching path with the shortest distance needs to be found out. Now assume that the title satisfies the following constraint:
a square (i, j), which may arrive from three directions, (i-1, j-1), (i-1, j) and (i, j-1), respectively, the distance values g (i, j) at (i, j) arriving from different directions are different, respectively defined as follows:
if the signal arrives from (i-1, j-1), g (i, j) ═ g (i-1, j-1) +2d (i, j)
If it arrives from (i-1, j), g (i, j) ═ g (i-1, j) + d (i, j)
If the result is from (i, j-1), g (i, j) ═ g (i-j, 1) +2d (i, j)
In the three directions, only the smallest one can be selected, and the strategy is as follows:
where g (i, j) indicates that 2 templates are all matched one after another starting from the starting component, i component in M and j component in T have been reached, and that the match to this step is the local cumulative minimum distance between the 2 templates. And d (i, j) or 2d (i, j) is added to the result of the previous matching, and then the minimum value is taken.
For example, fig. 4 is labeled after all the matching steps in fig. 3.
The calculation method is that if g (0,0) is assumed to be 0, then g (1,1) is g (0,0) +2d (1,1) is 0+2 × 2 is 4 (upper right italics 4 in the leftmost grid in the upper figure). The same holds true for g (2,2) ═ 9, and if calculated from g (1,2), g (2,2) ═ g (1,2) + d (2,2) ═ 5+4 ═ 9, because it is going vertically upward. If g (2,1) is counted, g (2,2) + d (2,2) ═ 7+4 ═ 11, because the transverse direction is towards the right. If calculated from g (1,1), g (2,2) ═ g (1,1) +2 × d (2,2) ═ 4+2 × 4 ═ 12. Because it is oblique to the past. In summary, the minimum value is 9, and all g (2,2) ═ 9. Others may be analogized.
The first row of results is calculated as in fig. 5, where each arrow indicates the direction from which the minimum originated. The results after the second row are calculated as in fig. 6. The answer is obtained after all the calculation, namely the distance between the 2 templates is 26 (the top right corner of the figure 6).
In step e, the multi-curve similarity detection method comprises the following steps:
(1) definition of multidimensional points:
ri is a multi-dimensional point with components of (a1, a2, …, an) and n components
Tj is a component (b1, b2, …, bn) of the multidimensional point, and n components are provided
Where ai, bi (i ═ 1., n) belong to the same parameter, e.g., both are sonic curves.
(2) Normalization of multidimensional points:
because the distances between different parameters are different due to different units of each curve, in order to reduce the influence caused by different units under the condition of unifying a multi-dimensional point, each parameter needs to be normalized, and thus all parameters of the multi-dimensional point are in the range of 0-1.
For each dimension of the parametric sequence, e.g., c ═ ci }, (e.g., a1 above, b1 is one of ci), the following normalization is performed:
3) calculation of distances between multidimensional points
d (i, j) represents the i row and j column values starting from the bottom left point, i.e. the Euclidean distance between the ith point of the original sequence R and the jth point (possibly a multidimensional point) of the test template T, having
The rest methods are completely the same as the single curve similarity detection algorithm.
In the step f, a stratigraphic layering model based on a single curve is established:
known (1) A log: a is { ai } { a1, a2, … a8}, where ai are column vectors, column vectors defining ci above, e.g., a1 depth, and so on.
Known (2) B log: b-bi, where bi are column vectors, column vectors defining the same ci, e.g., B1 depth, etc.
As is known, (3) geological and artificial layering of A LA＝{LA1,LA2,LA3,LA4,LA5In which L isAiAre column vectors, representing respectively: geologic stratification nameWeighing geological stratification depth, geological stratification thickness, artificial stratification depth and artificial stratification thickness.
Geological stratification of B, L, is known (4)B＝{LB1,LB2,B3In which L isBiAre column vectors, representing respectively: geological stratification name, geological stratification depth and geological stratification thickness.
For any j layers of A, by searching LARow j, the hierarchical data of the layer is obtained:
LA(j, 1) geological stratification name, LA(j, 2) geological stratification depth, LA(j, 3) geological stratification thickness, LA(j, 4) depth of Artificial layering, LA(j, 5) artificially layering the thickness.
According to LA(j, 4) artificial stratification depth, and a1(i.e., depth series for well A, N1 matrix, row number for serial number, data for depth in meters, N is total logging data point number) and find the j-th layer measurement row number k in A.
Taking discrete points of a k-w step k + w curve on A as a matching template patternA.
Find L and B onA(j, 1) the geological stratification number m (i.e. the mth layer on the B or the mth row of the B matrix) of the same geological stratification name, and further finding the geological stratification result of the B: depth LS(m, 2) and thickness LS(m，3)
According to LS(m, 2) geological stratification depth, taking depth L on BS(m，2)-wB1To LS(m，2)+wB2As a range for finding the best match.
The lower limit of the curve for x should not be lower than the depth of the m-1 th layer, i.e., LS(m，2)-wB1>LS(m-1, 2); also, the upper limit of the curve for x should not be greater than the depth of the m +1 th layer, i.e., LS(m，2)+wB2<LS(m+1，2)。wB1And wB2Typically 50 meters.
After the depth range of the B well search area x is obtained, the depth range can be further searched according to B1(i.e., B well)Depth series of X) to obtain the range of X numbers.
From the curve of x, traversing from beginning to end length (x) points, taking p as the center and the depth of each offset wa at the ith point p as the xi position (manually defining an initialized depth parameter when the algorithm is realized, generally taking 4 meters), jumping and taking points by step to obtain a matching template pattenB on B, then calculating the similar distance between pattenA and pattenB, and obtaining the result di(i＝1，2，3，...，length(x))。
Theoretically, all d are locatediIn dkIs equal to min ({ d)iH), then the depth value corresponding to the kth sequence number of the sub-curve x is the obtained layering point on B, which corresponds to the depth value corresponding to the jth layering point on a one-to-one.
In practice, however, the theoretical minimum distance index k point may not be a true best match due to the presence of noise.
Setting y points of the minimum distance curve, and taking the minimum distance points which are per% before the points to form a new sequence z (z is a matrix of two columns, the first column is a serial number, and the second column is a distance value); after z is sorted according to sequence numbers, the probability that the sequence of a plurality of sequence numbers contains a real layering point is higher when the sequence of the first column (namely, the sequence number column) of z is continuous on a mathematical integer. If the taken per% is larger, the number of points is more, the number of groups of points is possibly more, if K reference layering points are required, the group average point serial numbers of the first K continuous point numbers are taken, after the serial numbers are obtained, for example, the serial numbers are n, the automatic layering depth of the computer of the well B is B1(n), namely the number corresponding to the nth row of the well B depth matrix, and the unit is meter. By this, automatic layering is completed.
In the step g, a multi-curve-based stratigraphic layering model is established:
almost completely consistent with step f, the difference is that the matching curve (known layered curve A) and the matched curve (unknown, requiring layered curve B) are no longer a single curve, but are curve groups, and therefore, the calculation of the similarity distance between pattenA and pattenA adopts multidimensional calculation.
Example 2
a. Finding the wave crest and the wave trough of the curve;
b. calculating the relative depth difference and the measurement coordinate difference between the two wells;
c. smoothing the input curve;
d. detecting the similarity of the single curves;
e. detecting the similarity of multiple curves;
f. establishing a stratigraphic layering model based on a single curve;
g. and establishing a multi-curve-based stratigraphic layering model.
In the step a, the calculation process of finding the wave crest and the wave trough of the curve is as follows:
given that the curve X ═ { X1, X2, X3, … xn }, X1 ═ { X2, X3, X4, … xn, xn +1}, let X2 ═ X1-X, the difference between neighboring points is obtained, and the points at the peaks meet one of the following two characteristics:
(1) x2 is negative at position j and positive at j-1, then the j position is the peak;
(2) and if the position j of the X2 is negative and the position j-1 is 0, continuing to search from the position j-1 until the position X2(j-m) is not equal to 0, and if the position X2(j-m) is positive, the positions m +1 from the position j-m to the position j have the same value and are all located in a flat peak, and taking the midpoint as a peak point, namely (j + j-m)/2 and then rounding.
The trough calculation is the same as the peak calculation, and only the sign of the input curve needs to be reversed.
In the step B, in calculating the relative depth difference and the measurement coordinate difference between the two wells, because the two wells AB have a macroscopic similarity relationship, but the two wells start to measure different depths, the depth difference (meter, the relative displacement distance of B relative to a in the earth center direction) and the measurement coordinate difference (no unit number, the relative displacement coordinate of B relative to a in the earth center direction) between the two logging curves need to be calculated.
In step c, the input curve is smoothed by the following method:
let the input curve A, smooth width w, and curve B after smoothing. The first 1:1-w of A is not smooth, the last N-w: N of A is not smooth, only the w: N-w portion is smooth, then:
B(1:1-w)＝A(1:1-w)
B(N-w:N)＝A(N-w:N)
for any i range between w: N-w, there is
B (i) ═ mean (a (i-w: i + w)), mean is an average value.
In the above, the curve is once smoothed, and if the curve needs to be smoothed for many times, the curve after once smoothing is smoothed again, and so on.
In the step d, the single curve similarity detection method comprises the following steps:
assuming a standard reference template R, which is an M-dimensional vector, i.e., R ═ { R (1), R (2), … …, R (M), … …, R (M) }, each component may be a number or a smaller vector. There is a template T to be tested which is an N-dimensional vector, i.e. T ═ { T (1), T (2), … …, T (N), … …, T (N) } again each component may be a number or a smaller vector, note that M is not necessarily equal to N, but the dimensions of each component should be the same.
Since M is not necessarily equal to N, to calculate the similarity between R and T, the distance between each component of R and each component of T is first calculated to form a matrix of M × N. (for convenience, the number of rows is defined as the dimension M of the standard template and the number of columns is defined as the dimension N of the template to be tested).
The specific calculation method is as follows:
assume the standard template R is the letter ABCDEF (6) and the test template T is 1234 (4). The distances between the elements in R and T have been given as shown in fig. 3.
d (i, j) represents the i row and j column values starting from the bottom left point, i.e. the euclidean distance between the ith point of the original sequence R and the jth point of the test template T (here first considered to be a one-dimensional point).
d(i,j)＝(Ri-Tj)2
Since T is a template match, the order of the matches of the components has been determined, although not in a one-to-one correspondence. Because the lengths of the 2 templates are different, the corresponding matching relations are various, and the matching path with the shortest distance needs to be found out. Now assume that the title satisfies the following constraint:
a square (i, j), which may arrive from three directions, (i-1, j-1), (i-1, j) and (i, j-1), respectively, the distance values g (i, j) at (i, j) arriving from different directions are different, respectively defined as follows:
if the signal arrives from (i-1, j-1), g (i, j) ═ g (i-1, j-1) +2d (i, j)
If it arrives from (i-1, j), g (i, j) ═ g (i-1, j) + d (i, j)
If the result is from (i, j-1), g (i, j) ═ g (i, j-1) +2d (i, j)
In the three directions, only the smallest one can be selected, and the strategy is as follows:
where g (i, j) indicates that 2 templates are all matched one after another starting from the starting component, i component in M and j component in T have been reached, and that the match to this step is the local cumulative minimum distance between the 2 templates. And d (i, j) or 2d (i, j) is added to the result of the previous matching, and then the minimum value is taken.
In step e, the multi-curve similarity detection method comprises the following steps:
(1) definition of multidimensional points:
ri is a multi-dimensional point with components of (a1, a2, …, an) and n components
Tj is a component (b1, b2, …, bn) of the multidimensional point, and n components are provided
Where ai, bi (i ═ 1., n) belong to the same parameter, e.g., both are sonic curves.
(2) Normalization of multidimensional points:
because the distances between different parameters are different due to different units of each curve, in order to reduce the influence caused by different units under the condition of unifying a multi-dimensional point, each parameter needs to be normalized, and thus all parameters of the multi-dimensional point are in the range of 0-1.
For each dimension of the parametric sequence, e.g., c ═ ci }, (e.g., a1 above, b1 is one of ci), the following normalization is performed:
3) calculation of distances between multidimensional points
d (i, j) represents the i row and j column values starting from the bottom left point, i.e. the Euclidean distance between the ith point of the original sequence R and the jth point (possibly a multidimensional point) of the test template T, having
The rest methods are completely the same as the single curve similarity detection algorithm.
In the step f, a stratigraphic layering model based on a single curve is established:
known (1) A log: a is { ai } { a1, a2, … a8}, where ai are column vectors, column vectors defining ci above, e.g., a1 depth, and so on.
Known (2) B log: b-bi, where bi are column vectors, column vectors defining the same ci, e.g., B1 depth, etc.
As is known, (3) geological and artificial layering of A LA＝{LA1,LA2,LA3,LA4,LA5In which L isAiAre column vectors, representing respectively: geological stratification name, geological stratification depth, geological stratification thickness, artificial stratification depth and artificial stratification thickness.
Geological stratification of B, L, is known (4)B＝{LB1,LB2,B3In which L isBiAre column vectors, representing respectively: geological stratification name, geological stratification depth and geological stratification thickness.
For any j layers of A, by searching LARow j, the hierarchical data of the layer is obtained:
LA(j, 1) geological stratification name, LA(j, 2) geological stratification depth, LA(j, 3) geological stratification thickness, LA(j, 4) Artificial layeringDepth, LA(j, 5) artificially layering the thickness.
According to LA(j, 4) artificial stratification depth, and a1(i.e., depth series for well A, N1 matrix, row number for serial number, data for depth in meters, N is total logging data point number) and find the j-th layer measurement row number k in A.
Taking discrete points of a k-w step k + w curve on A as a matching template patternA.
Find L and B onA(j, 1) the geological stratification number m (i.e. the mth layer on the B or the mth row of the B matrix) of the same geological stratification name, and further finding the geological stratification result of the B: depth LS(m, 2) and thickness LS(m，3)
According to LS(m, 2) geological stratification depth, taking depth L on BS(m，2)-wB1To LS(m，2)-wB2As a range for finding the best match.
The lower limit of the curve for x should not be lower than the depth of the m-1 th layer, i.e.
LS(m，2)-wB1>LS(m-1, 2); also, the upper limit of the curve for x should not be greater than the depth of the m +1 th layer, i.e., LS(m，2)+wB2<LS(m+1，2)。wB1And wB2Typically 50 meters.
After the depth range of the B well search area x is obtained, the depth range can be further searched according to B1(i.e., depth series for B-well) the number range for X is obtained.
From the curve of x, traversing from beginning to end length (x) points, taking p as the center and the depth of each offset wa at the ith point p as the xi position (manually defining an initialized depth parameter when the algorithm is realized, generally taking 4 meters), jumping and taking points by step to obtain a matching template pattenB on B, then calculating the similar distance between pattenA and pattenB, and obtaining the result di(i＝1，2，3，...，length(x))。
Theoretically, all d are locatediIn dkIs equal to min ({ d)iH), then, the childThe depth value corresponding to the kth serial number of the curve x is the obtained layering point on B, and the depth value corresponds to the depth value corresponding to the jth layering point on A one to one.
In practice, however, the theoretical minimum distance index k point may not be a true best match due to the presence of noise. The distance curve has a plurality of valley regions, and the wider valleys contain more points and have a greater probability of containing true stratification points.
Setting y points of the minimum distance curve, and taking the minimum distance points which are per% before the points to form a new sequence z (z is a matrix of two columns, the first column is a serial number, and the second column is a distance value); after z is sorted according to sequence numbers, the probability that the sequence of a plurality of sequence numbers contains a real layering point is higher when the sequence of the first column (namely, the sequence number column) of z is continuous on a mathematical integer.
For example, given pattenA, a search range is determined on the well B, that is, for all points between the well B serial numbers 6532 to 7532, each point is partially taken from the left and right to form a patternB, the similar shortest similarity distance is calculated with patternA, so that 7532 + 6532+1 is 101 distances, the composition curve is shown in fig. 7, and theoretically, the lowest part of the following distance map should be the hierarchical part of the well B, but because of the influence of noise, the determination is not scientific.
The first 12% of all distances are taken as small points, marked with a "+" sign, as shown in FIG. 8.
These red + dots were then analyzed, and the consecutive numbers were grouped together and sorted by the number of consecutive dots, for a total of 5 groups after analysis in the upper graph, as shown in the following table.
As can be seen, the largest valley concentration region is in segments 6919-6965, 47 points are continuously located in the first 12% valley region, the average serial number of the 47 points is 6942, which is most likely to be close to the real layering point.
If the taken per% is larger, the number of points is more, the number of groups of points is possibly more, if K reference layering points are required, the group average point serial numbers of the first K continuous point numbers are taken, after the serial numbers are obtained, for example, the serial numbers are n, the automatic layering depth of the computer of the well B is B1(n), namely the number corresponding to the nth row of the well B depth matrix, and the unit is meter. By this, automatic layering is completed.
In the step g, a multi-curve-based stratigraphic layering model is established:
almost completely consistent with step f, the difference is that the matching curve (known layered curve A) and the matched curve (unknown, requiring layered curve B) are no longer a single curve, but are curve groups, and therefore, the calculation of the similarity distance between pattenA and pattenA adopts multidimensional calculation.
Claims (3)
1. A method for intelligently dividing stratum horizon based on logging information is characterized by comprising the following steps:
a. finding the wave crest and the wave trough of the curve;
b. calculating the relative depth difference and the measurement coordinate difference between the two wells;
c. smoothing the input curve;
d. detecting the similarity of the single curves;
e. detecting the similarity of multiple curves;
f. establishing a stratigraphic layering model based on a single curve;
g. establishing a multi-curve-based stratigraphic layering model;
in the step d, the single curve similarity detection method comprises the following steps:
let reference template R be an M-dimensional vector, i.e. R ═ { R (1), R (2), … …, R (M), … …, R (M) }, each component being a number or a smaller vector; there is a template T to be tested which is an N-dimensional vector, i.e., T ═ { T (1), T (2), … …, T (N), … …, T (N) } again each component is a number or a smaller vector, M is not necessarily equal to N, and the dimensions of each component are the same;
because M is not necessarily equal to N, the similarity of R and T is calculated, firstly, the distance between each component of R and each component of T is calculated to form a matrix of M x N;
in step e, the multi-curve similarity detection method comprises the following steps:
(1) definition of multidimensional points:
ri is a component of the multidimensional point (a1, a2, …, an), and n components are provided;
tj is a component (b1, b2, …, bn) of the multidimensional point, and n components are provided;
wherein ai, bi (i ═ 1., n) belong to the same parameter;
(2) normalization of multidimensional points:
normalizing all parameters, wherein all parameters of the multidimensional points are in the range of 0-1;
the parametric sequence for each dimension, e.g., c ═ { ci }, is normalized as follows:
(3) calculation of distances between multidimensional points
d (i, j) represents the i row and j column values with the bottom left point as the initial coordinate origin, namely the Euclidean distance between the ith point of the original sequence R and the jth point of the test template T, and
in the step f, a stratigraphic layering model based on a single curve is established:
known (1) A log: a ═ ai ═ a1, a2, … a8, where ai is the column vector;
known (2) B log: b ═ bi, where bi are all column vectors;
as is known, (3) geological and artificial layering of A LA＝{LA1,LA2,LA3,LA4,LA5In which L isAiAre column vectors, representing respectively: geological stratificationName, geological stratification depth, geological stratification thickness, artificial stratification depth, artificial stratification thickness;
geological stratification of B, L, is known (4)B＝{LB1,LB2,B3In which L isBiAre column vectors, representing respectively: geological stratification name, geological stratification depth and geological stratification thickness;
for any j layers of A, by searching LARow j, the hierarchical data of the layer is obtained:
LA(j, 1) geological stratification name, LA(j, 2) geological stratification depth, LA(j, 3) geological stratification thickness, LA(j, 4) depth of Artificial layering, LA(j, 5) artificial delamination thickness;
according to LA(j, 4) artificial stratification depth, and a1The depth sequence of the well A, the matrix N x1, the line number represents the serial number, the data represents the depth in the unit of meter, N is the total number of logging data points, and the serial number k of the measurement line of the jth layer in the well A is found;
taking k-w as step, k + w curve discrete point on A as a matching template pattern A;
find L and B onA(j, 1) the geological stratification sequence number m of the same geological stratification name, namely the mth layer on the B or the mth row of the B matrix, further finding the geological stratification result of the B: depth of field
LB(m, 2) and thickness LB(m，3)
According to LB(m, 2) geological stratification depth, taking depth L on BB(m，2)-wS1To
LB(m，2)+wB2The section of the logging curve x is used as a range for finding the best matching;
the lower limit of the curve for x should not be lower than the depth of the m-1 th layer, i.e.
LB(m，2)-wS1＞LB(m-1, 2); also, the upper limit of the curve for x should not be greater than the depth of the m +1 th layer, i.e., LB(m，2)+wS2＜LB(m+1，2)；wB1And wB2Taking 50 meters generally;
get the B wellAfter searching the depth range of region x, further based on b1Obtaining the sequence number range of X;
from the curve x, traversing from the beginning to the end of length (x) points, taking p as the center, shifting the depth of wa left and right by taking step to jump and get points at the ith point p as the xi point to obtain a matching template pattern B on B, then calculating the similarity distance between the pattern A and the pattern B, and obtaining the result di(i＝1，2，3，...，length(x))；
Locate all diIn dkIs equal to min ({ d)iH), then, the depth value corresponding to the kth sequence number of the sub-curve x is the obtained layering point on B, and the depth value corresponds to the jth layering point on a one-to-one;
setting y points of the minimum distance curve, and taking the minimum distance points which are per% before the points to form a new sequence z; after z is sequenced according to sequence numbers, the continuous condition of the first column of z on mathematical integers is observed, the possibility of containing real layering points is higher, if the taken per% is higher, the number of points is more, the number of groups of points is more possible, if K reference layering points are given, the number of the group average points with the continuous number of points arranged at the top K is taken, and after the sequence numbers are obtained, for example, the sequence number is n, the automatic layering depth of the computer of the well B is B1(n), namely the number corresponding to the nth row of the well B depth matrix, the unit is meter; thus, automatic layering is completed;
in the step g, a multi-curve-based stratigraphic layering model is established: consistent with the establishment of a single-curve stratum layered model, the difference is that the matching curve and the matched curve are not single curves but curve groups, and therefore, the similarity distance calculation of pattenA and pattenab adopts multidimensional calculation.
2. The method of claim 1, wherein in the step a, the calculation process for finding the peaks and valleys of the curve is as follows:
given that the curve X ═ { X1, X2, X3, … xn }, X1 ═ { X2, X3, X4, … xn, xn +1}, let X2 ═ X1-X, the difference between neighboring points is obtained, and the points at the peaks meet one of the following two characteristics:
(1) x2 is negative at position j and positive at j-1, then the j position is the peak;
(2) and if the position j of the X2 is negative and the position j-1 is 0, continuing to search from the position j-1 until the position X2(j-m) is not equal to 0, and if the position X2(j-m) is positive, the positions m +1 from the position j-m to the position j have the same value and are all located in a flat peak, and taking the midpoint as a peak point, namely (j + j-m)/2 and then rounding.
3. The method for intelligently partitioning formation horizons based on logging information according to claim 1, wherein in the step c, the method for smoothing the input curve comprises the following steps: and (3) setting an input curve A, a smooth width w and a smooth back curve B, wherein the input curve A is not smooth for the first 1:1-w of the input curve A, and is not smooth for the last N-w: N of the input curve A, and only the smooth w: N-w part is as follows:
B(1:1-w)＝A(1:1-w)
B(N-w:N)＝A(N-w:N)
for any i range between w: N-w, there is
B (i) ═ mean (a (i-w: i + w)), mean is an average value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711310475.0A CN108562950B (en) | 2017-12-11 | 2017-12-11 | Method for intelligently dividing stratum horizon based on logging information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711310475.0A CN108562950B (en) | 2017-12-11 | 2017-12-11 | Method for intelligently dividing stratum horizon based on logging information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108562950A CN108562950A (en) | 2018-09-21 |
| CN108562950B true CN108562950B (en) | 2020-01-03 |
Family
ID=63530190
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711310475.0A Active CN108562950B (en) | 2017-12-11 | 2017-12-11 | Method for intelligently dividing stratum horizon based on logging information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108562950B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109763814B (en) * | 2019-01-09 | 2022-05-13 | 浙江财经大学 | Stratum matching visual analysis method based on multi-dimensional logging data |
| CN110334317B (en) * | 2019-07-11 | 2020-07-17 | 东北石油大学 | Automatic extraction method for picture type logging curve data |
| CN110502569A (en) * | 2019-08-16 | 2019-11-26 | 浙江财经大学 | A kind of standard well screen based on Discrete Choice Model selects visual analysis method |
| CN110656934B (en) * | 2019-10-08 | 2024-04-30 | 中国石油天然气股份有限公司 | Compaction-removing stratum comparison method for tight sandstone reservoir |
| CN111472765B (en) * | 2020-04-24 | 2023-07-14 | 北京月新时代科技股份有限公司 | Stratum partitioning method and device for target well |
| CN116201535B (en) * | 2023-02-06 | 2024-02-09 | 北京月新时代科技股份有限公司 | Automatic dividing method, device and equipment for oil and gas reservoir target well sign stratum |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7280932B2 (en) * | 2004-09-07 | 2007-10-09 | Landmark Graphics Corporation | Method, systems, and computer readable media for optimizing the correlation of well log data using dynamic programming |
| US7502691B2 (en) * | 2007-07-31 | 2009-03-10 | Baker Hughes Incorporated | Method and computer program product for determining a degree of similarity between well log data |
| CN102052070B (en) * | 2010-12-13 | 2013-06-05 | 中国石油集团川庆钻探工程有限公司 | Complex lithologic clastic rock stratum component logging processing method |
| CN105572727A (en) * | 2014-10-16 | 2016-05-11 | 中国石油化工股份有限公司 | Reservoir fluid identification method based on pore fluid parameter frequency dependence inversion |
| CN106545337A (en) * | 2016-11-25 | 2017-03-29 | 西南石油大学 | A kind of sedimentary micro Logging Identification Method based on support vector machine |
-
2017
- 2017-12-11 CN CN201711310475.0A patent/CN108562950B/en active Active
Non-Patent Citations (1)
| Title |
|---|
| 长垣油田密度曲线的重构与应用;陈志德，等;《大庆石油地质与开发》;20151031;123-126 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108562950A (en) | 2018-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108562950B (en) | Method for intelligently dividing stratum horizon based on logging information | |
| CN109800863B (en) | Logging phase identification method based on fuzzy theory and neural network | |
| CN107545577B (en) | Sedimentary facies image segmentation method based on neural network | |
| CN109828304B (en) | Deep learning-based method for predicting lithologic sequence model by using seismic data | |
| CN112435363B (en) | Cutter wear state real-time monitoring method | |
| CN108897066B (en) | Carbonate rock crack density quantitative prediction method and device | |
| CN106529667A (en) | Logging facies identification and analysis method based on fuzzy depth learning in big data environment | |
| CN106257498A (en) | Zinc flotation work condition state division methods based on isomery textural characteristics | |
| CN108952699B (en) | Intelligent stratum lithology identification method in complex geological drilling process | |
| CN111191741A (en) | Rock classification constraint inheritance loss method of rock recognition deep learning model | |
| Gu et al. | Complex lithofacies identification using improved probabilistic neural networks | |
| CN104992178A (en) | Tight sandstone fluid type identification method based on support vector machine simulation cross plot | |
| CN109902697A (en) | Multi-target detection method, device and mobile terminal | |
| CN111722283B (en) | Stratum velocity model building method | |
| CN113762320A (en) | Method and apparatus for estimating lithofacies by learning well logs | |
| CN106446514A (en) | Fuzzy theory and neural network-based well-log facies recognition method | |
| CN116935067A (en) | Drilling fluid bubble identification method based on image | |
| CN103310205B (en) | A kind of Handwritten Numeral Recognition Method and device | |
| CN116299665A (en) | LSTM surface wave inversion method, device and medium | |
| Wang et al. | Using DTW to measure trajectory distance in grid space | |
| CN111598340B (en) | Thin sand body plane spread prediction method based on fractional order Hilbert transform | |
| CN102830430A (en) | Horizon speed modeling method | |
| CN111159869B (en) | Automatic calculation method for quantitative parameters between sandstone uranium deposit and ancient river course | |
| CN104462792B (en) | Log data lithological stratum numerical value reduction method | |
| Duan et al. | Seismic facies analysis based on deep encoder clustering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |