Geophysical prediction method for deep potassium-rich brine reservoir
By combining the method of Volume Active Learning Machine and Probability Neural Network, the main components of earthquake attributes are extracted, and the ant tracking algorithm and hierarchical analysis method are used to solve the problem of prediction of deep potassium-rich brine reservoirs, and the efficient brine layer space is achieved Prediction and crack detection improve prediction accuracy and recombination rate.
Patent Information
- Application Number
- CN202411055140.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-29
- Estimated Expiration
- Not applicable · inactive patent
Smart Images

Figure CN119046658A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of geological survey, and in particular to a geophysical prediction method for deep potassium-rich brine reservoirs. Background Art
[0002] Potassium-rich brine layers and polyhalite-containing potassium-rich layers are quite different in terms of sedimentary environment, reservoir development location, reservoir physical parameters, etc., and cannot be predicted using the same method. The interpretation results of drilling and logging data show that vertically, potassium-rich brine reservoirs are mainly developed in T 1 j 4-5 From the sedimentary analysis, T1 j 4-5 The middle and lower parts are mainly salt basins and gypsum flats, with high salt-gypsum content and undeveloped reservoirs; the upper part is mainly gypsum flats, with thick layers of gypsum flats locally developed, which is conducive to the development of potassium-rich brine reservoirs. According to drilling, most of the potassium-rich brine reservoirs are located in T 1 j 4-5 The reservoir thickness is relatively large in the brittle rock section of the upper cycle; it is not developed in the low-speed plastic rock layer, and the reservoir thickness is thin, with an average single-layer thickness of 3.85m. At the same time, thicker reservoirs are mostly located near faults, because fault activity can not only form cracks and improve reservoir properties, but also cause interlayer fluid flow, further connecting the reservoir space.
[0003] From the perspective of electrical characteristics, the dolomite reservoir section as a whole exhibits high speed, high gamma characteristics, and low resistivity characteristics. The AC and Gr values are no different from those of the adjacent dense dolomite and gypsum rocks. The denser dolomite limestone and anhydrite show weaker low-density characteristics, but are comparable to mudstone, gypsum-bearing dolomite limestone, and anhydrite rocks. From the velocity-density intersection diagram of the dolomite reservoir and the surrounding rock formations, its wave impedance is much higher than that of plastic rock formations such as rock salt and soft gypsum, and slightly lower than that of the surrounding brittle formations, but most of them overlap and are difficult to effectively distinguish.
[0004] Since the dolomite reservoir is thin and has a small difference in wave impedance with the surrounding rock, its impact on seismic characteristics is weak; the seismic characteristics are mainly affected by the response of the lithology combination of the surrounding rock (especially the salt-gypsum layer), and the reservoir may be located at any phase and any amplitude position; it is very difficult to predict the reservoir directly using the correspondence between the reservoir and seismic attributes.
[0005] To sum up, it is necessary to make further innovations on existing technologies. Summary of the invention
[0006] In view of the technical problems existing in the above-mentioned background technology, the present invention proposes a geophysical prediction method for deep potassium-rich brine reservoirs which has a reasonable concept, strong adaptability and high flexibility, and high computational efficiency, inversion result signal-to-noise ratio and continuity.
[0007] In order to solve the above technical problems, the present invention provides a geophysical prediction method for deep potassium-rich brine reservoirs, which mainly includes the following steps:
[0008] 1) Brine reservoir prediction
[0009] Based on the prediction of potassium-rich rock layers by the convolution limit learning machine, the prediction of brine identification factors is carried out; first, the resistivity curve is fused with the residual mineral content curve to make the potassium-rich brine layer a low-value anomaly; then the brine layer probabilistic neural network inversion based on principal component analysis is carried out to extract the principal components of various seismic attribute bodies, and the extracted principal components are used as new seismic attributes for subsequent attribute combination selection, and the nonlinear mapping relationship between the sensitivity curve and the selected seismic attributes is calculated using the probabilistic neural network algorithm to achieve spatial prediction of brine layers;
[0010] 2) Crack prediction
[0011] In view of the characteristics of brine reservoirs as dense carbonate rocks, the multi-scale fracture prediction results reflected by curvature, maximum likelihood and coherence attributes are taken as prior information, and the optimal weight value is determined by the hierarchical analysis method. The optimal weight value is added to the ant tracking algorithm as a heuristic factor for fracture detection, realizing the three-dimensional characterization of the overlapping development area of brine layers and fractures, and identifying the high-yield enrichment zone of potassium-rich brine.
[0012] The geophysical prediction method for deep potassium-rich brine reservoirs, wherein the specific process of using the hierarchical analysis method to determine the optimal weight value in step 2) is: taking the research object as a system, making decisions according to the thinking mode of decomposition, comparison, judgment and synthesis, that is, starting from the root node, all categories contained in the node are first divided into two subclasses, and then the subclasses are divided separately to obtain two updated subclasses, and so on, until it cannot be divided, and finally an inverted binary tree is obtained.
[0013] The geophysical prediction method for deep potassium-rich brine reservoirs, wherein the specific process of adding the optimal weight value to the ant tracking algorithm as a heuristic factor for crack detection in step 2) is:
[0014] First, the fractures and cracks detected in the edge detection and coherence body are fused to make them more refined and continuously protrude the target area; the coherence value and edge detection value are used as prior information, and some electronic ants are artificially set in the coherent data body. The prior information is used as a heuristic factor. During the ant walking and iteration process, the electronic ants will walk along the most suitable crack path and then release pheromones. There will be no increase in places without cracks. After several iterations, the pheromone content in the data body can indicate the distribution of faults and cracks;
[0015] The edge attributes are then calculated from the coherent data, and other prior information is added as a heuristic factor. The heuristic factor is composed as follows:
[0016] η=w1·(1-C)+w2·E+w3·O (1);
[0017] In the above formula (1), η is the heuristic factor; C is the coherence value; E is the edge detection value; O is the value of other prior information; w1, w2 and w3 are the weight coefficients of other prior information;
[0018] Then the third generation coherence algorithm based on eigenvalue is used to calculate:
[0019] C 3 =λ max / Σλ i (2);
[0020] In the above formula (2), λ max is the maximum value of the matrix eigenvalue, Σλ i is the trace of the matrix;
[0021] Then divide the seismic data into blocks of a certain size and place an artificial electronic ant in each small area. Initially, each artificial ant is placed in this small area in a pseudo-random selection, and the pseudo-random selection formula is the same as the probability selection formula when the ant searches for the next point, as shown in the following formula (3):
[0022]
[0023] In the above formula (3), P i is the probability of each artificial ant being placed at the starting point i in the region at the beginning; C i , C j are the coherence values of points i and j in the region, respectively, which are used as weights to calculate the probability of selecting the starting point;
[0024] A random P∈(0,1), when ΣP i >p, the i-th point is the starting point of this area, and the best value point in the area is taken as the starting point;
[0025] After the ant selects the starting point, it can continue to look for the next point before the termination condition is met. The probability formula for finding the next point is:
[0026]
[0027] In the above formula (4), R is the set of all points within the visibility range of the ant's current position (i, j); τ is the pheromone; α and β are the weight coefficients of the pheromone and the heuristic factor;
[0028] Set a random number k∈(0,1), and when ∑P(i,j)>k, the current point (i,j) is used as the next point for the ant to go to; in this process, if the ant has not been forced to terminate, the ant will search along the place it thinks is reasonable. After being forced to terminate, if the walking path meets the requirements, the walking path of this time is recorded. This path is the fault line drawn by the ant, and then the current path is compared with the historical optimal or global optimal path to selectively update the pheromone, replacing the historical optimal or global optimal path; when the ant walks, it is carried out as a single ant, and the pheromone update this time is also carried out as a single individual:
[0029] τ ij =τ ij ·(1-ρ) (5);
[0030] τ ij =Δτ+(1-ρ)·τ ij (6);
[0031] τ ij =τ ij +Δτ (7);
[0032] In the above formulas (5)-(7), τ ij is the accumulated pheromone size at point (i, j); Δτ is the increase in pheromone each time it is updated; ρ is the volatilization factor; ρ∈(0,1) is used to control the amount of pheromone volatilized in a single time.
[0033] The geophysical prediction method for deep potassium-rich brine reservoirs, wherein: the edge detection value E is calculated using an omnidirectional eight-directional Sobel operator.
[0034] The geophysical prediction method for deep potassium-rich brine reservoirs, wherein the specific process of three-dimensional carving in step 2) is: based on the fracture data body, the interval range of its fracture development density is 0-0.2, based on this, 0.12 is selected as the threshold value, and those above 0.12 are considered to be fracture-rich areas, and three-dimensional fracture carving is performed.
[0035] The geophysical prediction method for deep potassium-rich brine reservoirs, wherein the specific process of implementing the potassium-rich brine high-yield enrichment zone in the step 2) is as follows: considering the comprehensive structural characteristics analysis, selecting the contour line of 3200 meters as the boundary, and considering the area shallower than 3200 meters as the structural high position area; based on the RGB attribute fusion map and the three-dimensional carving of the fracture body, depicting the area with relatively large fracture density, the area with relatively dense fracture development in the high and steep position is the potassium-rich brine enrichment development area; the common area of the structural high position area and the potassium-rich brine enrichment development area is located as the potassium-rich brine enrichment area.
[0036] By adopting the above technical solution, the present invention has the following beneficial effects:
[0037] The geophysical prediction method for deep potassium-rich brine reservoirs of the present invention is reasonably conceived, has strong adaptability and high flexibility, and has high computational efficiency, inversion result signal-to-noise ratio and continuity.
[0038] Aiming at the problems of deep burial, rapid lateral change and relative density of potassium-rich brine carbonate reservoirs, as well as the low resistivity and high wave impedance response characteristics of potassium-rich brine reservoirs, the present invention integrates the fluid-lithology sensitivity curve of the potassium-rich brine reservoir to construct a brine identification factor; carries out probabilistic neural network inversion based on principal component analysis, calculates the nonlinear mapping relationship between the sensitivity curve and seismic data, and realizes the spatial prediction of brine layers.
[0039] In view of the characteristics of brine reservoirs as dense carbonate rocks, the present invention uses the multi-scale fracture prediction results reflected by curvature, maximum likelihood and coherence attributes as prior information, and adopts the analytic hierarchy process (AHP) to determine the optimal weight value and add it to the ant tracking algorithm as a heuristic factor, which significantly improves the fracture prediction accuracy, realizes the three-dimensional characterization of the overlapping development zone of brine layers and fractures, and implements the high-yield enrichment zone of potassium-rich brine, with a prediction composite rate of 80%. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0041] Figure 1 It is a comparative analysis diagram of the resistivity curve of Dawan 102 well involved in the geophysical prediction method for deep potassium-rich brine reservoirs of the present invention and the predicted section of the potassium salt sensitivity factor of Dawan 102;
[0042] Figure 2 It is a plane section comparison diagram of the Dawan 102-Puguang 11 potassium curve prediction and brine factor prediction results involved in the geophysical prediction method for deep potassium-rich brine reservoirs of the present invention;
[0043] Figure 3 It is a prediction plane diagram of sensitivity factors of the potassium-rich brine layers of Jia 4 and 5 sections involved in the geophysical prediction method of deep potassium-rich brine reservoirs of the present invention;
[0044] Figure 4 It is a schematic diagram of the fracture prediction analytic hierarchy process involved in the geophysical prediction method for deep potassium-rich brine reservoirs of the present invention;
[0045] Figure 5 It is a schematic diagram of eight-azimuth Sobel operators involved in the geophysical prediction method of deep potassium-rich brine reservoirs of the present invention;
[0046] Figure 6 It is a bottom boundary fracture attribute map of the Lei 1st section in the Puguang area involved in the geophysical prediction method for deep potassium-rich brine reservoirs of the present invention; wherein Figure 6 (a) is the maximum likelihood function attribute graph, (b) is the Semblance attribute graph, (c) is the conventional ant tracking crack graph, and (d) is the heuristic factor crack attribute graph;
[0047] Figure 7 A three-dimensional engraving (display effect in different directions) diagram of the cracks of the Jia 4th and 5th segments in the Puguang area involved in the geophysical prediction method for deep potassium-rich brine reservoirs of the present invention;
[0048] Figure 8 It is a regional map of the Puguang area involved in the geophysical prediction method of deep potassium-rich brine reservoirs of the present invention; wherein, Figure 8 In the figure, (a) is the high area map of Jia 45 segment, (b) is the fracture development area map of Jia 45 segment;
[0049] Fig. 9 This is the enrichment distribution map of potassium-rich brine in the Jia 4th and 5th sections involved in the geophysical prediction method for deep potassium-rich brine reservoirs of the present invention. DETAILED DESCRIPTION
[0050] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0051] The present invention is further explained below in conjunction with specific implementation modes.
[0052] The present embodiment provides a geophysical prediction method for a deep potassium-rich brine reservoir, which specifically comprises the following steps:
[0053] S100, brine reservoir prediction
[0054] As mentioned above, it is very difficult to directly predict dolomite thin reservoirs using seismic attributes. However, when polyhalite is dissolved in water to form brine deposits, its resistivity should be reduced. If the polyhalite deposits are sandwiched between gypsum, dolomite, limestone, etc., the resistivity should be very high. Therefore, the resistivity characteristics are also the key to distinguishing brine layers. The amount of potassium salt content and the level of resistivity are technical indicators for distinguishing brine layers. Based on the prediction of potassium-rich rock layers by the positive limit learning machine in the previous chapter, the prediction of brine identification factors is carried out. First, the resistivity curve is fused with the residual mineral content curve (1-potassium salt content) to make the potassium-rich brine layer a low-value anomaly; then the probabilistic neural network inversion of the brine layer based on principal component analysis is carried out, and the first principal component, the second principal component, ..., the Pth principal component (P>=1) of various seismic attribute bodies such as seismic amplitude, frequency, and wave impedance are extracted by this technology, and the extracted principal components are used as new seismic attributes for subsequent attribute combination selection. The nonlinear mapping relationship between the sensitivity curve and the selected seismic attribute is calculated by the probabilistic neural network algorithm to realize the spatial prediction of the brine layer. The method has strong adaptability and high flexibility, and has high computational efficiency, signal-to-noise ratio and continuity of inversion results.
[0055] Figure 1 The black curve in the left figure is the potash curve of Dawan 102 Well, and the red curve is the brine identification factor curve of Dawan 102 Well. The resistivity at the depth of 3175.16m-3194.46m in the left figure is low, and the potash content is high, which can be preliminarily judged as a brine point. Combined with the potash content profile predicted by the volume active limit learning, it is preliminarily judged that there is a brine layer at the time 1.8S. The resistivity at the depth of 3407.64m-3453.32m, 3459.86-3496.02mc in the left figure is high, and the potash content is high, which can be preliminarily judged as a polyhalite point. Combined with the potash content profile predicted by the volume active limit learning, it is preliminarily judged that there is a polyhalite layer at the time 1.87-1.90S.
[0056] Compare the prediction results of potassium salt content by convolutional neural network learning machine and the prediction results of brine identification factor by principal component analysis probabilistic neural network ( Figure 2 ), it can be seen that the profile identification results of the two are consistent, and both can identify the potassium-rich brine layer at the top of Jia 4-5 section (marked by black dotted line), and the drilling results are highly consistent. From the plane comparison results, the low-value anomaly areas (red) predicted by the brine identification factor are all high-value anomalies in the potassium content prediction results (marked by red wireframe), which are distributed in strip-like, star-like, and irregular clusters, which is consistent with the previous geological understanding. It shows that the prediction results of the brine identification factor can effectively predict the spatial distribution of the potassium-rich brine layer.
[0057] In order to better describe potassium-rich brine, the inversion results of potassium-rich sensitive factors and brine identification factors obtained by machine learning were comprehensively applied, and the thickness plane distribution characteristics of the potassium-rich brine layer in the target layer of the Jialing River 4-5 section were obtained by joint analysis ( Figure 3 ).
[0058] S200, Crack Prediction
[0059] The brine reservoir of the Jialingjiang Formation has a low porosity and is generally dense. Through the analysis of Well Chuan-25, Well Bei-2 and Well ZK001, it can be seen that in the dense carbonate rock system, superimposed fractures can improve the reservoir and make it easier to form brine-rich bodies. From the perspective of brine reservoirs, the formation of large-scale faults and the communication with the potash layer and the underlying strata are conducive to the upward movement of fluids and the full transformation and migration of polyhalite, while the development of medium and small-scale faults plays a role in communicating and improving the reservoir. Therefore, faults of different levels are crucial to the characterization of brine reservoirs.
[0060] From the perspective of fault identification methods, relying solely on fracture and crack prediction technologies such as coherence, similarity, and maximum likelihood functions is unable to achieve multi-scale fracture prediction in this area. Therefore, in view of the characteristics of the Jiasiwu brine reservoir as a dense carbonate rock, the multi-scale fracture prediction results reflected by curvature, maximum likelihood, and coherence attributes are used as prior information, and the optimal weight value is determined by the analytic hierarchy process (AHP) and added to the ant tracking algorithm as a heuristic factor to improve the accuracy of fracture prediction, achieve three-dimensional characterization of the superimposed development area of brine layers and fractures, and implement the high-yield enrichment zone of potassium-rich brine.
[0061] S210, Analytical Hierarchy Process
[0062] The analytic hierarchy process treats the research object as a system and makes decisions according to the thinking mode of decomposition, comparison, judgment and synthesis. It has become an important tool for system analysis developed after mechanism analysis and statistical analysis. The idea is: starting from the root node, all the categories contained in the node are divided into two subcategories, and then the subcategories are divided separately to obtain two updated subcategories, and so on, until it is impossible to divide, and finally an inverted binary tree can be obtained. In this paper, cracks are mainly divided into four scales, namely four types and two levels, which are divided into large-scale fractures, medium-scale fractures, small-scale fractures and micro-cracks ( Figure 4 ).
[0063] This method is easy to calculate, and the results obtained are simple and clear. It requires less quantitative data information and transforms complex decision-making steps into simple weight calculations.
[0064] S220, Ant-tracking crack detection based on heuristic factors
[0065] Ant colony algorithm is a biologically inspired swarm intelligence optimization algorithm, which was first proposed by Dorigo in 1992. Ant colony algorithm is a biologically inspired swarm intelligence optimization algorithm proposed by simulating the natural phenomenon that ants use biotin to move between food and nests. The basic idea of using ants to track and detect cracks is: first, the discontinuity of seismic signals is enhanced by edge detection methods, and then a large number of such electronic "ants" are set in the processed seismic data, and each "ant" moves forward along the possible discontinuity surface while emitting "pheromones". The "ants" moving forward along the fault can track the discontinuity surface. If they encounter the expected discontinuity surface, they will use pheromones to make very obvious marks, while those surfaces that are unlikely to be faults or cracks will not be marked or only marked less obviously. Finally, the discontinuity surface is tracked by different pheromone concentrations, so as to achieve the purpose of detecting faults or cracks.
[0066] The analytic hierarchy process (AHP) is used to determine the optimal weight value and add it to the ant tracking algorithm as a heuristic factor for ant tracking crack detection. The specific process is as follows: the fractures and cracks detected in the edge detection and coherence body are fused to make them more refined and continuously protrude from the target area; the coherence value and edge detection value are used as prior information, and some electronic ants are artificially set in the coherent data body. The prior information is used as a heuristic factor. During the ant walking and iteration process, the electronic ants will walk along the most suitable crack path and then release pheromones. There will be no increase in places where there are no cracks. In this way, after several iterations, the pheromone content in the data body can indicate the distribution of faults and cracks.
[0067] The edge attributes are calculated from the coherent data, and then other prior information is added as a heuristic factor. The heuristic factor is composed as follows:
[0068] η=w1·(1-C)+w2·E+w3·O (1);
[0069] Where: η is the heuristic factor; C is the coherence value; E is the edge detection attribute value; O is the value of other prior information; w1, w2 and w3 are the weight coefficients of other prior information.
[0070] The heuristic factor is not determined by a single factor, but by the linear weighting of several factors, which can retain global information and limit the solution to a reasonable range. The third-generation coherence algorithm based on eigenvalues is used to calculate the coherence value:
[0071] C 3 =λ max / ∑λ i (2);
[0072] Where: max is the maximum value of the matrix eigenvalue, ∑λi is the trace of the matrix.
[0073] The Sobel edge operator is selected to calculate the edge amplitude of seismic attributes, but since the usual calculation only calculates the X and Y directions, it cannot contain edge information well ( Figure 5 ). Therefore, the edge detection value of the present invention is calculated by using the omnidirectional eight-directional Sobel operator, and its calculation template is as follows Figure 5 As shown:
[0074] Conventional ant tracking randomly selects the starting point when spreading ants, and randomness carries strong uncertainty. If there are few or no ants in a certain area, even if the prior information and coherence values indicate that there are abnormal values in the area, the cracks in the area will not be detected. Taking the two-dimensional case as an example, the seismic data (image) is divided into blocks of a certain size (such as 3x3, 5x5) and an artificial ant is placed in each small area. Initially, each artificial ant is placed in this small area as a pseudo-random selection, and the pseudo-random selection formula is the same as the probability selection formula when the ant searches for the next point, as shown in formula (3):
[0075]
[0076] Where: P i is the probability of each artificial ant being placed at the starting point i in the area (assuming the ant moves from i to j); C i , C j are the coherence values of points i and j in the region, respectively, and are used as weights for calculating the probability of selecting the starting point.
[0077] A random P∈(0,1), when ∑P i >p, the i-th point is the starting point of this area. In order to facilitate rapid convergence, the best value point in the area (the best point may be the fracture crack) can also be selected as the starting point.
[0078] Fault line drawing:
[0079] After the ant selects the starting point, it can continue to look for the next point before the termination condition is met. The probability formula for finding the next point is:
[0080]
[0081] Where: R is the set of all points within the visibility range of the ant's current position (i, j); τ is the pheromone; α and β are the weight coefficients of the pheromone and the heuristic factor.
[0082] Similar to the pseudo-random selection of the starting point described above, a random number k∈(0,1) is set. When ∑P(i,j)>k, we use the current point (i,j) as the next point for the ant to go. In this process, if the ant has not been forced to terminate, the ant will search along the place it thinks is reasonable. After being forced to terminate, if the walking path meets the requirements, the walking path of this time is recorded. This path is the fault line drawn by the ant, and then the current path is compared with the historical optimal or global optimal path to selectively update the pheromone, replacing the historical optimal or global optimal. When the ant walks, it is carried out as a single ant. This pheromone update is also carried out on a single individual basis, which can eliminate the complicated statistical work.
[0083] τ ij =τ ij ·(1-ρ) (5);
[0084] τ ij =Δτ+(1-ρ)·τ ij (6);
[0085] τ ij =τ ij +Δτ (7);
[0086] Where: τ ij is the accumulated pheromone size at point (i, j); Δτ is the increase in pheromone each time it is updated; ρ is the volatilization factor; ρ∈(0,1) controls the amount of pheromone volatilized in a single time.
[0087] First, the pheromone evaporates according to formula (5), just like the actual biotin evaporation, and then increases according to formula (6). Combining formula (5) and formula (6) is formula (7). The pheromone update follows certain rules. The increased pheromone is less than the evaporated pheromone. The concentration of pheromone in the entire data (image) will become lower and lower, but the pheromone concentration at the fracture will become stronger relative to other positions as the number of iterations increases. After one iteration is completed, the path of the artificial ant may not be the optimal path, but as the number of iterations increases, the optimal path of the ant will be infinitely close to the fracture line.
[0088] Figure 6This is a plane map of the fractures in the Jia 4th and 5th sections of the Puguang area. It can be seen that both the northeast and northwest groups of fractures have been clearly characterized, showing the identification of faults and fractures in two major tectonic periods. Conventional fracture detection methods such as coherence, edge detection and conventional ant tracking algorithms have good recognition capabilities for large-scale fractures (fractures), but the resolution is low and the recognition accuracy for small-scale fractures is insufficient. These small-scale fracture development areas are often the focus of tight carbonate reservoir exploration. The conventional ant tracking fracture detection results can only identify the location and general shape of the fracture zone, and cannot reflect the detailed information inside the fracture zone. The improved fracture detection results have significantly improved resolution and clearly display small-scale fracture information, indicating that the ant tracking fracture detection method based on heuristic factors can greatly improve the accuracy of fracture detection.
[0089] S230, 3D engraving
[0090] With the improvement of computer performance, the 3D attribute carving technology has increasingly shown its superiority. By carving the seismic data body, we can observe the data body within the attribute range we need in a more three-dimensional and objective way. This topic is based on the fracture data body, and the interval range of its fracture development density is 0-0.2. Based on this, we choose 0.12 as the threshold value, and believe that the area above 0.12 is a fracture development-rich area, and perform 3D fracture carving, such as Figure 7 As shown, the left and right are three-dimensional sculptures in two different directions. From the figure, we can clearly see the high and low relationship of the structural parts of the Puguang area. At the same time, it is also easy to get that the cracks are mainly developed in the high and steep structural belt in the northeast direction, and the crack density on both sides of the structural belt is weaker.
[0091] S240, potassium-rich brine enrichment area
[0092] Through the previous single well dissection of Chuan 25 Well, Bei 2 Well and ZK001 Well, it is known that the reservoir of potassium-rich brine is mainly developed in the relatively high part of the structure. Considering the comprehensive structural characteristics, the contour line of 3200 meters is selected as the boundary, and the area shallower than 3200 meters is considered to be the high part of the structure. The following is obtained: Figure 8 As shown in a, the light yellow area is the structural high area, with a total area of 244 square kilometers. At the same time, based on the RGB attribute fusion map and the three-dimensional carving of the fracture body, the area with relatively large fracture density is depicted, such as Figure 8 As shown in b, the total area is 192 square kilometers. According to the previous analysis, the area with dense crack development in high and steep positions is our potassium-rich brine enrichment development area, so we will Figure 8 a and Figure 8 b common areas locate potassium-rich brine enrichment areas, such as Fig. 9 As shown, its total area is 105.9 square kilometers.
[0093] The invention has reasonable conception, strong adaptability and high flexibility, high computational efficiency, high signal-to-noise ratio and continuity of inversion results, and is suitable for promotion and application.
[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A geophysical prediction method for deep potassium-rich brine reservoirs, characterized in that: The geophysical prediction method mainly comprises the following steps: 1) Brine reservoir prediction Based on the prediction of potassium-rich rock layers by the convolution limit learning machine, the prediction of brine identification factors is carried out; first, the resistivity curve is fused with the residual mineral content curve to make the potassium-rich brine layer a low-value anomaly; then the brine layer probabilistic neural network inversion based on principal component analysis is carried out to extract the principal components of various seismic attribute bodies, and the extracted principal components are used as new seismic attributes for subsequent attribute combination selection, and the nonlinear mapping relationship between the sensitivity curve and the selected seismic attributes is calculated using the probabilistic neural network algorithm to achieve spatial prediction of brine layers; 2) Crack prediction In view of the characteristics of brine reservoirs as dense carbonate rocks, the multi-scale fracture prediction results reflected by curvature, maximum likelihood and coherence attributes are taken as prior information, and the optimal weight value is determined by the hierarchical analysis method. The optimal weight value is added to the ant tracking algorithm as a heuristic factor for fracture detection, realizing the three-dimensional characterization of the overlapping development area of brine layers and fractures, and identifying the high-yield enrichment zone of potassium-rich brine.
2. The geophysical prediction method for deep potassium-rich brine reservoirs according to claim 1, characterized in that: The specific process of using the hierarchical analysis method to determine the optimal weight value in the step 2) is: taking the research object as a system, making decisions according to the thinking mode of decomposition, comparison, judgment and synthesis, that is, starting from the root node, all the categories contained in the node are first divided into two subclasses, and then the subclasses are divided separately to obtain two updated subclasses, and so on, until it cannot be divided, and finally an inverted binary tree is obtained.
3. The geophysical prediction method for deep potassium-rich brine reservoirs according to claim 1, characterized in that: The specific process of adding the optimal weight value to the ant tracking algorithm as a heuristic factor for crack detection in step 2) is as follows: First, the fractures and cracks detected in the edge detection and coherence body are fused to make them more refined and continuously protrude the target area; the coherence value and edge detection value are used as prior information, and some electronic ants are artificially set in the coherent data body. The prior information is used as a heuristic factor. During the ant walking and iteration process, the electronic ants will walk along the most suitable crack path and then release pheromones. There will be no increase in places without cracks. After several iterations, the pheromone content in the data body can indicate the distribution of faults and cracks; The edge attributes are then calculated from the coherent data, and other prior information is added as a heuristic factor. The heuristic factor is composed as follows: η=w1·(1-C)+w2·E+w3·O (1); In the above formula (1), η is the heuristic factor; C is the coherence value; E is the edge detection value; O is the value of other prior information; w1, w2 and w3 are the weight coefficients of other prior information; Then the third generation coherence algorithm based on eigenvalue is used to calculate: C 3 =λ max / Sl i (2); In the above formula (2), λ max is the maximum value of the matrix eigenvalue, Σλ i is the trace of the matrix; Then divide the seismic data into blocks of a certain size and place an artificial electronic ant in each small area. Initially, each artificial ant is placed in this small area in a pseudo-random selection, and the pseudo-random selection formula is the same as the probability selection formula when the ant searches for the next point, as shown in the following formula (3): In the above formula (3), P i is the probability of each artificial ant being placed at the starting point i in the region at the beginning; C i , C j are the coherence values of points i and j in the region, respectively, which are used as weights to calculate the probability of selecting the starting point; A random P∈(0,1), when ΣP i >p, the i-th point is the starting point of this area, and the best value point in the area is taken as the starting point; After the ant selects the starting point, it can continue to look for the next point before the termination condition is met. The probability formula for finding the next point is: In the above formula (4), R is the set of all points within the visibility range of the ant's current position (i, j); τ is the pheromone; α and β are the weight coefficients of the pheromone and the heuristic factor; Set a random number k∈(0,1), and when ∑P(i,j)>k, the current point (i,j) is used as the next point for the ant to go to; in this process, if the ant has not been forced to terminate, the ant will search along the place it thinks is reasonable. After being forced to terminate, if the walking path meets the requirements, the walking path of this time is recorded. This path is the fault line drawn by the ant, and then the current path is compared with the historical optimal or global optimal path to selectively update the pheromone, replacing the historical optimal or global optimal path; when the ant walks, it is carried out as a single ant, and the pheromone update this time is also carried out as a single individual: t ij =t ij ·(1-p) (5); t ij =Δτ+(1-ρ)·τ ij (6); t ij =t ij +Δt (7); In the above formulas (5)-(7), τ ij is the accumulated pheromone size at point (i, j); Δτ is the increase in pheromone each time it is updated; ρ is the volatilization factor; ρ∈(0,1) is used to control the amount of pheromone volatilized in a single time.
4. The geophysical prediction method for deep potassium-rich brine reservoirs according to claim 3, characterized in that: The edge detection value E is calculated using an omnidirectional eight-directional Sobel operator.
5. The geophysical prediction method for deep potassium-rich brine reservoirs according to claim 1, characterized in that: The specific process of three-dimensional carving in step 2) is as follows: based on the fracture data volume, the interval range of the fracture development density is 0-0.
2. Based on this, 0.12 is selected as the threshold value, and the area above 0.12 is considered to be a fracture development-rich area, and three-dimensional fracture carving is performed.
6. The geophysical prediction method for deep potassium-rich brine reservoirs according to claim 1, characterized in that: The specific process of implementing the potassium-rich brine high-yield enrichment zone in the step 2) is as follows: considering the comprehensive structural characteristics analysis, selecting the contour line of 3200 meters as the boundary, and considering the area shallower than 3200 meters as the structural high position area; based on the RGB attribute fusion map and the three-dimensional carving of the fracture body, depicting the area with relatively large fracture density, the area with dense fracture development in high and steep positions is the potassium-rich brine enrichment development area; the common area of the structural high position area and the potassium-rich brine enrichment development area is located as the potassium-rich brine enrichment area.
Citation Information
Patent Citations
Brine storage layer prediction and resource quantity evaluation method for underground brine type potassium ore and lithium ore
CN112878998A
Earthquake reservoir comprehensive prediction method
CN115857047A
Lithium-rich brine spatial distribution prediction method and system and electronic equipment
CN117492108A
Method for determining space distribution of brine type potassium ore and lithium ore and method for estimating resource quantity
CN117849897A
Cited By
Method for predicting depth of gas-liquid interface for gas injection and brine discharge and related equipment
CN119777810A
