Five-in-one based oil and uranium exploration and prospecting method and device and electronic equipment

Abstract

The invention discloses a five-in-one oil-uranium exploration and prospecting technology, and the technology achieves the secondary development and utilization of a large amount of oil logging information of an oil field through comprehensive radioactivity anomaly screening, screens out the reliable radioactivity anomaly, and determines a uranium prospecting target region. When the natural gamma data of the shallow part of the oil field are missing and cannot be screened, an old well retest technology is adopted for supplement; through a natural gamma and quantitative gamma conversion technology, establishing a correlation between gamma anomaly of a uranium mineralization section verification hole and GR anomaly of an oil field hole, and assisting in estimating the uranium ore resource quantity; through a three-dimensional seismic data application technology, according to uranium exploration geological requirements, sandstone type uranium mine three-dimensional seismic data fault interpretation technology application countermeasures and methods are established; through typical ore deposit application research, typical sandstone type uranium metallogenic favorable conditions of main uranium-producing basins in northern China are analyzed, a technical system for exploration and evaluation of sandstone type uranium ore in a working area of a basin oilfield is established, accurate evaluation and pit determination are achieved, a favorable zone is determined, and exploration deployment is guided.

Description

Technical Field

[0001] The present invention relates to the field of geophysical technology, and in particular to a five-in-one oil and uranium prospecting method, device and electronic equipment. Background Art

[0002] In the current exploration of mineral resources, oil, gas and uranium are important energy and strategic resources, and their exploration and development are of great significance to national energy security and economic development. In recent years, a number of large and super-large sandstone-type uranium deposits have been discovered in the Mesozoic and Cenozoic basins in northern my country, and remarkable results have been achieved. The coexistence and interaction of sandstone-type uranium deposits and oil and gas in the same basin, as well as the idea of ​​"looking for uranium with oil and looking for oil with uranium" have also been widely recognized. According to the prospecting idea of ​​"exploring oil and uranium at the same time", the oil and gas drilling logging data are reinterpreted and secondary developed, abnormal boreholes are screened, key prospecting target areas are delineated and drilling verification is carried out, and uranium mineralized areas are discovered. The key to this idea is to determine the radioactive anomaly indicators of oil and gas boreholes, so as to check the radioactive anomaly characteristics of the boreholes (potential uranium mine holes, potential uranium mineralized holes, normal holes, and non-parameter holes) and conduct drilling verification. However, in actual work, the following problems are encountered:

[0003] (1) Through the collection and investigation of drilling data from major domestic oil fields, it is found that the developed oil fields actually retain relatively comprehensive and complete data. Through the database construction technology that has been mastered, the drilling and logging data can be exported as data files. Developers and technicians with professional backgrounds often need to spend a lot of time and energy to query and process data in their daily work. The screening of abnormal data often needs to be done manually. The defects of this method are very obvious. The query speed is slow, time-consuming and labor-intensive, and the efficiency is low. It cannot meet the requirements of efficient office work.

[0004] (2) Although most oilfield drilling data are complete and informative, the oil reservoirs for oil and gas field exploration and development are often deep, while the sandstone-type uranium prospecting target layers are shallow. Therefore, there are many missing logging data in the shallow part, which leads to the loss of radioactive information in the shallow part and makes it impossible to determine the mineralization of the strata. This restricts the control of the hidden mineral-bearing geological bodies in the region and delays the prospecting process.

[0005] (3) The radioactive information of oilfield logging data is measured in natural gamma (API), but this cannot effectively determine the radioactive intensity, nor can it be used as a basis for determining the potential mineralization of the borehole. It cannot be compared with known nuclear industry borehole logging information and lacks uniformity.

[0006] (4) As a relatively mature geophysical exploration method, seismic exploration technology has been widely used in resource exploration such as oil and natural gas, but it is less used in sandstone uranium exploration research. In addition, due to the complexity and diversity of geological structures, the fault identification results based on seismic data in the current sandstone uranium exploration often rely on the experience and judgment of the interpreters, which is subjective, and the algorithm often cannot accurately capture the characteristics of all faults.

[0007] (5) Many sandstone-type uranium deposits have been discovered in the Ordos-Erlian Basin. However, the ore-controlling factors, enrichment patterns, and mineralization models of each deposit are still incomplete, and the understanding is not systematic and comprehensive enough. Further sorting out is urgently needed to improve understanding.

[0008] That is, the traditional single exploration method is often difficult to meet the exploration needs of oil, gas and uranium at the same time, and there are problems such as long exploration cycle, high cost and low efficiency. Especially in the complex geological environment where sandstone uranium deposits coexist with oil and gas resources, a single exploration technology is often difficult to accurately identify the distribution of ore bodies, resulting in inaccurate resource assessment and affecting subsequent development decisions. Summary of the invention

[0009] In view of this, the purpose of the present invention is to provide a five-in-one oil and uranium prospecting method, device and electronic equipment to solve the problems raised in the above background technology. Specifically, the present invention provides a five-in-one oil and uranium prospecting method, device and electronic equipment based on data screening technology, old well retest technology, natural gamma and quantitative gamma conversion technology, application of three-dimensional seismic data and typical mineral deposit application to solve the problems raised in the above background technology.

[0010] Moreover, another object of the present invention is to provide a fault identification model based on deep learning applied to seismic exploration of sandstone-type uranium mines, which is suitable for sandstone-type uranium mines and significantly improves the accuracy of fault identification to solve the problems raised in the above-mentioned background technology.

[0011] Moreover, another object of the present invention is to provide a loss function suitable for a deep learning-based fault identification model used in sandstone-type uranium mine seismic exploration, which significantly improves the convergence speed of the deep learning-based fault identification model to solve the problems raised in the above-mentioned background technology.

[0012] In order to achieve the above object, the present invention adopts the following technical solution:

[0013] First, a method for prospecting for oil and uranium based on five-in-one exploration is provided, including:

[0014] Step S1: Conducting data screening, wherein step S1 includes:

[0015] Step S11: Collect, investigate and organize drilling information of the oil field;

[0016] Step S12: analyzing the borehole radioactivity anomaly standard and constructing radioactivity anomaly screening software;

[0017] Step S13: Based on the radioactivity anomaly screening software, the radioactivity anomaly standard of the borehole is used for identification, and radioactivity anomaly wells are screened out;

[0018] Step S2: retest the old well, wherein step S2 includes:

[0019] Step S21: establishing a casing-cement ring model;

[0020] Step S22: carrying out a cement sheath casing logging response test;

[0021] Step S23: analyzing the functional relationship between the counting rate and the cement sheath property parameters to determine the correction coefficient;

[0022] Step S24: obtaining accurate shallow stratum radioactivity information based on the correction coefficient;

[0023] Step S3: Conduct drilling construction verification to achieve the conversion of natural gamma into quantitative gamma, wherein step S3 includes:

[0024] Step S31: Collecting gamma data of verification holes and oilfield drilling holes;

[0025] Step S32: Correcting the verification hole gamma logging depth to the oilfield drilling hole gamma abnormal depth range;

[0026] Step S33: establishing a regression equation between natural gamma and quantitative gamma;

[0027] Step S34: using the regression equation to convert natural gamma to quantitative gamma and evaluate mineralization;

[0028] Step S4: Processing and interpreting 3D seismic data to analyze faults, sand bodies and structures in the mining area;

[0029] Step S5: Conduct typical mineral deposit comparison, wherein step S5 includes:

[0030] Step S51: Collect typical uranium deposit data;

[0031] Step S52: analyzing structural sedimentary evolution characteristics, uranium source conditions, sand bodies and sedimentary systems, ore body characteristics, and mineralization laws;

[0032] Step S53: Summarize the ore-controlling factors of typical sandstone-type uranium deposits;

[0033] Step S6: Based on the results of the above steps S3 to S5, determine the favorable uranium ore zones and evaluate the mineralization potential of the working area.

[0034] Furthermore, the step S4 further comprises:

[0035] S41: Determine the structural characteristics of the uranium mining area using seismic fine structural interpretation technology and structural and sedimentary evolution analysis technology;

[0036] S42: Use the seismic stage-grading fault fine characterization technology to determine the uranium fault, oil and gas fault and oil and gas migration channel;

[0037] S43: Determine the top and bottom condition characteristics of uranium sand bodies using sand body top and bottom seismic identification technology;

[0038] S44: Use formation reservoir fine description technology to determine the distribution, thickness and physical properties of uranium sandstone.

[0039] Furthermore, the use of seismic stage-by-stage classification fault fine characterization technology to determine the uranium fault, oil and gas fault, and oil and gas migration channel in step S42 includes:

[0040] A seismic deep learning model is used to finely characterize uranium mine faults. The seismic deep learning model includes an encoder-decoder module, and the encoder-decoder module includes:

[0041] Encoder unit, fourth Swin Transformer unit, decoder unit, skip connection unit;

[0042] The encoder unit includes: an image block segmentation unit, an image block embedding unit, a first Swin Transformer unit, a first image block merging unit, a second Swin Transformer unit, a second image block merging unit, a third Swin Transformer unit, and a third image block merging unit, and the image block segmentation unit outputs to the image block embedding unit, the image block embedding unit outputs to the first Swin Transformer unit, the first Swin Transformer unit outputs to the first image block merging unit, the first image block merging unit outputs to the second Swin Transformer unit, the second Swin Transformer unit outputs to the second image block merging unit, the second image block merging unit outputs to the third Swin Transformer unit, and the third Swin Transformer unit outputs to the third image block merging unit;

[0043] The third image block merging unit outputs to the fourth Swin Transformer unit, and the fourth Swin Transformer unit outputs to the first image block expansion unit;

[0044] The decoder unit includes: a first image block expansion unit, a fifth Swin Transformer unit, a second image block expansion unit, a sixth Swin Transformer unit, a third image block expansion unit, a seventh Swin Transformer unit, a final image expansion module, and a linear projection module, and the first image block expansion unit outputs to the fifth Swin Transformer unit, the fifth Swin Transformer unit outputs to the second image block expansion unit, the second image block expansion unit outputs to the sixth Swin Transformer unit, the sixth Swin Transformer unit outputs to the third image block expansion unit, the third image block expansion unit outputs to the seventh Swin Transformer unit, the seventh Swin Transformer unit outputs to the final image expansion module, and the final image expansion module outputs to the linear projection module;

[0045] The encoder unit is connected to the decoder unit through a skip connection unit.

[0046] Further, the encoder unit is connected to the decoder unit via a jump connection unit and comprises:

[0047] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the first jump connection unit, and then the output result of the first jump connection unit is concatted with the output of the first image block expansion unit and input into the fifth Swin Transformer unit;

[0048] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the second jump connection unit, and then the output result of the second jump connection unit is concatted with the output of the second image block expansion unit and input into the sixth Swin Transformer unit;

[0049] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the third jump connection unit. Then, the output result of the third jump connection unit is concatted with the output of the second image block expansion unit and input into the seventh Swin Transformer unit.

[0050] Further, the jump connection unit includes: a first jump connection unit, a second jump connection unit, and a third jump connection unit, and the first jump connection unit, the second jump connection unit and the third jump connection unit have the same structure;

[0051] The first jump connection unit includes: the eighth Swin Transformer unit and the channel mixing module;

[0052] The channel mixing module includes two branches, the first branch includes a maximum pooling layer and a 1×1 convolution layer, and the second branch includes an average pooling layer and a 1×1 convolution layer; the output result of the first branch and the output result of the second branch are added, and then passed to the 1×1 convolution layer and the sigmoid activation layer to obtain the output result of the channel mixing module;

[0053] The output result of the eighth Swin Transformer unit and the output result of the channel mixing module are multiplied to obtain the output result of the first jump connection unit.

[0054] Furthermore, the deep learning model includes two encoder-decoder modules.

[0055] The result of the last Swin Transformer unit in the decoder unit in the first encoder-decoder module is output to the first Swin Transformer unit in the encoder unit in the second encoder-decoder module, and a jump connection unit is used to establish a jump connection between the Swin Transformer unit in the decoder unit in the first encoder-decoder module and the Swin Transformer unit in the encoder unit in the second encoder-decoder module. The specific structure is as follows:

[0056] The earthquake deep learning model includes:

[0057] A first encoder-decoder module, a second encoder-decoder module;

[0058] The first encoder-decoder module comprises:

[0059] First encoder unit, fourth Swin Transformer unit, first decoder unit, skip connection unit;

[0060] The first encoder unit includes: an image block segmentation unit, an image block embedding unit, a first Swin Transformer unit, a first image block merging unit, a second Swin Transformer unit, a second image block merging unit, a third Swin Transformer unit, and a third image block merging unit, and the image block segmentation unit outputs to the image block embedding unit, the image block embedding unit outputs to the first Swin Transformer unit, the first Swin Transformer unit outputs to the first image block merging unit, the first image block merging unit outputs to the second Swin Transformer unit, and the second Swin

[0061] The Transformer unit outputs to the second image block merging unit, and the second image block merging unit outputs to the third Swin

[0062] The Transformer unit and the third Swin Transformer unit output to the third image block merging unit;

[0063] The third image block merging unit outputs to the fourth Swin Transformer unit, and the fourth Swin Transformer unit outputs to the first image block expansion unit;

[0064] The first decoder unit includes: a first image block expansion unit, a fifth Swin Transformer unit, a second image block expansion unit, a sixth Swin Transformer unit, a third image block expansion unit, a seventh Swin Transformer unit, a final image expansion module, and a linear projection module, and the first image block expansion unit is output to the fifth Swin Transformer unit, the fifth Swin Transformer unit is output to the second image block expansion unit, the second image block expansion unit is output to the sixth Swin Transformer unit, the sixth Swin Transformer unit is output to the third image block expansion unit, the third image block expansion unit is output to the seventh Swin Transformer unit, the seventh Swin Transformer unit is output to the final image expansion module, and the final image expansion module is output to the linear projection module;

[0065] And the seventh Swin Transformer unit outputs to the eighth Swin Transformer unit;

[0066] The second encoder-decoder module comprises:

[0067] The second encoder unit, the eleventh Swin Transformer unit, the second decoder unit, and the skip connection unit;

[0068] The second encoder unit includes: an eighth Swin Transformer unit, a fourth image block merging unit, a ninth Swin Transformer unit, a fifth image block merging unit, a tenth Swin Transformer unit, and a sixth image block merging unit, and the eighth Swin Transformer unit outputs to the fourth image block merging unit, the fourth image block merging unit outputs to the ninth Swin Transformer unit, the ninth Swin Transformer unit outputs to the fifth image block merging unit, the fifth image block merging unit outputs to the tenth Swin Transformer unit, and the tenth Swin Transformer unit outputs to the sixth image block merging unit;

[0069] And the sixth image block merging unit outputs to the eleventh Swin Transformer unit, and the eleventh Swin Transformer unit outputs to the fourth image block expanding unit;

[0070] The second decoder unit includes: a fourth image block expansion unit, a twelfth Swin Transformer unit, a fifth image block expansion unit, a thirteenth Swin Transformer unit, a sixth image block expansion unit, a fourteenth Swin Transformer unit, a final image expansion module, and a linear projection module, and the fourth image block expansion unit outputs to the twelfth Swin Transformer unit, the twelfth Swin Transformer unit outputs to the fifth image block expansion unit, the fifth image block expansion unit outputs to the thirteenth Swin Transformer unit, the thirteenth Swin Transformer unit outputs to the sixth image block expansion unit, the sixth image block expansion unit outputs to the fourteenth Swin Transformer unit, the fourteenth Swin Transformer unit outputs to the final image expansion module, and the final image expansion module outputs to the linear projection module;

[0071] The first encoder unit is connected to the first decoder unit via a skip connection unit;

[0072] The second encoder unit is connected to the second decoder unit via a skip connection unit;

[0073] The first decoder unit is connected to the first encoder unit via a skip connection unit;

[0074] Further, the first encoder unit is connected to the first decoder unit via a skip connection unit and comprises:

[0075] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the first jump connection unit, and then the output result of the first jump connection unit is concatted with the output of the first image block expansion unit and input into the fifth Swin Transformer unit;

[0076] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the second jump connection unit, and then the output result of the second jump connection unit is concatted with the output of the second image block expansion unit and input into the sixth Swin Transformer unit;

[0077] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the third jump connection unit. Then, the output result of the third jump connection unit is concatted with the output of the second image block expansion unit and input into the seventh Swin Transformer unit.

[0078] Further, the jump connection unit includes: a first jump connection unit, a second jump connection unit, and a third jump connection unit, and the first jump connection unit, the second jump connection unit and the third jump connection unit have the same structure;

[0079] Further, the second encoder unit is connected to the second decoder unit via a skip connection unit and comprises:

[0080] The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concatted and input into the fourth jump connection unit, and then the output result of the fourth jump connection unit is concatted with the output of the sixth image block expansion unit and input into the fourteenth Swin Transformer unit;

[0081] The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concat-ed and input to the fifth jump connection unit. Then, the output result of the fifth jump connection unit is concat-ed with the output of the fifth image block expansion unit and input to the thirteenth Swin Transformer unit.

[0082] The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concatted and input into the fifth jump connection unit, and then the output result of the sixth jump connection unit is concatted with the output of the fourth image block expansion unit and input into the twelfth Swin Transformer unit;

[0083] Further, the first decoder unit is connected to the first encoder unit via a skip connection unit and comprises:

[0084] The outputs of the fifth Swin Transformer unit, the sixth Swin Transformer unit, and the seventh Swin Transformer unit are concatted and input into the seventh jump connection unit, and then the output result of the seventh jump connection unit is concatted with the output of the fourth image block merging unit and input into the ninth Swin Transformer unit;

[0085] The outputs of the fifth Swin Transformer unit, the sixth Swin Transformer unit, and the seventh Swin Transformer unit are concatted and input into the eighth jump connection unit, and then the output result of the eighth jump connection unit is concatted with the output of the fifth image block merging unit and input into the tenth Swin Transformer unit;

[0086] And the structures of the first, second, third, fourth, fifth, sixth, seventh and eighth hop connection units are all the same.

[0087] Furthermore, the hybrid loss function used by the earthquake deep learning model is:

[0088]

[0089]

[0090] L=ω1L1+ω2L2, and:ω1+ω2=1

[0091] Among them, y i =sigmoid(xi ), xi is the input of the deep learning network, y i ′ is the pixel label, α is the ratio of the number of non-fault pixels to the total number of pixels; N is the total number of pixels in the seismic image, i is the pixel number, ω1 and ω2 are weight coefficients respectively.

[0092] Furthermore, the weight coefficients ω1 and ω2 in the hybrid loss function are dynamically acquired over time, and the calculation method is: in:

[0093]

[0094] L t =ω 1,t-1 L 1,t +ω 2,t-1 L 2,t

[0095] Among them, L j,t is the jth loss function at time t, u(L j,t-1 ) represents the average value of the j-th loss before time t-1,

[0096] σ(r i,t ) and μ(r i,t ) are respectively r before time t j,t The standard deviation and mean of , and the value of j is 1 or 2.

[0097] Secondly, a five-in-one oil and uranium prospecting device is provided, comprising:

[0098] Data screening unit: Carry out data screening, where the data screening unit includes:

[0099] The first screening module: collect, investigate and organize the drilling information of the oil field;

[0100] The second screening module: analyzing the drilling radioactivity anomaly standards and building radioactivity anomaly screening software;

[0101] The third screening module: based on the radioactive anomaly screening software, the radioactive anomaly screening module determines the borehole radioactive anomaly standard and screens out radioactive anomaly wells;

[0102] Old well retest unit: Carry out old well retest, among which, the old well retest unit includes:

[0103] The first retest module: establish the casing-cement ring model;

[0104] The second retest module: Carry out cement ring casing logging response test;

[0105] The third retest module: analyzes the functional relationship between the counting rate and the cement sheath property parameters to determine the correction coefficient;

[0106] The fourth retest module: obtaining accurate shallow stratum radioactivity information based on the correction coefficient;

[0107] Gamma conversion unit: Carry out drilling construction verification to realize the conversion of natural gamma to quantitative gamma. The gamma conversion unit includes:

[0108] The first gamma module: collects gamma data of verification holes and oil field drilling holes;

[0109] The second gamma module: corrects the gamma logging depth of the verification hole to the gamma anomaly depth range of the oil field borehole;

[0110] The third gamma module: establish the regression equation between natural gamma and quantitative gamma;

[0111] The fourth gamma module: use regression equation to convert natural gamma to quantitative gamma and evaluate mineralization;

[0112] Seismic processing and interpretation unit: Carry out 3D seismic data processing, interpretation and analysis of faults, sand bodies and structures in the mining area;

[0113] Deposit comparison unit: Carry out typical deposit comparison, where the deposit comparison unit includes:

[0114] The first comparison module: collect typical uranium deposit data;

[0115] The second comparison module: analysis of structural sedimentary evolution characteristics, uranium source conditions, sand bodies and sedimentary systems, ore body characteristics, and mineralization laws;

[0116] The third comparison module: summarize the ore-controlling factors of typical sandstone-type uranium deposits;

[0117] Comprehensive processing unit: based on the results of the data screening unit, the old well retesting unit, the gamma conversion unit, the seismic processing and interpretation unit, and the ore deposit comparison unit, the favorable uranium ore zones are determined and the mineralization potential of the working area is evaluated.

[0118] In a third aspect, an electronic device is also provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the oil and uranium prospecting method based on the five-in-one method are implemented.

[0119] In a fourth aspect, a computer storage medium is provided on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the oil and uranium prospecting method based on the five-in-one method are implemented.

[0120] Compared with the prior art, the present invention has the following beneficial effects:

[0121] 1. The present invention realizes secondary development and utilization of a large amount of oil logging data in oil fields through comprehensive radioactive anomaly screening through the five-in-one oil and uranium prospecting technology, screens out reliable radioactive anomalies, and determines the uranium prospecting target area; when the natural gamma data in the shallow part of the oil field is missing and cannot be screened, the old well re-survey technology is adopted to supplement it; through the natural gamma and quantitative gamma conversion technology, the correlation between the γ anomaly of the verification hole of the uranium mineralization section and the GR anomaly of the oil field hole is established to assist in estimating the uranium resource volume; through the three-dimensional seismic data application technology, according to the geological needs of uranium exploration, the application countermeasures and methods of the three-dimensional seismic data fault interpretation technology of sandstone-type uranium mines are established; through the application research of typical ore deposits, the favorable conditions for typical sandstone-type uranium mineralization in the main uranium-producing basins in northern my country are analyzed, and a sandstone-type uranium mine exploration and evaluation technology system for the working area of ​​the basin oil field is established to accurately evaluate and determine the depression, determine the favorable zone, and guide the exploration deployment.

[0122] 2. Based on the actual uranium mine seismic exploration, the present invention has developed a new deep learning neural network model for the three-dimensional seismic data fault interpretation technology. This model combines the upper and lower information together to improve the resolution ability. It can not only predict the main faults, but also better avoid most non-fault discontinuities, thereby improving the resolution ability of sandstone-type uranium mine faults.

[0123] 3. The present invention develops a new hybrid loss function in combination with the deep learning neural network model for uranium mine earthquake fault identification. The hybrid loss function can accelerate the model convergence speed, improve the model convergence adaptability, and improve the accuracy of fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0124] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0125] Figure 1 A flowchart for realizing the five-in-one oil and uranium prospecting method according to an embodiment of the present invention;

[0126] Figure 2 A stratigraphic structural characteristic map of a sandstone-type uranium deposit in depth domain provided according to an embodiment of the present invention;

[0127] Figure 3 A seismic profile of a sandstone-type uranium mine provided according to an embodiment of the present invention;

[0128] Figure 4 A deep learning neural network model structure (single encoder-decoder) for fault / fracture prediction of sandstone-type uranium deposits based on seismic data according to an embodiment of the present invention;

[0129] Figure 5 A skip connection unit structure in a deep learning neural network model for fault / fracture prediction of sandstone-type uranium deposits based on seismic data according to an embodiment of the present invention;

[0130] Figure 6 A deep learning neural network model structure (two encoders-decoders) for fault / fracture prediction of sandstone-type uranium deposits based on seismic data according to an embodiment of the present invention;

[0131] Figure 7 The Swin transformer module structure in the deep learning neural network model for fault / fracture prediction of sandstone-type uranium deposits based on seismic data provided by an embodiment of the present invention;

[0132] Figure 8 2D prediction results (vertical profile) of a deep learning neural network model for fault / fracture prediction based on seismic data for sandstone-type uranium deposits provided in an embodiment of the present invention;

[0133] Fig. 9 2D prediction results (lateral profile) of a deep learning neural network model for fault / fracture prediction based on seismic data in sandstone-type uranium deposits according to an embodiment of the present invention;

[0134] Fig.10 3D prediction results of a deep learning neural network model for fault / fracture prediction based on seismic data for sandstone-type uranium deposits according to an embodiment of the present invention;

[0135] Fig.11 A graph showing the variation of the weight coefficient of the loss function of the deep learning neural network model for fault / fracture prediction of sandstone-type uranium deposits based on seismic data over time according to an embodiment of the present invention;

[0136] Fig.12 A diagram of a five-in-one oil and uranium prospecting device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0137] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0138] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0139] Figure 1 The figure shows a flowchart of a five-in-one oil and uranium prospecting method according to an embodiment of the present invention. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0140] A five-in-one oil and uranium prospecting method, including:

[0141] Step S1: Conducting data screening, wherein step S1 includes:

[0142] Step S11: Collect, investigate and organize drilling information of the oil field;

[0143] Step S12: analyzing the borehole radioactivity anomaly standard and constructing radioactivity anomaly screening software;

[0144] Step S13: Based on the radioactivity anomaly screening software, the radioactivity anomaly standard of the borehole is used for identification, and radioactivity anomaly wells are screened out;

[0145] Step S2: retest the old well, wherein step S2 includes:

[0146] Step S21: establishing a casing-cement ring model;

[0147] Step S22: carrying out a cement sheath casing logging response test;

[0148] Step S23: analyzing the functional relationship between the counting rate and the cement sheath property parameters to determine the correction coefficient;

[0149] Step S24: obtaining accurate shallow stratum radioactivity information based on the correction coefficient;

[0150] Step S3: Conduct drilling construction verification to achieve the conversion of natural gamma into quantitative gamma, wherein step S3 includes:

[0151] Step S31: Collecting gamma data of verification holes and oilfield drilling holes;

[0152] Step S32: Correcting the verification hole gamma logging depth to the oilfield drilling hole gamma abnormal depth range;

[0153] Step S33: establishing a regression equation between natural gamma and quantitative gamma;

[0154] Step S34: using the regression equation to convert natural gamma to quantitative gamma and evaluate mineralization;

[0155] Step S4: Processing and interpreting 3D seismic data to analyze faults, sand bodies and structures in the mining area;

[0156] Step S5: Conduct typical mineral deposit comparison, wherein step S5 includes:

[0157] Step S51: Collect typical uranium deposit data;

[0158] Step S52: analyzing structural sedimentary evolution characteristics, uranium source conditions, sand bodies and sedimentary systems, ore body characteristics, and mineralization laws;

[0159] Step S53: Summarize the ore-controlling factors of typical sandstone-type uranium deposits;

[0160] Step S6: Based on the results of the above steps S3 to S5, determine the favorable uranium ore zones and evaluate the mineralization potential of the working area.

[0161] Furthermore, the step S4 further comprises:

[0162] S41: Determine the structural characteristics of the uranium mining area using seismic fine structural interpretation technology and structural and sedimentary evolution analysis technology;

[0163] S42: Use the seismic stage-grading fault fine characterization technology to determine the uranium fault, oil and gas fault and oil and gas migration channel;

[0164] S43: Determine the top and bottom condition characteristics of uranium sand bodies using sand body top and bottom seismic identification technology;

[0165] S44: Use formation reservoir fine description technology to determine the distribution, thickness and physical properties of uranium sandstone.

[0166] In particular, in step S1, data screening technology is an effective means to obtain deep uranium mineralization information. On the basis of liberating productivity, it is possible to further study abnormal data, form a "technical system for screening and evaluating radioactive anomalies in oil and gas wells", quickly, accurately and efficiently screen and process logging data, quickly obtain abnormal information, lock in favorable exploration areas, and conduct uranium exploration by re-drilling and verification of the optimal radioactive survey area, so as to submit a uranium mine development base and finally develop sandstone uranium mines.

[0167] The abnormal screening of oil wells with natural gamma logging curves in shallow layers in each depression in the oil field is based on the data screening practice standard values ​​in the table below, and is based on the radioactive anomaly screening software. Among them, the most favorable screening abnormal section is more than 500m deep. On this basis, wells with anomalies greater than 1000API and thickness generally greater than 1m are selected as ultra-high abnormal wells. According to experience, the ultra-high abnormal wells screened can generally reach the level of industrial uranium mine holes.

[0168] Table 1 Criteria for data screening

[0169]

[0170] In particular, in step S2, when retesting old wells, based on the Compton effect of radioactive logging, multiple cement sheath models with different thicknesses and densities are established to simulate downhole conditions, and the cement sheath absorption coefficient values ​​of different models are used for fitting. The functional relationship between the radioactive counting rate and the cement sheath density and the cement sheath thickness is obtained through simulation calculation. Analysis shows that the radioactive counting rate and the cement sheath density and thickness are linearly related within a certain range, and the comprehensive functional relationship between the counting rate and the cement sheath attribute parameters is derived. Finally, the influence coefficient of the cement sheath on the gamma logging count rate is found, the specific correction coefficient is obtained, and a correction demonstration experiment is carried out to make the corrected data infinitely close to the true value of the formation, and the old oil wells that have been measured in the past and will be measured are reinterpreted and corrected to restore the real formation radioactive effect.

[0171] After testing, the comprehensive correction coefficient of cement sheath is expressed as:

[0172] Y=100*Y1 / (202.628-61.447*ρ(s)-0.627*h(s))

[0173] Where: ρ(s) represents cement sheath density, unit is g / cm3, h(s) represents cement sheath thickness, unit is mm; and it satisfies 6.15≤h(s)≤50.15; 1.63≤ρ(s)≤1.94;

[0174] In particular, in step S3, when converting natural gamma to quantitative gamma, the correlation between the natural gamma data (in API) of the Ordos Basin (Changqing exploration area) and the Erlian Basin (Huayou exploration area) and the nuclear industry radioactive quantitative gamma measurement data (in cps) is established, and the interpreted uranium grade (in %) and uranium content (ppm) of the oil field boreholes are further calculated, which is of great significance for guiding the deployment of the next drilling work and even estimating the prospective resource volume within the exploration right scope of the entire Ordos-Erlian Basin oil field.

[0175] The analysis shows that the mudstone uranium anomaly section is mainly controlled by the carbon content and the adsorption of clay minerals in the mudstone, and the influencing factors are relatively simple; while the sandstone uranium anomaly section is highly heterogeneous in the vertical direction, is subject to more geological interference factors, has a high degree of geological complexity, and has a complex correlation pattern. The present invention establishes the correlation between the GR logging data of the mudstone section of the oilfield borehole and the radioactive γ logging data of the verification hole. In the study, mudstone-type uranium mineralization and anomaly clues with a thickness of about 3.0m (395-398m) and stable distribution on the profile were found in the Saihan Formation of the block. The oilfield hole 7M72-F130 and the verification hole 10D72-1 anomaly section were selected as the research objects. The final conversion model used is: Y=6.147+0.218X1, X1 represents the GR value of the oilfield hole. The following table is the analysis result table of the inferred γ of the oilfield hole in the Erlian Basin and the measured γ of the verification hole:

[0176] Table 2 Analysis results of inferred γ and measured γ in verification wells of Erlian Basin oilfield

[0177] Abnormal depth m Oilfield Hole GR Oilfield hole estimation cps Verification hole measured cps 243.00 249.26 60.49 51.68 243.13 285.90 68.47 61.81 243.25 317.31 75.32 81.11 243.38 326.47 77.32 98.89 243.50 307.09 73.09 91.70 243.63 266.91 64.33 71.81 243.75 223.72 54.92 50.01 243.88 192.45 48.10 34.61 244.00 171.71 43.58 36.61

[0178] In particular, in step S4, the interpretation of seismic data of in-situ leaching sandstone uranium deposits can be based on the main work objectives of in-situ leaching sandstone uranium deposit exploration, namely, uranium source conditions, uranium migration and uranium precipitation geological environment issues, through three-dimensional seismic data interpretation work, analyze and study the spatial characteristics of regional structure, structural development history, lithology, sedimentary phases, etc., so as to summarize the mineralization law, establish a mineralization model, and guide the selection of peripheral areas. According to uranium mineralization elements and combined with geological needs, the main application directions of seismic technology are divided into six aspects: structural and sedimentary evolution analysis, fault characterization, structural description, top and bottom plate condition analysis, sand body distribution and physical property prediction, and favorable area evaluation.

[0179] Figure 2 The following is a characteristic diagram of the stratigraphic structure of a sandstone-type uranium ore in the depth domain according to an embodiment of the present invention. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0180] The present invention develops a new deep learning neural network model by combining the actual uranium seismic exploration to predict the main faults. According to the distribution morphology and spatial characteristics of the strata and faults explained in the time domain, combined with the time-depth relationship of the typical wells in the region, the time-depth conversion on the plane is performed to establish the stratigraphic structural characteristics in the depth domain. The bottom interface of the ore-bearing sand body in Zhenyuan area is high in the west and low in the east. There are two structural depressions in the north and south. The strata in the main uranium exploration area are relatively flat, with a maximum height difference of about 50m and a maximum fault throw of about 40m. The structural characteristics of the top interface of the Luohe Formation are similar to those of the bottom interface of the ore-bearing sand body, and have a certain degree of inheritance.

[0181] Figure 3The seismic profile of a sandstone-type uranium mine provided according to an embodiment of the present invention is shown. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0182] The PY-11 well area is located in a flat and low-lying area. The bottom boundary of the upper section of the Huanhe Formation is between 616 and 636 meters above sea level, and the formation angle is less than 1.5°; the bottom boundary of the lower section of the Huanhe Formation is between 440 and 460 meters above sea level, and the formation angle is less than 1°. The well area is less than 1 km away from the Tongyuan Fault. The PY-7 well area is located in the middle of two northeast strike-slip faults. The strata are flat and low-lying. The bottom boundary of the upper section of the Huanhe Formation is between 430 and 440 meters above sea level, less than 600 meters away from the southern fault, and less than 1 km away from the northern fault. Both faults are connected to the Chang 7 fault. The PY-11 well area is close to the PY-7 well area, about 2.5 km apart. According to the regional tectonic unit division, the area is located in the southwest of the Ordos Basin, and the tectonic unit belongs to the Tianhuan Sag. The overall structure of the work area is relatively flat and faults are well developed. Through the analysis of the currently completed industrial uranium mines, uranium is mostly enriched in the low-lying and gentle zones, and is close to the Tongyuan Fault. Except for the PY-6 well, the other well areas all meet the requirements of low-lying and gentle formations and less than 1km from the fault. At the same time, compared with several key wells, the PY-6 well is also an industrial well with a lower uranium content per square meter. Therefore, it is unanimously believed that low-lying areas with gentle structures and areas less than 1km from the Tongyuan Fault are conducive to uranium enrichment and mineralization.

[0183] In particular, in step S5, the Erlian Basin is an important uranium-producing basin in northern China in the comparison of typical mineral deposits. The main uranium deposit type in the basin is the ancient river valley sandstone type. The mineral deposits discovered so far include the Bayanwula large-scale, Hadatuo large-scale, and Saihang Gaobi medium-sized. The main uranium mineralization in the ancient river valley is mainly located in the upper part of the Saihan Formation, which is controlled by the braided river sand body in the upper part of the Saihan Formation in the ancient river valley. The main mineralization type of the Bayanwula deposit is the ancient river valley phreatic-interlayer oxidation zone type, the Saihang Gaobi is mainly the ancient river valley phreatic oxidation zone type, and the Hadatuo deposit is mainly the interlayer oxidation zone type.

[0184] Through the analysis of the mineralization conditions of typical sandstone-type uranium deposits in the Erlian Basin, various mineralization conditions are directly or indirectly reflected in the main controlling factors. The typical sandstone-type uranium mineralization and enrichment rules of the Erlian Basin are summarized, and it is found that the mineralization rules of the deposits have commonalities and characteristics. The commonalities are reflected in: paleostructure, paleogeomorphology, target layer, paleoclimate, and sedimentary system; the characteristics are reflected in: uranium source, sedimentary facies, hydrological conditions, and phreatic-interlayer oxidation.

[0185] It is believed that the main controlling factors of sandstone-type uranium mineralization in the Erlian Basin are: uranium source, large-scale sand body, and redox zone (compensation and drainage). Therefore, the analysis and research on the Erren Nur and Naomugen working areas focused on the uranium source location, sand body layer location, and redox location.

[0186] Figure 4 The structure of a deep learning neural network model for fault / fracture prediction based on seismic data of sandstone-type uranium deposits according to an embodiment of the present invention is shown. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows: The seismic deep learning model includes: an encoder-decoder module, and the encoder-decoder module includes:

[0187] Encoder unit, fourth Swin Transformer unit, decoder unit, skip connection unit;

[0188] The encoder unit includes: an image block segmentation unit, an image block embedding unit, a first Swin Transformer unit, a first image block merging unit, a second Swin Transformer unit, a second image block merging unit, a third Swin Transformer unit, and a third image block merging unit, and the image block segmentation unit outputs to the image block embedding unit, the image block embedding unit outputs to the first Swin Transformer unit, the first Swin Transformer unit outputs to the first image block merging unit, the first image block merging unit outputs to the second Swin Transformer unit, the second Swin Transformer unit outputs to the second image block merging unit, the second image block merging unit outputs to the third Swin Transformer unit, and the third Swin Transformer unit outputs to the third image block merging unit;

[0189] The third image block merging unit outputs to the fourth Swin Transformer unit, and the fourth Swin Transformer unit outputs to the first image block expansion unit;

[0190] The decoder unit includes: a first image block expansion unit, a fifth Swin Transformer unit, a second image block expansion unit, a sixth Swin Transformer unit, a third image block expansion unit, a seventh Swin Transformer unit, a final image expansion module, and a linear projection module, and the first image block expansion unit outputs to the fifth Swin Transformer unit, the fifth Swin Transformer unit outputs to the second image block expansion unit, the second image block expansion unit outputs to the sixth Swin Transformer unit, the sixth Swin Transformer unit outputs to the third image block expansion unit, the third image block expansion unit outputs to the seventh Swin Transformer unit, the seventh Swin Transformer unit outputs to the final image expansion module, and the final image expansion module outputs to the linear projection module;

[0191] The encoder unit is connected to the decoder unit through a skip connection unit.

[0192] In particular, the details of the deep learning neural network model are still Figure 5 The description of the mark in is not expanded in detail. The principle is explained as follows. Specifically, the input seismic image x is first divided into image block slices (Patch partition) and embedded with a fixed-length tag vector (Patch embedding). The tag vector is the basic information unit in the network architecture. For the two-dimensional and three-dimensional network architectures, we set the image block slice size to 4×4 and 4×4×4, respectively, and the length of the embedding vector C is 64. Without loss of generality, taking the two-dimensional seismic image as an example, the input dimension of the encoder branch is H / 4×W / 4×C. The encoder and decoder branches have a similar three-layer structure. A single level of the encoder consists of two consecutive Swin Transformer unit blocks to implement the self-attention mechanism, followed by an image block slice merging module (patch merging block) for downsampling the output. The image block merging module (patch merging block) connects the features of two adjacent patches, which expands the receptive field of the next (lower) level of the encoder, enabling the next level of the encoder to focus on larger-scale features. The input and output sizes of the i-th level of the encoder are W / (4×2 i-1 )×2 i-1 C and H / (4×2 i-1 )×W / (4×2 i )×2 1 C. Similarly, a single level of the decoder consists of two consecutive Swin-transformer blocks, followed by a patch-expanding block that upsamples the output to restore the patch resolution of the upper layer. The encoder and decoder also have two consecutive Swin transformer blocks and a patch expansion block. At the top level of the decoder branch, the last image patch is expanded (Final expanding), followed by a fully connected linear projection layer and a sigmoid activation layer, outputting an image mask showing the pixel-level fault probability. The output size (H×W) is the same as the input seismic image.

[0193] Furthermore, in the neural network deep learning model of the present invention, for each level of decoder, the input is the combination of the output of the next level decoder and the output of the skip connection unit. The present invention does not use simple skip connections to transmit fault information between the encoder and decoder at the same level. The present invention innovatively provides the decoder with multi-dimensional information from each level of the encoder to enhance the neural network. This architecture enables the network to learn fault features from different dimensions at the same time.

[0194] Figure 5 The figure shows the skip connection unit structure in the deep learning neural network model for fault / fracture prediction based on seismic data of sandstone-type uranium deposits according to an embodiment of the present invention. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0195] Further, the encoder unit is connected to the decoder unit via a jump connection unit and comprises:

[0196] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the first jump connection unit, and then the output result of the first jump connection unit is concatted with the output of the first image block expansion unit and input into the fifth Swin Transformer unit;

[0197] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the second jump connection unit, and then the output result of the second jump connection unit is concatted with the output of the second image block expansion unit and input into the sixth Swin Transformer unit;

[0198] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the third jump connection unit. Then, the output result of the third jump connection unit is concatted with the output of the second image block expansion unit and input into the seventh Swin Transformer unit.

[0199] Further, the jump connection unit includes: a first jump connection unit, a second jump connection unit, and a third jump connection unit, and the first jump connection unit, the second jump connection unit and the third jump connection unit have the same structure;

[0200] The first jump connection unit includes: the eighth Swin Transformer unit and the channel mixing module;

[0201] The channel mixing module includes two branches, the first branch includes a maximum pooling layer and a 1×1 convolution layer, and the second branch includes an average pooling layer and a 1×1 convolution layer; the output result of the first branch and the output result of the second branch are added, and then passed to the 1×1 convolution layer and the sigmoid activation layer to obtain the output result of the channel mixing module;

[0202] The output result of the eighth Swin Transformer unit and the output result of the channel mixing module are multiplied to obtain the output result of the first jump connection unit.

[0203] The principle is explained as follows. Specifically, the output of all levels of the encoder is first downsampled, upsampled, or directly passed to the Concat module before entering the i-th level of the skip connection unit module. The copy is then divided into two inputs, one feeding two Swin transformer modules for calculating spatial self-attention and one feeding a channel mixing module for calculating channel self-attention. The Swin transformer block output is then projected to the same size H / (4×2 i-1 )×W / (4×2 i )×2 1 C is output. The channel mixing block has two branches with maximum pooling and average pooling to reduce spatial information, and then a 1×1 convolution layer is applied to perform channel self-attention. The outputs of the two branches are added together and then passed through another 1×1 convolution layer and a sigmoid activation layer to obtain the channel attention result. Finally, the output of the skip connection unit is the multiplication of the channel attention and spatial attention.

[0204] Figure 6 Another structure of a deep learning neural network model for fault / fracture prediction based on seismic data of sandstone-type uranium deposits according to an embodiment of the present invention is shown. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0205] The network architecture includes Figure 6 The two encoder-decoder modules are shown.

[0206] The result of the last Swin Transformer unit in the decoder unit in the first encoder-decoder module is output to the first Swin Transformer unit in the encoder unit in the second encoder-decoder module, and a jump connection unit is used to establish a jump connection between the Swin Transformer unit in the decoder unit in the first encoder-decoder module and the Swin Transformer unit in the encoder unit in the second encoder-decoder module. The specific structure is as follows:

[0207] The earthquake deep learning model includes:

[0208] A first encoder-decoder module, a second encoder-decoder module;

[0209] The first encoder-decoder module comprises:

[0210] First encoder unit, fourth Swin Transformer unit, first decoder unit, skip connection unit;

[0211] The first encoder unit includes: an image block segmentation unit, an image block embedding unit, a first Swin Transformer unit, a first image block merging unit, a second Swin Transformer unit, a second image block merging unit, a third Swin Transformer unit, and a third image block merging unit, and the image block segmentation unit outputs to the image block embedding unit, the image block embedding unit outputs to the first Swin Transformer unit, the first Swin Transformer unit outputs to the first image block merging unit, the first image block merging unit outputs to the second Swin Transformer unit, and the second Swin

[0212] The Transformer unit outputs to the second image block merging unit, and the second image block merging unit outputs to the third Swin

[0213] The Transformer unit and the third Swin Transformer unit output to the third image block merging unit;

[0214] The third image block merging unit outputs to the fourth Swin Transformer unit, and the fourth Swin Transformer unit outputs to the first image block expansion unit;

[0215] The first decoder unit includes: a first image block expansion unit, a fifth Swin Transformer unit, a second image block expansion unit, a sixth Swin Transformer unit, a third image block expansion unit, a seventh Swin Transformer unit, a final image expansion module, and a linear projection module, and the first image block expansion unit is output to the fifth Swin Transformer unit, the fifth Swin Transformer unit is output to the second image block expansion unit, the second image block expansion unit is output to the sixth Swin Transformer unit, the sixth Swin Transformer unit is output to the third image block expansion unit, the third image block expansion unit is output to the seventh Swin Transformer unit, the seventh Swin Transformer unit is output to the final image expansion module, and the final image expansion module is output to the linear projection module;

[0216] And the seventh Swin Transformer unit outputs to the eighth Swin Transformer unit;

[0217] The second encoder-decoder module comprises:

[0218] The second encoder unit, the eleventh Swin Transformer unit, the second decoder unit, and the skip connection unit;

[0219] The second encoder unit includes: an eighth Swin Transformer unit, a fourth image block merging unit, a ninth Swin Transformer unit, a fifth image block merging unit, a tenth Swin Transformer unit, and a sixth image block merging unit, and the eighth Swin Transformer unit outputs to the fourth image block merging unit, the fourth image block merging unit outputs to the ninth Swin Transformer unit, the ninth Swin Transformer unit outputs to the fifth image block merging unit, the fifth image block merging unit outputs to the tenth Swin Transformer unit, and the tenth Swin Transformer unit outputs to the sixth image block merging unit;

[0220] And the sixth image block merging unit outputs to the eleventh Swin Transformer unit, and the eleventh Swin Transformer unit outputs to the fourth image block expanding unit;

[0221] The second decoder unit includes: a fourth image block expansion unit, a twelfth Swin Transformer unit, a fifth image block expansion unit, a thirteenth Swin Transformer unit, a sixth image block expansion unit, a fourteenth Swin Transformer unit, a final image expansion module, and a linear projection module, and the fourth image block expansion unit outputs to the twelfth Swin Transformer unit, the twelfth Swin Transformer unit outputs to the fifth image block expansion unit, the fifth image block expansion unit outputs to the thirteenth Swin Transformer unit, the thirteenth Swin Transformer unit outputs to the sixth image block expansion unit, the sixth image block expansion unit outputs to the fourteenth Swin Transformer unit, the fourteenth Swin Transformer unit outputs to the final image expansion module, and the final image expansion module outputs to the linear projection module;

[0222] The first encoder unit is connected to the first decoder unit via a skip connection unit;

[0223] The second encoder unit is connected to the second decoder unit via a skip connection unit;

[0224] The first decoder unit is connected to the first encoder unit via a skip connection unit;

[0225] Further, the first encoder unit is connected to the first decoder unit via a skip connection unit and comprises:

[0226] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the first jump connection unit, and then the output result of the first jump connection unit is concatted with the output of the first image block expansion unit and input into the fifth Swin Transformer unit;

[0227] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the second jump connection unit, and then the output result of the second jump connection unit is concatted with the output of the second image block expansion unit and input into the sixth Swin Transformer unit;

[0228] The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the third jump connection unit. Then, the output result of the third jump connection unit is concatted with the output of the second image block expansion unit and input into the seventh Swin Transformer unit.

[0229] Further, the jump connection unit includes: a first jump connection unit, a second jump connection unit, and a third jump connection unit, and the first jump connection unit, the second jump connection unit and the third jump connection unit have the same structure;

[0230] Further, the second encoder unit is connected to the second decoder unit via a skip connection unit and comprises:

[0231] The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concatted and input into the fourth jump connection unit, and then the output result of the fourth jump connection unit is concatted with the output of the sixth image block expansion unit and input into the fourteenth Swin Transformer unit;

[0232] The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concat-ed and input to the fifth jump connection unit. Then, the output result of the fifth jump connection unit is concat-ed with the output of the fifth image block expansion unit and input to the thirteenth Swin Transformer unit.

[0233] The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concatted and input into the fifth jump connection unit, and then the output result of the sixth jump connection unit is concatted with the output of the fourth image block expansion unit and input into the twelfth Swin Transformer unit;

[0234] Further, the first decoder unit is connected to the first encoder unit via a skip connection unit and comprises:

[0235] The outputs of the fifth Swin Transformer unit, the sixth Swin Transformer unit, and the seventh Swin Transformer unit are concatted and input into the seventh jump connection unit, and then the output result of the seventh jump connection unit is concatted with the output of the fourth image block merging unit and input into the ninth Swin Transformer unit;

[0236] The outputs of the fifth Swin Transformer unit, the sixth Swin Transformer unit, and the seventh Swin Transformer unit are concatted and input into the eighth jump connection unit, and then the output result of the eighth jump connection unit is concatted with the output of the fifth image block merging unit and input into the tenth Swin Transformer unit;

[0237] In particular, the first, second, third, fourth, fifth, sixth, seventh and eighth hop connection units have the same structure.

[0238] In particular, Figure 6 The model principle of each encoder-decoder unit in the two encoder-decoder units shown is the same as Figure 4 The principles of the single encoder-decoder unit model shown are the same and will not be repeated here.

[0239] Experiments show that by setting up these two encoder-decoder modules, the features of the fault can be better extracted for learning and training, thereby achieving high-precision fault feature extraction.

[0240] Figure 7 The structure of the Swin transformer module in the deep learning neural network model for fault / fracture prediction based on seismic data of sandstone-type uranium deposits provided in accordance with an embodiment of the present invention is shown. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0241] Figure 7 The codes LN, W-MSA, SW-MSA, MLP-GELU, etc. are explained as follows:

[0242] Layer Normalization (LN): In Swin Transformer, LayerNorm is used to normalize the input of the layer to stabilize and accelerate the training process. It is applied before each Multi-Head Self-Attention (MSA) module and each MLP, which helps improve the training efficiency and performance of the model.

[0243] Window-based Multi-Head Self-Attention (W-MSA): W-MSA is one of the core components of Swin Transformer, which divides the image into non-overlapping windows and applies the self-attention mechanism independently within each window. This method reduces the amount of computation because the self-attention is only performed within the local window instead of on the entire image. This design allows the model to capture local features while maintaining computational efficiency.

[0244] Shifted Window Multi-Head Self-Attention (SW-MSA): To overcome the problem of information isolation between windows in W-MSA, Swin Transformer introduces SW-MSA. By sliding between adjacent windows, SW-MSA allows information to be transferred between windows, thereby enhancing the model's ability to capture global information. W-MSA and SW-MSA are used alternately in Swin Transformer to balance the processing of local and global information.

[0245] Multi-Layer Perceptron (MLP-GELU): In each Transformer block of Swin Transformer, MLP is used to further process the features after the self-attention mechanism. MLP usually contains two fully connected layers, with GELU (Gaussian Error Linear Unit) as the activation function in the middle. MLP is applied after the self-attention layer to increase the nonlinear expression ability of the model. The GELU activation function is widely used due to its excellent performance in deep learning, which helps the model better learn complex feature representations.

[0246] Here, the MLP layer has two fully connected linear layers with Gaussian error linear unit activation function. The skip connection that adds the input of the Swin transformer module and the output of the W-MSA / SW-MSA is intended to pass the information to the second LN and MLP. Another skip connection adds the input of the second LN and the output of the MLP and passes the information to the next Swin transformer module.

[0247] The key to the Swin transformer block is W-MSA, SW-MSA. The input of W-MSA, SW-MSA is a large token matrix. In order to reduce the computational cost, the token matrix is ​​divided into 8×8 windows according to the actual position in the original image. Multi-head self-attention is calculated in parallel in each window. The self-attention mechanism is:

[0248]

[0249] where Q, K, and V are the query matrix, key matrix, and value matrix respectively, computed by three linear projections of the label matrix. The linear projection layer is a matrix with different weights W. Q , W K and W V A fully connected layer. k is the number of features used for normalization. The self-attention mechanism computes the coherence of each feature between different embedded tag vectors; therefore, it can naturally measure the global coherence between tag vectors. In order to incorporate the coherence between image patches in different windows, a shift window mechanism is implemented. Here, the windows used to compute multi-head self-attention are cyclically shifted four levels downward and rightward. If the window moves out of the boundary, it will fetch the missing patches from the top and left. This cyclic shift mechanism ensures that the self-attention mechanism is not limited to the local window and maintains the long-range and global connections between image patches.

[0250] Figure 8 The 2D prediction results (vertical profile) of the deep learning neural network model for fault / fracture prediction based on seismic data of sandstone-type uranium deposits provided in accordance with an embodiment of the present invention are shown. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0251] Figure 8 (a) is the fault prediction result of a single encoder-decoder. Figure 8 (b) The fault prediction results of the two encoder-decoders. It can be seen from the figure that both the single encoder-decoder and the two encoder-decoders achieved good fault recognition results.

[0252] Fig. 9 The 2D prediction results (cross-section) of the deep learning neural network model for fault / fracture prediction based on seismic data of sandstone-type uranium deposits provided in accordance with an embodiment of the present invention are shown. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0253] Fig. 9 (a) is the fault prediction result of a single encoder-decoder. Fig. 9 (b) The fault prediction results of the two encoder-decoders. It can be seen from the figure that both the single encoder-decoder and the two encoder-decoders achieved good fault recognition results.

[0254] Fig.10 The 3D prediction results of the deep learning neural network model for fault / fracture prediction based on seismic data of sandstone-type uranium deposits provided in accordance with an embodiment of the present invention are shown. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0255] Fig.10(a) is the fault prediction result of a single encoder-decoder. Fig.10 (b) The fault prediction results of the two encoder-decoders. It can be seen from the figure that both the single encoder-decoder and the two encoder-decoders achieved good fault recognition results.

[0256] The loss function used in the present invention is introduced below:

[0257] In deep learning, the loss function is a crucial component that measures the degree of difference between the model's predictions and the actual results. That is, the performance of the model is evaluated by calculating the degree of inconsistency between the predicted values ​​and the actual values. During the training process, the goal of the model is to minimize the value of the loss function by adjusting its parameters. The model can update each parameter through the back-propagation algorithm to reduce the loss between the actual value and the predicted value. In this way, the predicted values ​​generated by the model will gradually move closer to the actual value, thereby achieving the purpose of learning.

[0258] Loss functions are mainly divided into several types according to the task type, including:

[0259] Loss function based on distance metric: This type of loss function usually maps the input data to a feature space based on distance metric (such as Euclidean space, Hamming space, etc.), and then uses a suitable loss function to measure the distance between the true value of the sample and the model prediction value in the feature space. The smaller the distance, the better the prediction performance of the model. Common loss functions based on distance metric include MSE, LAE loss, LSE loss, etc.

[0260] MSE: Calculates the average of the sum of squares of the differences between the predicted value and the true value. MSE gives a larger penalty for larger errors and is more sensitive to outliers.

[0261] LAE loss: Also called minimizing the absolute error (LAE), it calculates the sum of the absolute values ​​of the differences between the predicted value and the true value. LAE loss is relatively insensitive to outliers, and its gradient is discontinuous at zero, which helps to produce sparse solutions (ie, the weights of some features are zero).

[0262] LSE loss: Also known as the least square error (LSE), it is the minimization of the sum of the squares of the differences between the target value and the estimated value. LSE loss is more sensitive to outliers, and its gradient is continuous at zero.

[0263] Loss function based on probability distribution measurement: This type of loss function converts the similarity between samples into the probability of a random event, and determines the similarity between the two by measuring the distance between the true distribution of the sample and its estimated distribution. It is particularly commonly used in classification problems. Common loss functions based on probability distribution measurement include cross entropy loss, KL divergence function, etc.

[0264] Cross entropy loss: It is used to measure the difference between the predicted probability distribution and the true label probability distribution. In the binary classification problem, the cross entropy loss can be expressed as the sum of the negative logarithms of the product of the probability of the predicted positive class and the true label (the same treatment is applied to the negative class). In the multi-classification problem, the cross entropy loss compares the predicted probability of each category with the probability distribution of the true label one by one and sums them up. The smaller the value of the cross entropy loss, the better the model performance.

[0265] KL divergence function: also known as relative entropy or information divergence, it is used to measure the difference between two probability distributions. KL divergence is non-negative and takes the value of zero if and only if the two distributions are exactly the same. In deep learning, KL divergence is often used in regularization terms to encourage the probability distribution of the model output to be close to the prior distribution.

[0266] Based on the need for a deep learning model for fault identification in seismic exploration of uranium sandstone-type uranium deposits, the present invention proposes a mixed loss function used in a seismic deep learning model:

[0267]

[0268] L=ω1L1+ω2L2, and:ω1+ω2=1

[0269] Among them, y i =sigmoid(x i ), xi is the input of the deep learning network, y i ′ is the pixel label, α is the ratio of the number of non-fault pixels to the total number of pixels; N is the total number of pixels in the seismic image, i is the pixel number, ω1 and ω2 are weight coefficients respectively.

[0270] In order to accurately measure the prediction performance of the network model, the embodiment of the present invention uses four accuracy assessment indicators, namely precision, recall, F1 score (F1_Score) and mean intersection over union (MloU), to determine the weights.

[0271] Precision is the ratio of positive samples predicted to be actually positive. It measures the accuracy of the positive samples predicted by the model. Calculation formula: Among them: TP (True Positive) is the true positive example, that is, the number of samples correctly predicted by the model as positive category. FP (False Positive) is the false positive example, that is, the number of samples incorrectly predicted by the model as positive category.

[0272] Recall is the ratio of positive samples to positive samples. It measures the ability of the model to capture all positive samples. Calculation formula: Among them, FN (False Negative) is the false negative example, that is, the number of positive samples that the model mistakenly predicts as negative categories.

[0273] The F1 score is the harmonic mean of precision and recall, which attempts to find a balance between precision and recall. Calculation formula: mIoU is the average of the intersection over union (IoU) of multiple categories and is used to measure the segmentation performance of the model, especially in multi-category segmentation tasks. Calculation formula: Where N is the number of categories and IoUi is the intersection over union of the i-th category.

[0274] Table 3 shows the comparison results of weight values ​​used in different hybrid loss functions:

[0275]

[0276] As can be seen from Table 3, the first group of weights has the worst effect, which obviously has a side effect and the network fitting effect is poor. When the weight ratio is 0.2 and 0.8, all evaluation indicators are the highest, that is, when the L1 loss function accounts for 0.2 and the L2 loss function accounts for 0.8, the network convergence is the best.

[0277] Furthermore, the present invention has found in the research that when the weight coefficients in the mixed loss function are fixed, when processing data with imbalanced categories, the use of fixed weight coefficients may cause the model to be biased towards the majority class, thereby ignoring the feature learning of the minority class, causing the model to overfit to certain categories during the training process, especially when the category distribution is uneven, the model may over-focus on the majority class and ignore the minority class. The model may perform well on the training set, but has poor generalization ability on a wider data set. Therefore, the present invention proposes that the weight coefficients ω1 and ω2 in the mixed loss function are variable, and the calculation method is:

[0278] in:

[0279]

[0280] L t =ω 1,t-1 L 1,t +ω 2,t-1 L 2,t

[0281] Among them, L j,t is the jth loss function at time t, u(L j,t-1 ) represents the average value of the j-th loss before time t-1,

[0282] σ(r i,t ) and μ(r i,t ) are respectively r before time t j,t The standard deviation and mean of , and the value of j is 1 or 2.

[0283] The weight coefficient can change according to the prediction difficulty of the data instance, allowing the model to adjust its focus on instances of different difficulty during training, thereby improving the generalization ability of the model, improving the calibration performance of the model, and making the probability of the model output more reliable.

[0284] Fig.11 The figure shows the weight coefficient of the loss function of the deep learning neural network model for fault / fracture prediction of sandstone-type uranium deposits based on seismic data according to an embodiment of the present invention. For ease of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0285] Fig.11 It shows that the weights of two loss functions (loss1 (L1) and loss2 (L2)) are adjusted as the number of training rounds (epochs) changes during the training of deep learning models. Initial stage: In the initial stage of training, the weight of loss1 drops rapidly, from close to 0.5 to about 0.4. At the same time, the weight of loss2 rises rapidly, from close to 0.5 to about 0.6. This shows that in the initial stage, the model prefers to optimize loss2 because loss2 is more helpful for model learning in the early stage. Mid-term stage: In this stage, the weight of loss1 continues to drop slowly, but the rate of decline slows down. The weight of loss2 reaches a plateau in the mid-term stage and then continues to rise slowly. This means that as training progresses, the contribution of loss1 to the model gradually decreases, while loss2 continues to maintain its importance. Late stage: In the later stage of training, the weight of loss1 continues to drop, but the change is small, and finally stabilizes at about 0.2. The weight of loss2 continues to rise in the later stage, and finally approaches 0.8, showing its importance in the later stage of model training. This trend may indicate that as the model approaches convergence, loss2 becomes more critical to improving model performance. Fig.11 It shows that the weights of the two components in the hybrid loss function change dynamically as the training process progresses to optimize the model performance. This adjustment strategy helps the model better balance the contributions of different loss functions during training, thereby improving the final model performance.

[0286] Fig.12 The figure shows a five-in-one oil and uranium prospecting device provided according to an embodiment of the present invention. For the convenience of description, only the parts related to the embodiment of the present invention are shown, which are described in detail as follows: The five-in-one oil and uranium prospecting device comprises:

[0287] Data screening unit: Carry out data screening, where the data screening unit includes:

[0288] The first screening module: collect, investigate and organize the drilling information of the oil field;

[0289] The second screening module: analyzing the drilling radioactivity anomaly standards and building radioactivity anomaly screening software;

[0290] The third screening module: based on the radioactive anomaly screening software, the radioactive anomaly screening module determines the borehole radioactive anomaly standard and screens out radioactive anomaly wells;

[0291] Old well retest unit: Carry out old well retest, among which, the old well retest unit includes:

[0292] The first retest module: establish the casing-cement ring model;

[0293] The second retest module: Carry out cement ring casing logging response test;

[0294] The third retest module: analyzes the functional relationship between the counting rate and the cement sheath property parameters to determine the correction coefficient;

[0295] The fourth retest module: obtaining accurate shallow stratum radioactivity information based on the correction coefficient;

[0296] Gamma conversion unit: Carry out drilling construction verification to realize the conversion of natural gamma to quantitative gamma. The gamma conversion unit includes:

[0297] The first gamma module: collects gamma data of verification holes and oil field drilling holes;

[0298] The second gamma module: corrects the gamma logging depth of the verification hole to the gamma anomaly depth range of the oil field borehole;

[0299] The third gamma module: establish the regression equation between natural gamma and quantitative gamma;

[0300] The fourth gamma module: use regression equation to convert natural gamma to quantitative gamma and evaluate mineralization;

[0301] Seismic processing and interpretation unit: Carry out 3D seismic data processing, interpretation and analysis of faults, sand bodies and structures in the mining area;

[0302] Deposit comparison unit: Carry out typical deposit comparison, where the deposit comparison unit includes:

[0303] The first comparison module: collect typical uranium deposit data;

[0304] The second comparison module: analysis of structural sedimentary evolution characteristics, uranium source conditions, sand bodies and sedimentary systems, ore body characteristics, and mineralization laws;

[0305] The third comparison module: summarize the ore-controlling factors of typical sandstone-type uranium deposits;

[0306] Comprehensive processing unit: based on the results of the data screening unit, the old well retesting unit, the gamma conversion unit, the seismic processing and interpretation unit, and the ore deposit comparison unit, the favorable uranium ore zones are determined and the mineralization potential of the working area is evaluated.

[0307] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the oil and uranium prospecting method based on the five-in-one method are implemented.

[0308] An embodiment of the present invention further provides a computer storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the oil and uranium prospecting method based on the five-in-one method are implemented.

[0309] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0310] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0311] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0312] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A five-in-one oil and uranium prospecting method, characterized in that: include: Step S1: Conducting data screening, wherein step S1 includes: Step S11: Collect, investigate and organize drilling information of the oil field; Step S12: analyzing the borehole radioactivity anomaly standard and constructing radioactivity anomaly screening software; Step S13: Based on the radioactivity anomaly screening software, the radioactivity anomaly standard of the borehole is used for identification, and radioactivity anomaly wells are screened out; Step S2: retest the old well, wherein step S2 includes: Step S21: establishing a casing-cement ring model; Step S22: carrying out a cement sheath casing logging response test; Step S23: analyzing the functional relationship between the counting rate and the cement sheath property parameters to determine the correction coefficient; Step S24: obtaining accurate shallow stratum radioactivity information based on the correction coefficient; Step S3: Conduct drilling construction verification to achieve the conversion of natural gamma into quantitative gamma, wherein step S3 includes: Step S31: Collecting gamma data of verification holes and oilfield drilling holes; Step S32: Correcting the verification hole gamma logging depth to the oilfield drilling hole gamma abnormal depth range; Step S33: establishing a regression equation between natural gamma and quantitative gamma; Step S34: using the regression equation to convert natural gamma to quantitative gamma and evaluate mineralization; Step S4: Processing and interpreting 3D seismic data to analyze faults, sand bodies and structures in the mining area; Step S5: Conduct typical mineral deposit comparison, wherein step S5 includes: Step S51: Collect typical uranium deposit data; Step S52: analyzing structural sedimentary evolution characteristics, uranium source conditions, sand bodies and sedimentary systems, ore body characteristics, and mineralization laws; Step S53: Summarize the ore-controlling factors of typical sandstone-type uranium deposits; Step S6: Based on the results of the above steps S1 to S5, determine the favorable uranium ore zones and evaluate the mineralization potential of the working area.

2. The oil and uranium prospecting method based on the five-in-one method according to claim 1 is characterized in that: The step S4 further comprises: S41: Determine the structural characteristics of the uranium mining area using seismic fine structural interpretation technology and structural and sedimentary evolution analysis technology; S42: Use the seismic stage-grading fault fine characterization technology to determine the uranium fault, oil and gas fault and oil and gas migration channel; S43: Determine the top and bottom condition characteristics of uranium sand bodies using sand body top and bottom seismic identification technology; S44: Use formation reservoir fine description technology to determine the distribution, thickness and physical properties of uranium sandstone.

3. The oil and uranium prospecting method based on the five-in-one method according to claim 2 is characterized in that: The step S42 of using the seismic stage-by-stage classification fault fine characterization technology to determine the uranium fault, oil and gas fault, and oil and gas migration channel includes: A seismic deep learning model is used to finely characterize uranium mine faults. The seismic deep learning model includes an encoder-decoder module, and the encoder-decoder module includes: Encoder unit, fourth Swin Transformer unit, decoder unit, skip connection unit; The encoder unit includes: an image block segmentation unit, an image block embedding unit, a first Swin Transformer unit, a first image block merging unit, a second Swin Transformer unit, a second image block merging unit, a third Swin Transformer unit, and a third image block merging unit, and the image block segmentation unit outputs to the image block embedding unit, the image block embedding unit outputs to the first Swin Transformer unit, the first Swin Transformer unit outputs to the first image block merging unit, the first image block merging unit outputs to the second Swin Transformer unit, the second Swin Transformer unit outputs to the second image block merging unit, the second image block merging unit outputs to the third Swin Transformer unit, and the third Swin Transformer unit outputs to the third image block merging unit; The third image block merging unit outputs to the fourth Swin Transformer unit, and the fourth Swin Transformer unit outputs to the first image block expansion unit; The decoder unit includes: a first image block expansion unit, a fifth Swin Transformer unit, a second image block expansion unit, a sixth Swin Transformer unit, a third image block expansion unit, a seventh Swin Transformer unit, a final image expansion module, and a linear projection module, and the first image block expansion unit outputs to the fifth Swin Transformer unit, the fifth Swin Transformer unit outputs to the second image block expansion unit, the second image block expansion unit outputs to the sixth Swin Transformer unit, the sixth Swin Transformer unit outputs to the third image block expansion unit, the third image block expansion unit outputs to the seventh Swin Transformer unit, the seventh Swin Transformer unit outputs to the final image expansion module, and the final image expansion module outputs to the linear projection module; The encoder unit is connected to the decoder unit through a skip connection unit.

4. The oil and uranium prospecting method based on the five-in-one method according to claim 3 is characterized in that: The encoder unit is connected to the decoder unit via a skip connection unit and comprises: The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the first jump connection unit, and then the output result of the first jump connection unit is concatted with the output of the first image block expansion unit and input into the fifth Swin Transformer unit; The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the second jump connection unit, and then the output result of the second jump connection unit is concatted with the output of the second image block expansion unit and input into the sixth Swin Transformer unit; The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concat-ed and input into the third jump connection unit. Then, the output result of the third jump connection unit is concat-ed with the output of the second image block expansion unit and input into the seventh Swin Transformer unit.

5. The oil and uranium prospecting method based on the five-in-one method according to claim 4 is characterized in that: The jump connection unit includes: a first jump connection unit, a second jump connection unit, and a third jump connection unit, and the first jump connection unit, the second jump connection unit and the third jump connection unit have the same structure; The first jump connection unit includes: the eighth Swin Transformer unit and the channel mixing module; The channel mixing module includes two branches, the first branch includes a maximum pooling layer and a 1×1 convolution layer, and the second branch includes an average pooling layer and a 1×1 convolution layer; the output result of the first branch and the output result of the second branch are added, and then passed to the 1×1 convolution layer and the sigmoid activation layer to obtain the output result of the channel mixing module; The output result of the eighth Swin Transformer unit and the output result of the channel mixing module are multiplied to obtain the output result of the first jump connection unit.

6. The oil and uranium prospecting method based on the five-in-one method according to claim 2 is characterized in that: The earthquake deep learning model includes: A first encoder-decoder module, a second encoder-decoder module; The first encoder-decoder module comprises: First encoder unit, fourth Swin Transformer unit, first decoder unit, skip connection unit; The first encoder unit includes: an image block segmentation unit, an image block embedding unit, a first Swin Transformer unit, a first image block merging unit, a second Swin Transformer unit, a second image block merging unit, a third Swin Transformer unit, and a third image block merging unit, and the image block segmentation unit outputs to the image block embedding unit, the image block embedding unit outputs to the first Swin Transformer unit, the first Swin Transformer unit outputs to the first image block merging unit, the first image block merging unit outputs to the second Swin Transformer unit, the second Swin Transformer unit outputs to the second image block merging unit, the second image block merging unit outputs to the third Swin Transformer unit, and the third Swin Transformer unit outputs to the third image block merging unit; The third image block merging unit outputs to the fourth Swin Transformer unit, and the fourth Swin Transformer unit outputs to the first image block expansion unit; The first decoder unit includes: a first image block expansion unit, a fifth Swin Transformer unit, a second image block expansion unit, a sixth Swin Transformer unit, a third image block expansion unit, a seventh Swin Transformer unit, a final image expansion module, and a linear projection module, and the first image block expansion unit is output to the fifth Swin Transformer unit, the fifth Swin Transformer unit is output to the second image block expansion unit, the second image block expansion unit is output to the sixth Swin Transformer unit, the sixth Swin Transformer unit is output to the third image block expansion unit, the third image block expansion unit is output to the seventh Swin Transformer unit, the seventh Swin Transformer unit is output to the final image expansion module, and the final image expansion module is output to the linear projection module; And the seventh Swin Transformer unit outputs to the eighth Swin Transformer unit; The second encoder-decoder module comprises: The second encoder unit, the eleventh Swin Transformer unit, the second decoder unit, and the skip connection unit; The second encoder unit includes: an eighth Swin Transformer unit, a fourth image block merging unit, a ninth Swin Transformer unit, a fifth image block merging unit, a tenth Swin Transformer unit, and a sixth image block merging unit, and the eighth Swin Transformer unit outputs to the fourth image block merging unit, the fourth image block merging unit outputs to the ninth Swin Transformer unit, the ninth Swin Transformer unit outputs to the fifth image block merging unit, the fifth image block merging unit outputs to the tenth Swin Transformer unit, and the tenth Swin Transformer unit outputs to the sixth image block merging unit; And the sixth image block merging unit outputs to the eleventh Swin Transformer unit, and the eleventh Swin Transformer unit outputs to the fourth image block expanding unit; The second decoder unit includes: a fourth image block expansion unit, a twelfth Swin Transformer unit, a fifth image block expansion unit, a thirteenth Swin Transformer unit, a sixth image block expansion unit, a fourteenth Swin Transformer unit, a final image expansion module, and a linear projection module, and the fourth image block expansion unit outputs to the twelfth Swin Transformer unit, the twelfth Swin Transformer unit outputs to the fifth image block expansion unit, the fifth image block expansion unit outputs to the thirteenth Swin Transformer unit, the thirteenth Swin Transformer unit outputs to the sixth image block expansion unit, the sixth image block expansion unit outputs to the fourteenth Swin Transformer unit, the fourteenth Swin Transformer unit outputs to the final image expansion module, and the final image expansion module outputs to the linear projection module; The first encoder unit is connected to the first decoder unit via a skip connection unit; The second encoder unit is connected to the second decoder unit via a skip connection unit; The first decoder unit is connected to the first encoder unit via a skip connection unit; Further, the first encoder unit is connected to the first decoder unit via a skip connection unit and comprises: The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the first jump connection unit, and then the output result of the first jump connection unit is concatted with the output of the first image block expansion unit and input into the fifth Swin Transformer unit; The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the second jump connection unit, and then the output result of the second jump connection unit is concatted with the output of the second image block expansion unit and input into the sixth Swin Transformer unit; The outputs of the first Swin Transformer unit, the second Swin Transformer unit, and the third Swin Transformer unit are concatted and input into the third jump connection unit, and then the output result of the third jump connection unit is concatted with the output of the second image block expansion unit and input into the seventh Swin Transformer unit; Further, the jump connection unit includes: a first jump connection unit, a second jump connection unit, and a third jump connection unit, and the first jump connection unit, the second jump connection unit and the third jump connection unit have the same structure; Further, the second encoder unit is connected to the second decoder unit via a skip connection unit and comprises: The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concatted and input into the fourth jump connection unit, and then the output result of the fourth jump connection unit is concatted with the output of the sixth image block expansion unit and input into the fourteenth Swin Transformer unit; The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concat-ed and input to the fifth jump connection unit. Then, the output result of the fifth jump connection unit is concat-ed with the output of the fifth image block expansion unit and input to the thirteenth Swin Transformer unit. The outputs of the eighth Swin Transformer unit, the ninth Swin Transformer unit, and the tenth Swin Transformer unit are concat-ed and input to the fifth jump connection unit. Then, the output result of the sixth jump connection unit is concat-ed with the output of the fourth image block expansion unit and input to the twelfth Swin Transformer unit. Further, the first decoder unit is connected to the first encoder unit via a skip connection unit and comprises: The outputs of the fifth Swin Transformer unit, the sixth Swin Transformer unit, and the seventh Swin Transformer unit are concatted and input into the seventh jump connection unit, and then the output result of the seventh jump connection unit is concatted with the output of the fourth image block merging unit and input into the ninth Swin Transformer unit; The outputs of the fifth Swin Transformer unit, the sixth Swin Transformer unit, and the seventh Swin Transformer unit are concat-ed and input into the eighth jump connection unit. Then, the output result of the eighth jump connection unit is concat-ed with the output of the fifth image block merging unit and input into the tenth Swin Transformer unit.

7. The oil and uranium prospecting method based on the five-in-one method according to claim 5 or 6, characterized in that: The hybrid loss function used by the earthquake deep learning model is: L=ω1L1+ω2L2, and:ω1+ω2=1 Among them, y i =sigmoid(x i ), xi is the input of the deep learning network, y i ' is the pixel label, α is the ratio of the number of non-fault pixels to the total number of pixels; N is the total number of pixels in the seismic image, i is the pixel number, ω1 and ω2 are weight coefficients respectively.

8. The oil and uranium prospecting method based on the five-in-one method according to claim 7 is characterized in that: The weight coefficients ω1 and ω2 in the hybrid loss function are dynamically acquired over time, and the calculation method is: in: L t =ω 1,t-1 L 1,t +oh 2,t-1 L 2,t Among them, L j,t is the jth loss function at time t, u(L j,t-1 ) represents the average value of the j-th loss before time t-1, σ(r i,t ) and μ(r i,t ) are respectively r before time t j,t The standard deviation and mean of , and the value of j is 1 or 2.

9. A five-in-one oil and uranium prospecting device, characterized in that: include: Data screening unit: Carry out data screening, where the data screening unit includes: The first screening module: collect, investigate and organize the drilling information of the oil field; The second screening module: analyzing the drilling radioactivity anomaly standards and building radioactivity anomaly screening software; The third screening module: based on the radioactive anomaly screening software, the radioactive anomaly screening module determines the borehole radioactive anomaly standard and screens out radioactive anomaly wells; Old well retest unit: Carry out old well retest, among which, the old well retest unit includes: The first retest module: establish the casing-cement ring model; The second retest module: Carry out cement ring casing logging response test; The third retest module: analyzes the functional relationship between the counting rate and the cement sheath property parameters to determine the correction coefficient; The fourth retest module: obtaining accurate shallow stratum radioactivity information based on the correction coefficient; Gamma conversion unit: Carry out drilling construction verification to realize the conversion of natural gamma to quantitative gamma. The gamma conversion unit includes: The first gamma module: collects gamma data of verification holes and oil field drilling holes; The second gamma module: corrects the gamma logging depth of the verification hole to the gamma anomaly depth range of the oil field borehole; The third gamma module: establish the regression equation between natural gamma and quantitative gamma; The fourth gamma module: use regression equation to convert natural gamma to quantitative gamma and evaluate mineralization; Seismic processing and interpretation unit: Carry out 3D seismic data processing, interpretation and analysis of faults, sand bodies and structures in the mining area; Deposit comparison unit: Carry out typical deposit comparison, where the deposit comparison unit includes: The first comparison module: collect typical uranium deposit data; The second comparison module: analysis of structural sedimentary evolution characteristics, uranium source conditions, sand bodies and sedimentary systems, ore body characteristics, and mineralization laws; The third comparison module: summarize the ore-controlling factors of typical sandstone-type uranium deposits; Comprehensive processing unit: based on the results of the data screening unit, the old well retesting unit, the gamma conversion unit, the seismic processing and interpretation unit, and the ore deposit comparison unit, the favorable uranium ore zones are determined and the mineralization potential of the working area is evaluated.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the five-in-one oil and uranium prospecting method according to any one of claims 1 to 8 are implemented.

Images