An intelligent identification method for comprehensive sweet spots in tight conglomerate reservoirs

Abstract

The present invention discloses a method for intelligently identifying comprehensive sweet spots in tight conglomerate reservoirs, and relates to the fields of oil and gas field development and machine learning technology. The method comprises the following steps: obtaining sweet spot evaluation parameters and indicators of a target well and an adjacent well, wherein the sweet spots include geological sweet spots, engineering sweet spots, and economic sweet spots; performing feature extraction and dimensionality reduction on the sweet spot evaluation parameters, and setting the adjacent well data as a training set, and setting the target well data as a prediction set; using the training set to train the gated cyclic unit to obtain a prediction model, and combining the prediction set to obtain the values ​​of the geological sweet spot evaluation index, engineering sweet spot evaluation index, and economic sweet spot evaluation index of the target well; obtaining the benefit ratio value of the target well under various depth conditions based on the entropy weight method and the VIKOR method, and predicting the comprehensive sweet spots at various depths based on the benefit ratio value. The method of the present invention can predict and identify the comprehensive sweet spots during drilling.

Description

Technical Field

[0001] The present invention relates to the field of oil and gas field development and machine learning technology, and in particular to a method for intelligently identifying comprehensive sweet spots in a tight conglomerate reservoir. Background Art

[0002] With the continuous deepening of unconventional oil and gas exploration, development and research, the concept of "sweet spot" has been continuously enriched and improved, and currently mainly includes geological sweet spots, engineering sweet spots, economic sweet spots, etc. Existing sweet spot research mainly focuses on geological sweet spots, engineering sweet spots and comprehensive sweet spots that organically integrate geology and engineering, and is widely used in well site deployment and segmented clustering optimization. Domestic sweet spot identification research mainly focuses on shale reservoirs, tight sandstones, etc., but does not involve tight conglomerates; in fact, tight conglomerates are significantly different from shale in terms of lithology and reservoir formation conditions. In addition, the characteristics of poor reservoir physical properties, strong heterogeneity, rock plasticity, and lack of natural fractures bring challenges to sweet spot identification.

[0003] Traditional sweet spot identification mostly adopts a comprehensive evaluation method. For geological sweet spots, the structural style, fracture development, reservoir physical properties, reservoir oil / gas content and other characteristics are considered. For engineering sweet spots, the brittleness index, ground stress and rock mechanics parameters are mainly considered. These studies have made the connotation of "sweet spots" more in-depth and clear, and also produced a series of invention patents related to sweet spot identification. Existing invention patents are mainly based on traditional mathematical models, multivariate linear regression, shallow neural networks and other methods to evaluate or predict sweet spots. Most of them are aimed at a single geological sweet spot or engineering sweet spot. A few patents consider the comprehensive evaluation of "geology-engineering" double sweet spots, but none of them directly consider economic-related evaluation indicators. In addition, existing invention patents have strict requirements on data. The evaluation indicators used as input often need to be obtained through prior calculation or interpretation. In addition, the intermediate process may also involve human decision-making. Therefore, the intelligence level of sweet spot identification needs to be improved.

[0004] Artificial intelligence methods have been widely used in the prediction of rock properties and geomechanical parameters, and have promoted the development of intelligent identification of sweet spots. However, there are few studies on the application of artificial intelligence methods to "sweet spots" in China, and economic-related evaluation indicators are not directly considered. At the same time, there are few studies on prediction while drilling. Summary of the invention

[0005] To solve at least one of the above problems, the present invention proposes a method for intelligently identifying comprehensive sweet spots in tight conglomerate reservoirs.

[0006] The technical solution of the present invention is: a method for intelligently identifying comprehensive sweet spots in tight conglomerate reservoirs, comprising the following steps:

[0007] S1. Obtain geological data, engineering data, and production profile data of the target well and adjacent wells, and divide them into geological sweet spot evaluation parameters and indicators, engineering sweet spot evaluation parameters and indicators, and economic sweet spot evaluation parameters and indicators;

[0008] S2, perform feature extraction and dimensionality reduction on the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters and economic sweet spot evaluation parameters, and set the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters, economic sweet spot evaluation parameters and corresponding evaluation indicators of the adjacent wells after dimensionality reduction as a training set, and set the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters, economic sweet spot evaluation parameters and corresponding evaluation indicators of the target well after dimensionality reduction as a prediction set;

[0009] S3, using the training set to train the gated cyclic unit to obtain a prediction model, based on the prediction model, combined with the prediction set to obtain the values ​​of the geological sweet spot evaluation index, engineering sweet spot evaluation index, and economic sweet spot evaluation index of the target well;

[0010] S4, based on the entropy weight method, the weights of various parameters in the geological sweet spot evaluation index, the engineering sweet spot evaluation index, and the economic sweet spot evaluation index are obtained, and then the geological sweetness factor, the engineering sweetness factor, and the economic sweetness factor at each depth of the reservoir are calculated using the predicted value in S3;

[0011] S5. Based on the entropy weight method, the weights of the geological sweetness factor, engineering sweetness factor and economic sweetness factor are obtained. Combined with the results of S4, the VIKOR method is used to obtain the benefit ratio value of the target well under various depth conditions, and the comprehensive sweet spot at each depth is predicted based on the benefit ratio value; among them, the smaller the benefit ratio value, the closer the target well is to the comprehensive sweet spot under the depth condition.

[0012] Beneficial effects: Relying only on conventional logging parameters and fracturing construction parameters under certain depth conditions during drilling, a quantitative prediction model for the geological, engineering, and economic sweet spots of tight conglomerate reservoirs was established based on deep learning methods. The entropy weight method (EM) was used to objectively weight the geological, engineering, and economic sweetness factors that quantitatively characterize the sweet spots. The geological, engineering, and economic sweet spots were then weighted and combined through the multi-criteria compromise solution sorting method (VIKOR), forming an integrated "geology-engineering-economy" intelligent identification method for tight conglomerate reservoirs. The present invention can realize the identification of sweet spots in tight conglomerate reservoirs while drilling, without the need for human decision-making, and intelligently select perforation positions with good physical properties, easy fracturing, and high output, providing technical support for improving the efficiency and intelligence of tight conglomerate oil and gas development. DETAILED DESCRIPTION

[0013] The specific implementation modes of the present invention will be described clearly and completely below with reference to examples. Obviously, the examples described are only some embodiments of the present invention, rather than all embodiments.

[0014] A method for intelligently identifying a comprehensive sweet spot in a tight conglomerate reservoir comprises the following steps:

[0015] S1. Obtain geological data, engineering data, and production profile data of the target well and adjacent wells, and divide them into geological sweet spot evaluation parameters and indicators, engineering sweet spot evaluation parameters and indicators, and economic sweet spot evaluation parameters and indicators;

[0016] Specifically, in this step, the geological sweet spot evaluation indicators are total hydrocarbon content, mud content, reservoir thickness, porosity, permeability, oil saturation, and fault distance; the engineering sweet spot evaluation indicators are fracture pressure and geostress difference coefficient; and the economic sweet spot evaluation indicator is oil production.

[0017] The geological sweet spot evaluation parameters are depth, reservoir thickness, fault distance, density, well diameter, resistivity, natural gamma, natural potential, acoustic time difference, and compensated neutron; the engineering sweet spot evaluation parameters are depth, density, well diameter, resistivity, natural gamma, natural potential, acoustic time difference, and compensated neutron; the economic sweet spot evaluation parameters are density, well diameter, resistivity, natural gamma, natural potential, acoustic time difference, compensated neutron, total liquid volume, total sand volume, acid consumption, pre-flushing liquid ratio, and average construction displacement.

[0018] For developed adjacent wells, it is easier to obtain the above-mentioned evaluation parameters and indicators, which are common parameters in this field and can be obtained through drilling logging, engineering data and production data; for target wells, it is easier to obtain the above-mentioned evaluation parameters through drilling logging while drilling.

[0019] S2, perform feature extraction and dimensionality reduction on the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters and economic sweet spot evaluation parameters, and set the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters, economic sweet spot evaluation parameters and corresponding evaluation indicators of the adjacent wells after dimensionality reduction as a training set, and set the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters, economic sweet spot evaluation parameters and corresponding evaluation indicators of the target well after dimensionality reduction as a prediction set;

[0020] In this step, the feature extraction and dimensionality reduction operation is mainly performed through an autoencoder, wherein the autoencoder includes an encoder and a decoder, the encoder compresses parameters into low-dimensional features, and the decoder reconstructs the low-dimensional features back to the original data; the encoder and the decoder are both composed of an input layer, two hidden layers and an output layer, and the hidden layer contains two fully connected layers.

[0021] In this embodiment, considering that the number of parameters of geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters and economic sweet spot evaluation parameters are different, when using the autoencoder to perform feature extraction and dimensionality reduction operations, although the structures of the encoder and decoder are the same, the number of nodes inside them has certain differences.

[0022] For the geological sweet spot evaluation parameters, there are a total of 10 parameters. In its encoder, the input dimension is 10, the number of nodes in the two hidden layers is 64 and 32 respectively, and the activation functions of the two hidden layers are ReLU, and the output dimension is 5. In its decoder, the input features are the low-dimensional features output by the encoder, the number of nodes in the two hidden layers is 32 and 64 respectively, and the activation functions of the two hidden layers are ReLU, and the reconstructed geological sweet spot evaluation parameters are output to verify the feature extraction effect of the encoder on the geological sweet spot evaluation parameters.

[0023] For the engineering sweet spot evaluation parameters, there are a total of 8 parameters. In the encoder, the input dimension is 8, the number of nodes in the two hidden layers is 32 and 16 respectively, and the activation functions of the two hidden layers are both ReLU, and the output dimension is 5. In the decoder, the input features are the low-dimensional features of the encoder's output, the number of nodes in the two hidden layers is 16 and 32 respectively, and the activation functions are both ReLU. The reconstructed engineering sweet spot evaluation parameters are output to verify the encoder's feature extraction effect on the engineering sweet spot evaluation parameters.

[0024] For the economic sweet spot evaluation parameters, there are a total of 12 parameters. In the encoder, the input dimension is 12, the number of nodes in the two hidden layers is 64 and 32 respectively, and the activation functions of the two hidden layers are ReLU, and the output dimension is 5. In the decoder, the input features are the low-dimensional features output by the encoder, the number of nodes in the two hidden layers is 32 and 64 respectively, and the activation functions of the two hidden layers are ReLU. The reconstructed economic sweet spot evaluation parameters are output to verify the encoder's feature extraction effect on the economic sweet spot evaluation parameters.

[0025] For the training set, there are geological sweet spot training set, engineering sweet spot training set and economic sweet spot training set, and the prediction set includes geological sweet spot prediction set, engineering sweet spot prediction set and economic sweet spot prediction set;

[0026] Among them, the geological sweet spot training set includes the features of the geological sweet spot evaluation parameters of the adjacent wells after dimension reduction and the geological sweet spot evaluation index, the engineering sweet spot training set includes the features of the engineering sweet spot evaluation parameters of the adjacent wells after dimension reduction and the engineering sweet spot evaluation index, and the economic sweet spot training set includes the features of the economic sweet spot evaluation parameters of the adjacent wells after dimension reduction and the economic sweet spot evaluation index. The geological sweet spot prediction set includes the features of the geological sweet spot evaluation parameters of the target well after dimension reduction, the engineering sweet spot prediction set includes the features of the engineering sweet spot evaluation parameters of the target well after dimension reduction, and the economic sweet spot prediction set includes the features of the economic sweet spot evaluation parameters of the target well after dimension reduction.

[0027] S3, using the training set to train the gated cyclic unit to obtain a prediction model, based on the prediction model, combined with the prediction set to obtain the values ​​of the geological sweet spot evaluation index, engineering sweet spot evaluation index, and economic sweet spot evaluation index of the target well;

[0028] As for the gated recurrent unit, it is a conventional model in the art, so its specific structure and operation mode are not described in detail here. In particular, in this embodiment, when the gated recurrent unit is trained using the training set, the geological sweet spot training set, the engineering sweet spot training set and the economic sweet spot training set are used to train the gated recurrent unit, and the final training obtains the geological sweet spot prediction module, the engineering sweet spot prediction module and the economic sweet spot prediction module, respectively. These three prediction modules constitute the prediction model of this embodiment.

[0029] After obtaining the prediction model, the geological sweet spot prediction set, engineering sweet spot prediction set and economic sweet spot prediction set are substituted into the geological sweet spot prediction module, engineering sweet spot prediction module and economic sweet spot prediction module respectively to predict the sweet spot evaluation indicators of different depths of the target well.

[0030] S4, based on the entropy weight method, obtain the weights of each parameter in the geological sweet spot evaluation index, engineering sweet spot evaluation index, and economic sweet spot evaluation index, and then use the predicted value in S3 to calculate the geological sweetness factor, engineering sweetness factor, and economic sweetness factor at each depth of the reservoir; this step includes the following sub-steps:

[0031] S41. Normalize the parameters in the geological sweet spot evaluation index, the engineering sweet spot evaluation index, and the economic sweet spot evaluation index. Normalization is a common data processing method in this field, so the specific processing method will not be described in detail here.

[0032] S42. Make negative indicators positive: , where is the positive value of the jth normalized negative evaluation index, is the jth normalized negative indicator;

[0033] S43. Based on the entropy weight method, the weights of various parameters in the geological sweet spot evaluation index, the engineering sweet spot evaluation index, and the economic sweet spot evaluation index are obtained;

[0034] S44. Calculate the geological sweetness factor, engineering sweetness factor and economic sweetness factor at each depth.

[0035] Obtaining the weight of each indicator by entropy weight method is a common operation in this field, so the specific method is not described here. Among them, the geological sweetness factor is calculated by the following formula: , where is the predicted value of the evaluation index of the jth geological sweet spot at the i-th depth output by the geological sweet spot gated recurrent unit module, is the geological sweetness factor of the target well at the i-th depth, is the weight of the evaluation index of the jth geological sweet spot of the target well at the i-th depth, n ego is the number of geological sweet spot evaluation indicators, and j=1,2,……,n ego ; i∈I m , I m is the depth range of the horizontal section of the target well.

[0036] The engineering sweetness factor is calculated by the following formula: , where is the predicted value of the kth engineering sweet spot evaluation index at the i-th depth output by the engineering sweet spot gated recurrent unit module, is the engineering sweetness factor of the target well at the i-th depth, is the weight of the kth engineering sweet spot evaluation index of the target well at the i-th depth, is the number of engineering sweet spot evaluation indicators, and k=1,2,……,n eng ; i∈I m .

[0037] As for the economic sweetness factor, since the only economic dessert evaluation index in this embodiment is the oil yield, the final normalized oil yield is the economic sweetness factor.

[0038] S5. Obtain the weights of geological sweetness factor, engineering sweetness factor and economic sweetness factor based on the entropy weight method, combine the results of S4, use the VIKOR method to obtain the benefit ratio value of the target well under various depth conditions, and predict the comprehensive sweet spot of each depth based on the benefit ratio value; wherein, the smaller the benefit ratio value, the closer the target well is to the comprehensive sweet spot under the depth condition. In this embodiment, the VIKOR method includes the following steps:

[0039] Compute positive and negative ideal solutions: =( , , ), =( , , ), where represents the positive ideal solution, represents the geological sweetness factor, Represents the engineering sweetness factor, represents the economic sweetness factor, represents a negative ideal solution;

[0040] Calculate the group benefit value and individual regret value: , , where represents the group benefit value of the i-th depth, is the weight of the jth sweetness factor, is the value of the jth sweetness factor under the i-th depth condition; represents the individual regret value of the i-th depth;

[0041] Calculate the benefit ratio value: , where represents the benefit ratio value of the i-th candidate cluster perforation position, , Indicates the maximum group benefit value and the minimum group benefit value at all depths. , represents the maximum individual regret value and the minimum individual regret value in all depths, For the compromise coefficient, we choose 0.5, which means that we pay attention to the integrated sweet spot index of "geology, engineering, and economy" and consider the single sweet spot index.

[0042] Conventional methods may only focus on a single sweet spot, or only consider the comprehensive sweet spots of geology and engineering for simple weighted summation, lacking certain references and standards. In this embodiment, our method organically integrates geology, engineering, and economic sweet spots, and uses the VIKOR method to determine the optimal solution and the most cracked solution under ideal conditions in the target well area. In the process of sweet spot identification, the two objective standards of optimal solution and most cracked solution are used as references for quantitative evaluation, avoiding artificial division of standards and subjective decision-making.

[0043] The above description is only a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A method for intelligently identifying comprehensive sweet spots in tight conglomerate reservoirs, characterized in that: The following steps are involved: S1. Obtain geological data, engineering data, and production profile data of the target well and adjacent wells, and divide them into geological sweet spot evaluation parameters and indicators, engineering sweet spot evaluation parameters and indicators, and economic sweet spot evaluation parameters and indicators; the geological sweet spot evaluation indicators are total hydrocarbon content, mud content, reservoir thickness, porosity, permeability, oil saturation, and fault distance; the engineering sweet spot evaluation indicators are fracture pressure and geostress difference coefficient; and the economic sweet spot evaluation indicator is oil production; S2, perform feature extraction and dimensionality reduction on the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters and economic sweet spot evaluation parameters, and set the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters, economic sweet spot evaluation parameters and corresponding evaluation indicators of the adjacent wells after dimensionality reduction as a training set, and set the geological sweet spot evaluation parameters, engineering sweet spot evaluation parameters, economic sweet spot evaluation parameters and corresponding evaluation indicators of the target well after dimensionality reduction as a prediction set; S3, using the training set to train the gated cyclic unit to obtain a prediction model, based on the prediction model, combined with the prediction set to obtain the values ​​of the geological sweet spot evaluation index, engineering sweet spot evaluation index, and economic sweet spot evaluation index of the target well; S4, based on the entropy weight method, obtain the weights of each parameter in the geological sweet spot evaluation index, engineering sweet spot evaluation index, and economic sweet spot evaluation index, and then use the predicted value in S3 to calculate the geological sweetness factor, engineering sweetness factor, and economic sweetness factor at each depth of the reservoir; this step includes the following sub-steps: S41, normalizing each parameter in the geological sweet spot evaluation index, the engineering sweet spot evaluation index, and the economic sweet spot evaluation index; S42. Make negative indicators positive: , where It is j The normalized negative evaluation index is the positive value. For the j A normalized negative indicator; S43. Based on the entropy weight method, the weights of various parameters in the geological sweet spot evaluation index, the engineering sweet spot evaluation index, and the economic sweet spot evaluation index are obtained; S44. Calculate the geological sweetness factor, engineering sweetness factor and economic sweetness factor at each depth: , where is the geological sweetness factor, is the output of the Geodesic Gated Recurrent Unit module. i The predicted value of the evaluation index of the jth geological sweet spot at the depth, The target well i A geological sweetness factor at depth, The target well is i The weight of the evaluation index of the jth geological sweet spot at the depth, n ego is the number of geological sweet spot evaluation indicators, and j=1,2,……, n ego ; i ∈ I m , I m is the depth range of the horizontal section of the target well; , where It is the engineering sweetness factor. It is the output of the engineering sweet spot gated recurrent unit module. i The depth of k The predicted value of the engineering sweet spot evaluation index, The target well i A deep engineering sweetness factor, The target well is i The depth of k The weight of the engineering sweet spot evaluation index, is the number of engineering sweet spot evaluation indicators, and k =1,2,……, n eng ; i ∈ I m ; The normalized oil yield is the economic sweetness factor; S5. Based on the entropy weight method, the weights of the geological sweetness factor, engineering sweetness factor and economic sweetness factor are obtained. Combined with the results of S4, the VIKOR method is used to obtain the benefit ratio value of the target well under various depth conditions, and the comprehensive sweet spot at each depth is predicted based on the benefit ratio value; among them, the smaller the benefit ratio value, the closer the target well is to the comprehensive sweet spot under the depth condition.

2. The method according to claim 1, characterized in that In S1, the geological sweet spot evaluation parameters are depth, reservoir thickness, fault distance, density, well diameter, resistivity, natural gamma, natural potential, acoustic time difference and compensated neutron; the engineering sweet spot evaluation parameters are depth, density, well diameter, resistivity, natural gamma, natural potential, acoustic time difference and compensated neutron; the economic sweet spot evaluation parameters are density, well diameter, resistivity, natural gamma, natural potential, acoustic time difference, compensated neutron, total liquid volume, total sand volume, acid consumption, pre-flushing liquid ratio and average construction displacement.

3. The method according to claim 1, characterized in that: In S2, the input parameters of the geological sweet spot evaluation index, the engineering sweet spot evaluation index, and the economic sweet spot evaluation index are reduced in dimension through an autoencoder, and the autoencoder includes an encoder and a decoder. The encoder compresses the parameters into low-dimensional features, and the decoder reconstructs the low-dimensional features back to the original data; the encoder and the decoder are both composed of an input layer, two hidden layers and an output layer, and the hidden layer includes two fully connected layers.

4. The method according to claim 1, characterized in that: In S2, the training set includes a geological sweet spot training set, an engineering sweet spot training set and an economic sweet spot training set, and the prediction set includes a geological sweet spot prediction set, an engineering sweet spot prediction set and an economic sweet spot prediction set; in S3, the prediction model includes a geological sweet spot prediction module, an engineering sweet spot prediction module and an economic sweet spot prediction module.

5. The method according to claim 1, characterized in that In S5, the VIKOR method includes the following sub-steps: Compute positive and negative ideal solutions: =( , , ), =( , , ), where represents the positive ideal solution, represents the geological sweetness factor, Represents the engineering sweetness factor, represents the economic sweetness factor, represents a negative ideal solution; Calculate the group benefit value and individual regret value: , , where Indicates i The group benefit value of a depth, It is j The weight of the sweetness factor, It is i The depth condition j The value of the sweetness factor; Indicates i Individual regret values ​​of depth; Calculate the benefit ratio value: , where Indicates i The benefit ratio of the candidate cluster perforation positions is , Indicates the maximum group benefit value and the minimum group benefit value at all depths. , represents the maximum individual regret value and the minimum individual regret value in all depths, is the compromise coefficient.

Images