A residual oil classification method based on random forest-fuzzy mathematics

Abstract

The present invention discloses a residual oil classification method based on random forest-fuzzy mathematics, comprising the following steps: S1: collecting geological data; S2: screening control factors and constructing residual oil classification standards; S3: classifying each well group of the oil layer through a random forest algorithm and performing correlation calculation; S4: creating a decision tree to determine the residual oil type of the well group, and dividing the residual oil type of each sublayer through a fuzzy mathematics algorithm. The beneficial effect of the present invention is that through residual oil classification, random forest algorithm and fuzzy mathematics algorithm, the key types of residual oil in each sublayer are divided, thereby providing a basis for the evaluation and comprehensive adjustment of subsequent residual oil potential tapping plans.

Description

Technical Field

[0001] The invention relates to the technical field of oil and gas field development, in particular to a residual oil classification method based on random forest-fuzzy mathematics. Background Art

[0002] At present, the distribution of residual oil in reservoirs is still a challenge that is difficult to completely solve in the global petroleum industry. How to characterize the distribution of residual oil in macroscopic and microscopic heterogeneous reservoirs after displacement has become a hot topic in international conferences.

[0003] Existing residual oil classification technologies, such as residual oil morphology classification based on deep learning, require a lot of computing resources and time costs when processing high-dimensional data and complex models; for example, the residual oil classification method for oil fields in the ultra-high water-cut period has more complex and uneven distribution of residual oil, and there are various nonlinear interactions between various influencing factors, which makes accurate classification difficult; for example, seismic exploration technology uses seismic data to classify residual oil, but since seismic exploration technology usually has limited resolution of deep geological structures and inaccurate residual oil detection in heterogeneous formations, it may lead to low residual oil classification accuracy. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a residual oil classification method based on random forest-fuzzy mathematics.

[0005] The purpose of the present invention is achieved by the following technical solution: A residual oil classification method based on random forest-fuzzy mathematics comprises the following steps:

[0006] S1: Collect geological data;

[0007] S2: Screening control factors and constructing residual oil classification standards;

[0008] S3: Classify each well group in the oil layer through the random forest algorithm and calculate the correlation;

[0009] S4: Create a decision tree to determine the type of remaining oil in the well group, and divide the remaining oil type of each sublayer through fuzzy mathematical algorithm.

[0010] Preferably, in step S1, the geological data includes static data and dynamic data, the static data includes reservoir thickness, oil saturation, porosity, original formation pressure, net-to-gross ratio, microstructure, oil-bearing height, fault, viscosity, kh value, vertical permeability difference and plane permeability difference; the dynamic data includes oil well recovery degree, phase permeability curve, oil well location, oil and water well production data and injection and production well network.

[0011] Preferably, in step S2, the control factors include microstructure, fault, kh value, vertical permeability difference, plane permeability difference, injection-production well pattern and injection-production flow field.

[0012] Preferably, in step S3, the correlation calculation step is:

[0013] S31: Calculation Statistics are used to assign different extraction probabilities to each attribute. Suppose the category set Y has q different values , attribute A has p different values ,

[0014] ;

[0015] ;

[0016] ;

[0017] in, For the sample set to meet and The number of samples, For the sample set to meet The number of samples, For the sample set to meet The number of samples, is the total number of samples;

[0018] If A and Y are independent, then:

[0019] ;

[0020] ;

[0021] in, Express satisfaction and The probability of the number of samples is Express satisfaction The probability of the number of samples is Express satisfaction The probability of the number of samples;

[0022] S32: Construction through the independence of A and Y Statistics,

[0023] ;

[0024] ;

[0025] in, is the chi-square value obtained by performing chi-square statistics on the eigenvalues ​​A and Y, is the expected frequency;

[0026] Proportional to the correlation, when building a decision tree, each node to be split calculates the statistics of each feature , and use the weights as the probability of feature sampling, Statistics standardization, feature A n The weight is:

[0027] ;

[0028] in, Represents the random sampling probability of eigenvalue A in the subspace.

[0029] Preferably, in step S4, the fuzzy mathematics evaluation process is:

[0030] S41: Determine the judgment set, factor set and weight;

[0031] S42: Obtaining the degree of membership;

[0032] S43: Transformation by fuzziness calculate , and the final classification result is obtained: According to the maximum membership principle, when At maximum, ,but The corresponding remaining oil type is the remaining oil distribution type of this sublayer.

[0033] Preferably, in step S42, according to the random forest algorithm, the correlation of each factor is used as the element of the fuzzy relationship matrix ,

[0034] ;

[0035] in, It is an indicator evaluation matrix.

[0036] The present invention has the following advantages: the present invention realizes the division of key types of residual oil in each sublayer through residual oil classification, random forest algorithm and fuzzy mathematical algorithm, thereby providing a basis for subsequent residual oil potential tapping program evaluation and comprehensive adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a structural schematic diagram of the residual oil classification method process;

[0038] Figure 2 A schematic diagram of random forest classification and a schematic diagram of the structure of a single decision tree training process in a random forest;

[0039] Figure 3It is a structural diagram of the calculation results of the correlation of various control factors of the A well group in the G layer;

[0040] Figure 4 This is a schematic diagram of the structure of the remaining oil saturation distribution of the A well group in the G layer. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions 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 part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0042] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0044] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0045] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, or the positions or positional relationships in which the product of the invention is usually placed when in use, or the positions or positional relationships commonly understood by those skilled in the art, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0046] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0047] In this embodiment, if Figure 1 As shown, a residual oil classification method based on random forest-fuzzy mathematics includes the following steps:

[0048] S1: Collect geological data;

[0049] S2: Screening control factors and constructing residual oil classification standards;

[0050] S3: Classify the well groups in the oil layer through the random forest algorithm and calculate the correlation; specifically, the random forest is based on Decision Tree is the basic classifier, and the combined classifier is obtained after ensemble learning, where It is a random variable sequence based on the bagging idea and feature subspace idea of ​​random forest.

[0051] S4: Create a decision tree to determine the type of remaining oil in the well group, and divide the remaining oil types of each sub-layer through fuzzy mathematical algorithms. Through the remaining oil classification, random forest algorithm and fuzzy mathematical algorithms, the key types of remaining oil in each sub-layer can be divided, thus providing a basis for the evaluation and comprehensive adjustment of the subsequent remaining oil potential tapping plan.

[0052] Further, in step S1, the geological data includes static data and dynamic data, the static data includes reservoir thickness, oil saturation, porosity, original formation pressure, net-to-gross ratio, microstructure, oil-bearing height, fault, viscosity, kh value, vertical permeability difference and plane permeability difference; the dynamic data includes oil well recovery degree, phase permeability curve, oil well location, oil and water well production data and injection and production well network. Specifically, the specific values ​​of static data are shown in Table 1:

[0053] Table 1 Static data table

[0054] Reservoir thickness (m) Average porosity Original formation pressure (kpa) Net to Gross Ratio Microstructure (m) 122 0.206 11376.6 0-1 4-6 Number of faults Viscosity (mPa·s) Average kh value (D mm) Vertical permeability difference Plane permeability difference 10 0.6-9.8 12.46 17.5 38.2

[0055] Furthermore, in step S2, the control factors include micro-structure, fault, kh value, longitudinal permeability difference, plane permeability difference, injection-production well network and injection-production flow field. Specifically, a classification system based on 7 types of residual oil is established, namely, micro-structure type, fault shielding type, thin-differential reservoir type, interlayer interference type, plane interference type, imperfect well network type and flow field symmetry retention type. Micro-structure: the micro-length of the structure with micro-amplitude high points in the well group is used as the standard; fault: the proportion of faults in the well group and the distance from the well are used as the standard; kh value: the kh value measured in the well group is used as the standard; longitudinal permeability difference: the longitudinal permeability difference measured in the well group is used as the standard; plane permeability difference: the plane permeability difference measured in the well group is used as the standard; injection-production well network: the degree of perfection of the well network in the well group is used as the standard; injection-production flow field: the proportion of dominant seepage channels in the well group is used as the standard.

[0056] In this embodiment, if Figure 2 As shown, in step S3, the correlation calculation step is:

[0057] S31: Calculation Statistics are used to assign different extraction probabilities to each attribute. Suppose the category set Y has q different values , attribute A has p different values ,

[0058] ;

[0059] ;

[0060] ;

[0061] in, For the sample set to meet and The number of samples, For the sample set to meet The number of samples, For the sample set to meet The number of samples, is the total number of samples;

[0062] If A and Y are independent, then:

[0063] ;

[0064] ;

[0065] in, Express satisfaction and The probability of the number of samples is Express satisfaction The probability of the number of samples is Express satisfaction The probability of the number of samples;

[0066] S32: Construction through the independence of A and Y Statistics,

[0067] ;

[0068] ;

[0069] in, is the chi-square value obtained by performing chi-square statistics on the eigenvalues ​​A and Y, is the expected frequency;

[0070] Proportional to the correlation, when building a decision tree, each node to be split calculates the statistics of each feature , and use the weights as the probability of feature sampling, Statistics standardization, feature A n The weight is:

[0071] ;

[0072] in, Represents the random sampling probability of the eigenvalue A in the subspace. Specifically, the weighted sampling method is used to extract the attribute subset, so as to increase the probability of the attribute related to the classification being selected. For example, taking the classification of the remaining oil type of well group A in the G sublayer as an example, the remaining oil type is selected according to the remaining oil classification standard, and the micro-structure, fault, kh value, vertical permeability difference, plane permeability difference, injection-production well network, and injection-production flow field are attribute nodes. The correlation calculation is carried out through the weighted sampling method of the attribute subset, and then the random forest algorithm is used to calculate the remaining oil type of each decision tree. Finally, the main type of remaining oil in the well group is determined by voting, that is, the well group A in the G sublayer is defined as fault-blocked remaining oil, such as Figure 3 and Figure 4 shown.

[0073] In this embodiment, in step S4, the fuzzy mathematics evaluation process is:

[0074] S41: Determine the judgment set, factor set and weight;

[0075] S42: Obtain the degree of membership; further, according to the random forest algorithm, the correlation of each factor is used as the element of the fuzzy relationship matrix ,

[0076] ;

[0077] in, is - indicator evaluation matrix;

[0078] S43: Transformation by fuzziness calculate , and the final classification result is obtained: According to the maximum membership principle, when At maximum, ,but The corresponding residual oil type is the residual oil distribution type of the sublayer. Specifically, in the classification of the main types of residual oil distribution, the use of a single factor is too simple, and there are many uncertain factors in the classification process that cause "ambiguity", so by introducing a mathematical fuzzy discrimination method to discriminate the main types of residual oil, it is possible to effectively characterize the effect of each factor on the residual oil distribution type, achieve a comprehensive judgment of multiple factors, and determine the main types of residual oil in each sublayer. For example, taking the G sublayer as an example, the correlation of each control factor of each well group calculated by random forest is used as the matrix R, and the ratio of the residual oil reserves of each well group is used as the weight coefficient of each well group, and the matrix product of the public step S42 is used for fuzzy comprehensive evaluation, that is,

[0079] Weight coefficient matrix A=[a1a2 ...a 48 ]=[0.0232 0.0514 ...0.0318];

[0080] ;

[0081] The final evaluation result is: B=[0.0232 0.169 0.384 0.132 0.155 0.0601 0.0767]. It can be determined that the thin and poor reservoir type is the main type of remaining oil distribution in the G sublayer. The above algorithm is used to carry out fuzzy mathematical evaluation with the sublayer as the scale to determine the main type of remaining oil in each sublayer.

[0082] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A residual oil classification method based on random forest-fuzzy mathematics, characterized in that: The following steps are involved: S1: Collect geological data; S2: Screening control factors and constructing residual oil classification standards; S3: Classify each well group in the oil layer through the random forest algorithm and calculate the correlation; S4: Create a decision tree to determine the type of remaining oil in the well group, and divide the remaining oil type of each sublayer by fuzzy mathematical algorithm; In step S1, the geological data includes static data and dynamic data, wherein the static data includes reservoir thickness, oil saturation, porosity, original formation pressure, net-to-gross ratio, microstructure, oil-bearing height, fault, viscosity, KH value, vertical permeability difference and plane permeability difference; the dynamic data includes oil well recovery degree, phase permeability curve, oil well location, oil and water well production data and injection-production well network; In step S2, the control factors include microstructure, fault, kh value, vertical permeability difference, plane permeability difference and injection-production well pattern; In step S3, the correlation calculation step is: S31: Calculation Statistics are used to assign different extraction probabilities to each attribute. Suppose the category set Y has q different values , attribute A has p different values , in, For the sample set to meet and The number of samples, For the sample set to meet The number of samples, For the sample set to meet The number of samples, is the total number of samples; If A and Y are independent, then: ； ； in, Express satisfaction and The probability of the number of samples is Express satisfaction The probability of the number of samples is Express satisfaction The probability of the number of samples; S32: Construction through the independence of A and Y Statistics, in, is the chi-square value obtained by performing chi-square statistics on the eigenvalues ​​A and Y, is the expected frequency; Proportional to the correlation, when building a decision tree, each node to be split calculates the statistics of each feature , and use the weights as the probability of feature sampling, Statistics standardization, feature A n The weight is: ； in, represents the random sampling probability of eigenvalue A in the subspace; In step S4, the fuzzy mathematics evaluation process is: S41: Determine the judgment set, factor set and weight; S42: Obtain membership degree; S43: Transformation by fuzziness calculate , and the final classification result is obtained: According to the maximum membership principle, when At maximum, ,but The corresponding remaining oil type is the small layer remaining oil distribution type.

2. The residual oil classification method based on random forest-fuzzy mathematics according to claim 1 is characterized in that: In step S42, according to the random forest algorithm, the correlation of each factor is used as the element of the fuzzy relationship matrix , ； in, It is an indicator evaluation matrix.

Images