Method for improving prediction precision of geological parameter electrical property multivariate model
By combining closed coring, nuclear magnetic resonance logging, and conventional logging in the oilfield, a single rhythmic subdivision standard and an electrical multivariate fitting model were established, which solved the problem of low interpretation accuracy of electrical prediction models and achieved high-precision prediction of geological parameters.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2026-04-10
AI Technical Summary
In existing technologies, when establishing electrical prediction models using a limited number of data sources, the interpretation accuracy is low, making it difficult to effectively apply conventional logging data during oilfield development.
By combining closed coring, nuclear magnetic resonance logging, and conventional logging, a single rhythmic subdivision standard was established. The electrical values were extracted using the morphological geometric mean method and the data point electrical average method. The parameter weighting coefficients were determined, and a multivariate fitting model of geological parameter electrical properties was established.
It significantly improves the prediction accuracy of porosity, permeability and bound water saturation. The accuracy of the porosity model is improved to over 92%, and the relative error of the permeability model is reduced to below 50%.
Smart Images
Figure CN121835104A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of oil reservoir engineering, in particular to a method for improving the prediction accuracy of a geological parameter electrical multi-element model. BACKGROUND
[0002] Reservoir geological parameters mainly include porosity, permeability, irreducible water saturation, and current water saturation parameters, among which, the permeability prediction is the most difficult.
[0003] Current geological parameter models are mainly divided into two categories: one depends on core experiment data, and the other is based on conventional logging. These two types of models at least have the following problems:
[0004] For the model depending on core experiment data, it cannot be widely used in conventional logging; for the model based on conventional logging, it mainly relies on big data analysis. However, in the actual oilfield development process, the data that can be used for modeling in the hands of researchers are only a small amount of core data and special logging data such as nuclear magnetic logging. Therefore, it is the most urgent and important problem for oilfields to use a small amount of data sources to establish an electrical property prediction model and improve the interpretation accuracy, and to apply conventional logging curves to interpret reservoir geological characteristics. SUMMARY
[0005] The present application is aimed at the problem of low interpretation accuracy of the existing method for establishing an electrical property prediction model using a small amount of data sources, and provides a method for improving the prediction accuracy of a geological parameter electrical multi-element model. The method for improving the prediction accuracy of a geological parameter electrical multi-element model can establish a geological parameter electrical multi-element fitting model, and greatly improve the accuracy.
[0006] The present application solves the problem by the following technical scheme: the method for improving the prediction accuracy of a geological parameter electrical multi-element model comprises the following steps:
[0007] S1: sealed coring is performed in a target work area to detect the required geological parameters of the reservoir;
[0008] S2: nuclear magnetic logging is performed in the target work area to detect the required geological parameters of the reservoir;
[0009] S3: conventional logging is performed in the target work area to collect the required logging data of the reservoir;
[0010] S4: based on the electrical response characteristics of step S3, the target layer of modeling and model prediction is standardized, and the single rhythm subdivision layer standard and the electrical property value standard of the subdivision layer are established;
[0011] S5: based on the obtained geological parameters and logging data, different electrical property value methods are matched through different data sources;
[0012] S6: the weight coefficients of each parameter are determined;
[0013] S7: geological parameter electrical property multi-fitting model precision evaluation.
[0014] Further, the detected geological parameters include porosity, permeability and irreducible water saturation; and the collected logging data include natural gamma ray logging curve GR, density logging curve DEN, micro-electrode logging curve RM, acoustic logging curve HAC and spontaneous potential logging curve SP.
[0015] Further, the single rhythm subdivision layer standard in step S4 is that the micro-potential curve return of 14% is taken as the standard for subdividing layers.
[0016] Further, in the electrical property value setting standard in step S4, the electrical property value setting method includes three methods of fluctuation peak value method, peak average method and morphological geometric average value method.
[0017] Further, the electrical property value setting method is preferably the morphological geometric average value method.
[0018] When the electrical property value setting is in the obvious interlayer, the three methods of fluctuation peak value method, peak average method and morphological geometric average value method are compared, and the calculation results are similar.
[0019] When the electrical property value setting is in the transition type rhythm caused by lithology change and the interlayer is not obvious, the morphological geometric average value method is selected, and the morphological geometric average value method is more consistent with the electrical property response characteristics in the logging principle and has higher prediction accuracy.
[0020] Further, the feature that the interlayer is not obvious is that the micro-potential curve return is in the range of greater than 5% and less than 14%.
[0021] Further, the step S5 of matching different electrical property value setting methods through different data sources includes:
[0022] S501: based on the core data in step 1, in order to improve the fitting precision of core detected geological parameters and electrical property, the single rhythm morphological extreme value method is used to extract the electrical property value;
[0023] S502: based on the nuclear magnetic data in step 2, in order to improve the fitting precision of nuclear magnetic logging detected geological parameters and electrical property, the multi-rhythm sampling point value average method is used to extract the electrical property value.
[0024] Further, the specific steps of the single rhythm morphological extreme value method for extracting the electrical property value include:
[0025] First, the core data of the coring well is classified into a single rhythm;
[0026] Second, the core detected geological parameters are normalized and calculated.
[0027] Third, the core representative is calculated.
[0028] Fourth, select reservoirs with high representative cores, extract the geometric mean of rhythmic morphology and perform electrical fitting with the standard core values, select sensitive parameters and establish a univariate model;
[0029] Furthermore, specific methods for optimizing sensitive parameters and establishing a univariate model include:
[0030] Optimize sensitive parameters and geological parameters to create scatter plots;
[0031] Then add a trend line for the data points, and set the regression mathematical formula for the trend line based on the principle of maximizing the correlation coefficient; thus, a univariate model is established.
[0032] The formula can be a linear, power function, exponential, or logarithmic function;
[0033] The preferred method for the sensitive parameters is as follows: Based on descriptive analysis of well logging interpretation principles, the microelectrode is preferably a permeability-sensitive parameter. The microelectrode responds to the permeability difference based on the difference in formation resistivity amplitude corresponding to different detection depths of microgradients and micropotentials; the larger the amplitude difference, the better the permeability. Spontaneous potential is preferably a sensitive parameter, as it affects formation permeability based on the difference in formation water and mud salinity. Density is preferably a sensitive parameter, as density responds to the size of formation pores, and gamma responds to the size of formation micropores; the two work together to indirectly reflect permeability. Acoustic logging is preferably a sensitive parameter, as acoustic logging responds to the density of the formation and can also indirectly reflect permeability.
[0034] Furthermore, the method for extracting electrical values using the multi-rhythm sampling point numerical averaging method includes: rhythmic morphological geometric average method, extreme value method, and data point electrical averaging method.
[0035] Furthermore, the preferred method for extracting electrical values using the multi-rhythm sampling point numerical averaging method is the data point electrical averaging method.
[0036] The specific method of the data point electrical averaging method is as follows: read an electrical value at certain intervals from the logging curve, and read all the electrical values.
[0037] Based on all the electrical values read, calculate the average value of all the above electrical values;
[0038] And / or,
[0039] If the interval between the logging curves is 0.05m, then the total number of electrical values read is: layer thickness / 0.05.
[0040] Furthermore, step S6 determines the weight coefficients of each parameter, and the methods include: mathematical equation solving method, statistical least squares method, correction coefficient method and goodness-of-fit weight coefficient method.
[0041] Furthermore, the method for evaluating the accuracy of the geological parameter electrical multivariate fitting model in step S7 includes:
[0042] S701: Based on the univariate fitting model of each parameter, the coefficients are matched by the goodness-of-fit weighting coefficient method to establish a multivariate model.
[0043] S702: Evaluate the multivariate model according to the model evaluation criteria.
[0044] Furthermore, the goodness-of-fit weighting coefficient method specifically includes:
[0045] Obtaining the goodness of fit R 2 The goodness of fit R 2 This represents the degree of influence of the independent variable on the dependent variable;
[0046] The ratio of the univariate goodness of fit of the independent variable to the sum of the goodness of fits is determined as the weight coefficient of the multiple regression.
[0047] Furthermore, the regression formula for the goodness-of-fit weight coefficients is as follows:
[0048] D=(R 2 / Σ(R 2 );
[0049] In the formula: D is the multiple regression coefficient, R 2 The goodness of fit of the univariate model for each parameter.
[0050] Furthermore, based on the univariate fitting model for each parameter, and using the goodness-of-fit weighted coefficient method to match coefficients, the specific methods for establishing a multivariate model include:
[0051] Assume the univariate models for each parameter are linear, power function, logarithmic, and exponential function, respectively, with the following relationships: y = a1x1 + b1; y = a2x2 0.5507 ; y = a3ln(x) - b3; y = a4e b4x The multiple regression formula is: D1*(a1x1+b1)+D2*(a2x2) 0.5507 )+D3*(a3ln(x)-b3)+D4*(a4e b4x );
[0052] In the formula: a1 is the coefficient of a univariate function, and b1 is the constant of a univariate function;
[0053] a2 represents the coefficients of the power function, and b2 represents the constant of the power function.
[0054] a3 is the coefficient of the logarithmic function, and b3 is the constant of the logarithmic function;
[0055] a4 represents the coefficients of the exponential function, and b4 represents the constant of the exponential function.
[0056] D1 is a monomial regression coefficient; D2 is a power function regression coefficient; D3 is a logarithmic function regression coefficient; and D4 is an exponential function regression coefficient.
[0057] Further, the model evaluation criteria comprise:
[0058] The core data non-modeling data source for model evaluation;
[0059] The model application object is a single rhythm subdivided layer.
[0060] The core well participates in the evaluation, and the core is not selected subjectively.
[0061] The error layer caused by the low representativeness of the small number of cores is excluded, and the exclusion rate is not more than 15% of the whole well sample.
[0062] The existing problems of the electrical property fitting method in the industry are: first, the model element innovation is mainly used, that is, the fitting parameters are increased or changed to improve the precision; second, the multiple model regression coefficient is an empirical constant, that is, a correction coefficient, and the correction coefficient has data source pertinence, so that the model application effect is not good. The existing technology of the geological parameter electrical property calculation model does not have research on improving the fitting precision from the electrical property value method, and there is no professional calculation method for the multiple model regression coefficient.
[0063] Compared with the above background art, the present application can have the following beneficial effects:
[0064] The method for improving the prediction precision of the geological parameter electrical multiple model of the present application has the following advantages: first, the single rhythm subdivided layer standard is established to standardize the target layer of modeling and model prediction; second, the monomial correlation coefficient of the geological parameter is improved by changing the electrical property value method; and third, the monomial fitting model fitting goodness proportion of each parameter is used as a parameter weight matching coefficient, and thus the geological parameter electrical multiple fitting model is established, and the precision is greatly improved.
[0065] The electrical property value is extracted by using the data point electrical property average method, and the nuclear magnetic logging is a logging while drilling nuclear magnetic resonance logging, the object detected by the nuclear magnetic logging is a thick sandstone layer combined with multiple rhythms, and the thick sandstone layer may further comprise a certain thickness of mudstone interlayer or calcareous interlayer, the electrical property of the mudstone interlayer or calcareous interlayer is greatly different from that of the sandstone, and the data point electrical property average method can fully reflect the electrical property characteristics; the data point electrical property average method is used, and the nuclear magnetic data modeling of the layer segment with less rhythm or less interlayer in the detected layer segment is optimized, and the fitting precision is greatly improved.
[0066] The application uses the sensitive parameter to establish a unary model by using the electric value innovation method (the shape geometric mean method is used in the conventional logging, the shape extreme value method is used in the sealed coring well, and the sampling point value average method is used in the nuclear magnetic logging), and then uses the statistical least square method, the mathematical equation method, the correction coefficient method and the fitting goodness weight coefficient method to establish a multi-element model.
[0067] The method for improving the prediction accuracy of the electric multi-element model of the geological parameter improves the prediction accuracy of the porosity model to more than 92%, improves the prediction accuracy of the irreducible water saturation model to more than 90%, and reduces the relative error of the permeability model from one order of magnitude to less than 50%. BRIEF DESCRIPTION OF DRAWINGS
[0068] Figure 1 The technical roadmap for improving the prediction accuracy of the electric multi-element model of the geological parameter;
[0069] Figure 2 The well logging interpretation result map of the research area C of the embodiment;
[0070] Figure 3 The single point and electric extreme value correlation graph of the rock sample of the embodiment;
[0071] Figure 4 The single rhythm average value and electric shape geometric mean value correlation graph of the rock sample of the embodiment;
[0072] Figure 5 The microelectrode and irreducible water saturation correlation graph of the embodiment;
[0073] Figure 6 The density and irreducible water saturation correlation graph of the embodiment;
[0074] Figure 7 The natural potential and irreducible water saturation correlation graph of the embodiment;
[0075] Figure 8 The gamma and irreducible water saturation correlation graph of the embodiment;
[0076] Figure 9 The microelectrode and irreducible water saturation correlation graph of the embodiment;
[0077] Figure 10 The density and irreducible water saturation correlation graph of the embodiment;
[0078] Figure 11 The natural potential and irreducible water saturation correlation graph of the embodiment;
[0079] Figure 12 Gamma and irreducible water saturation correlation graph for the embodiment of the present application;
[0080] Figure 13 Acoustic wave and irreducible water saturation correlation graph for the embodiment of the present application;
[0081] Figure 14 Permeability and irreducible water saturation correlation graph for the embodiment of the present application. DETAILED DESCRIPTION
[0082] In order to make the purpose, technical scheme and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the drawings.
[0083] As shown in the figure, a method for improving the prediction accuracy of an electrical multi-element model of a geological parameter comprises the following steps: Figure 1
[0084] S1: sealed coring is performed in a target work area to detect the geological parameters required for the reservoir;
[0085] S2: nuclear magnetic logging is performed in the target work area to detect the geological parameters required for the reservoir;
[0086] The geological parameters detected in the steps S1 and S2 include porosity, permeability, irreducible water saturation and the like.
[0087] S3: conventional logging is performed in the target work area to collect logging data required for the reservoir;
[0088] The collected logging data includes natural gamma logging curve GR, density logging curve DEN, micro-electrode logging curve RM, acoustic logging curve HAC, natural potential logging curve SP and the like.
[0089] S4: based on the electrical response characteristics of the step S3, the target layer of modeling and model prediction is standardized, and the single rhythm subdivision layer standard and the electrical property value standard of the subdivision layer are established;
[0090] The single rhythm subdivision layer standard is that the 10% micro-potential curve return is taken as the standard of the subdivision layer;
[0091] The electrical property value standard in the single rhythm is that the commonly used electrical property value method is the comparison of the three methods of fluctuation peak value method, peak average method and morphological geometric average value. When the interlayer is obvious, the calculation results of the three methods are similar. When the interlayer is not obvious (the micro-potential curve return is in the range of greater than 5% and less than 14%), the transition type rhythm is caused by lithology change, the morphological geometric average is more consistent with the electrical response characteristics in the logging principle, and the prediction accuracy is also higher.
[0092] S5: based on the obtained geological parameters and logging data, different electrical property value methods are matched through different data sources; the specific method comprises:
[0093] S501: Based on the core data of step 1, in order to improve the fitting accuracy of core detection geological parameters and electrical properties, a single rhythm form extreme value method is used to extract electrical properties;
[0094] The specific steps of the single rhythm form extreme value method for extracting electrical properties include:
[0095] First, the core data of the coring well is classified into a single rhythm;
[0096] Second, the core detection geological parameters are normalized and calculated;
[0097] Third, the core representativeness is calculated;
[0098] Fourth, the reservoir where the core with high representativeness is located is optimized, the rhythm form geometric mean value and the core standard value are extracted for electrical property fitting, the sensitive parameters are optimized, and a unary model is established;
[0099] The optimization method of the sensitive parameters is: according to the descriptive analysis of the logging interpretation principle, the microelectrode is optimized as the permeability sensitive parameter, the microelectrode is the response of the permeability difference according to the amplitude difference of the formation resistivity corresponding to the different detection depths of the micro gradient and the micro potential, the greater the amplitude difference, the better the permeability; the spontaneous potential is optimized as the sensitive parameter, the spontaneous potential is the difference between the formation water and the mud salinity affecting the formation permeability; the density is optimized as the sensitive parameter, the density is the response of the size of the formation pore, the gamma is the response of the size of the small pore of the formation, and the two cooperate with each other to indirectly reflect the permeability; the acoustic wave is optimized as the sensitive parameter, the acoustic logging responds to the compactness of the formation, and can also indirectly reflect the permeability.
[0100] S502: Based on the data of step 2 nuclear magnetic resonance, in order to improve the fitting accuracy of the nuclear magnetic logging detection geological parameters and electrical properties; a multi-rhythm sampling point numerical average method is used to extract electrical properties.
[0101] The method for extracting electrical properties by the multi-rhythm sampling point numerical average method includes: a rhythm form geometric average method, an extreme value method and a data point electrical property average method.
[0102] The data point electrical property average method is used, the nuclear magnetic logging is the nuclear magnetic resonance logging while drilling, the object detected is a thick sandstone layer combined by multiple rhythms, which may also include a certain thickness of mudstone interlayer or calcareous interlayer, the electrical properties of which are quite different from those of the sandstone, the data point electrical property average method can fully reflect the electrical properties; in order to greatly improve the fitting accuracy, the data point electrical property average method is used, and the nuclear magnetic data modeling of the layer section with less rhythm or less interlayer in the detection layer section is optimized
[0103] S6: Determine the weight coefficients of each parameter;
[0104] Methods for determining the weight coefficients of each parameter include: mathematical equation solving, statistical least squares method, correction coefficient method, and goodness-of-fit weight coefficient method.
[0105] The calculation process of the goodness-of-fit weighting coefficient method is as follows:
[0106] Obtaining the goodness of fit R 2 The goodness of fit R 2 This represents the degree of influence of the independent variable on the dependent variable;
[0107] The ratio of the univariate goodness of fit of the independent variable to the sum of the goodness of fits is determined as the weight coefficient of the multiple regression.
[0108] The regression formula for the goodness-of-fit weight coefficients is as follows:
[0109] D=(R 2 / Σ(R 2 );
[0110] In the formula: D is the multiple regression coefficient, R 2 The goodness of fit of the univariate model for each parameter.
[0111] S7: Accuracy evaluation of multivariate fitting models for geological parameters; specific methods include:
[0112] S701: Based on the univariate fitting model of each parameter, the coefficients are matched by the goodness-of-fit weighting coefficient method to establish a multivariate model.
[0113] S702: Evaluate the multivariate model according to the model evaluation criteria.
[0114] The model evaluation criteria include:
[0115] (1) Core data used for model evaluation is not a modeling data source;
[0116] (2) The model is applied to a single prosodic subdivision layer;
[0117] (3) Core samples are taken from wells that are traversed for evaluation; core samples are not selected subjectively.
[0118] (4) Exclude error layers caused by small number of core samples or low representativeness, with an exclusion rate not exceeding 15% of all well samples.
[0119] S703: Evaluation of model accuracy using different regression coefficients. The regression coefficient method with the highest model accuracy is selected to evaluate the model accuracy.
[0120] The accuracy of the model is evaluated by using the above four different regression coefficient methods (mathematical equation solving method, statistical least squares method, correction coefficient method, and goodness-of-fit weight coefficient method).
[0121] The fitting degree weight coefficient method is compared with four different regression coefficient methods, and the fitting degree weight coefficient method has the highest permeability prediction accuracy, and thus a breakthrough is achieved in improving the permeability interpretation accuracy.
[0122] Embodiment 1
[0123] The technical process for improving the prediction accuracy of the electrical property multi-element model of the geological parameter is specifically described by taking a target work area as an example, and includes the following steps:
[0124] Step 1, sealed coring is performed in the target work area to detect the geological parameters such as porosity and permeability.
[0125] The core can be analyzed by experiments to obtain the geological parameters such as porosity, permeability and irreducible water saturation (see Table 1).
[0126] Table 1 Core detection data table of well A in the study area
[0127]
[0128] Step 2, nuclear magnetic logging is performed in the target work area to detect the geological parameters such as porosity and permeability.
[0129] The nuclear magnetic resonance logging is a new logging technology, and is not affected by the rock skeleton minerals, and can calculate the parameters such as porosity, permeability and irreducible water saturation (see Table 2).
[0130] Table 2 Nuclear magnetic logging interpretation data table of well B in the study area
[0131]
[0132] Step 3, based on the electrical response characteristics, a single rhythm subdivision layer standard and an electrical property value standard of the subdivided layer are established.
[0133] Due to the heterogeneity of the reservoir, the logging curve is composed of a series of fluctuation peaks, each fluctuation peak is a rhythm, represents a sand-mud alternation process, and corresponds to a group of different geological parameters. Therefore, whether it is the process of applying the data source modeling or the process of applying the model prediction, the division standard and the value method of the target layer are important, and the subdivision layer standard is not unified, which will cause the core single-point modeling to be applied to the heterogeneous multi-point reservoir, or the nuclear magnetic logging multi-rhythm modeling to be applied to the single-rhythm interpretation, and thus the model fitting accuracy is high, but the actual application accuracy is low. The model application accuracy will also decrease due to the different value methods of the target layer.
[0134] Therefore, the subdivision layer standard is established, the micro-potential curve return of 14% is taken as the standard of the subdivision layer, and the electrical property value is extracted by the morphological geometric mean method, so as to ensure the unity of the single-rhythm modeling and the single-rhythm prediction, and lay a foundation for ensuring the model accuracy.
[0135] Figure 2 The well logging interpretation results of well C in the study area are shown in Fig. 1. Figure 2 For example, the 0.9m layer in the middle should be subdivided into two layers at 0.6m, the bottom 0.3m layer is obviously reduced in acoustic wave, which is calcareous sandstone layer, and the porosity and permeability are reduced. The parameter prediction accuracy of the subdivided layer is reduced due to the error. After the layer is subdivided, the relative error of permeability prediction of the upper 0.6m layer is reduced by 5.7%, and the relative error of permeability prediction of the lower 0.3m layer is reduced to 17%(Table 3).
[0136] The standard of electrical property value within a single rhythm: The commonly used methods for electrical property value are peak fluctuation method, peak average method and geometric average of shape. When the interlayer is obvious, the calculation results of the three methods are similar. When the interlayer is not obvious and the transition type rhythm is caused by lithology change, the geometric average of shape is more consistent with the electrical property response characteristics in logging principle, and the prediction accuracy is also higher.
[0137] For example, the 0.9m layer in the middle should be subdivided into two layers at 0.6m, the bottom 0.3m layer is obviously reduced in acoustic wave, which is calcareous sandstone layer, and the porosity and permeability are reduced. The parameter prediction accuracy of the subdivided layer is reduced due to the error. After the layer is subdivided, the relative error of permeability prediction of the upper 0.6m layer is reduced by 5.7%, and the relative error of permeability prediction of the lower 0.3m layer is reduced to 17%(Table 3). Figure 2 For example, the 0.9m layer in the middle should be subdivided into two layers at 0.6m, the bottom 0.3m layer is obviously reduced in acoustic wave, which is calcareous sandstone layer, and the porosity and permeability are reduced. The parameter prediction accuracy of the subdivided layer is reduced due to the error. After the layer is subdivided, the relative error of permeability prediction of the upper 0.6m layer is reduced by 5.7%, and the relative error of permeability prediction of the lower 0.3m layer is reduced to 17%(Table 3).
[0138] Table 3 Influence of 0.9m layer subdivision standard on geological parameter interpretation accuracy of well C in the study area
[0139]
[0140] Step 4, different data sources match different electrical property value methods.
[0141] 4.1, single rhythm shape extreme value method is used to extract electrical property value, and the fitting accuracy of core detection geological parameters and electrical property is improved.
[0142] Rock sample is a rock specimen collected at a certain depth point of the reservoir, which is used for indoor various reservoir parameter inspection and experimental analysis. Its advantages are high inspection accuracy, and its disadvantages are poor representativeness in heterogeneous reservoirs. The current application of core data is still mainly subjective point selection, which has poor representativeness.
[0143] This project takes single rhythm as a unit, standardizes the core data within each rhythm, normalizes the core inspection data, and optimizes the reservoir with high representativeness to participate in model fitting with 0.1m interval one rock core as a standard. This is actually the standardization of dependent variables in mathematical model(see Table 4).
[0144] The specific steps include: 1. classifying the core data of the coring well into a single rhythm; 2. normalizing the core detected geological parameters; 3. calculating the core representativeness; 4. selecting the core in the reservoir with high representativeness, extracting the geometric mean value of the rhythm form, and performing electrical fitting with the core standard value to select sensitive parameters and establish a unary model.
[0145] Table 4 shows the dependent variable and independent variable standardization processing table when modeling with core data sources
[0146]
[0147] Taking porosity as an example, the original core single-point value is fitted with the electrical extreme value method, Figure 3 The fitting relationship curve of the rock sample single point and the electrical extreme value is shown in FIG. 2. Figure 3 It can be seen that the fitting coefficient R 2 of porosity and spontaneous potential is only 0.2069 Figure 3 ); Figure 4 The correlation fitting curve of the average value of the core in the single rhythm and the geometric mean value of the electrical form is shown in FIG. 3. In the figure, the fitting coefficient R 2 of porosity and spontaneous potential is increased to 0.4086 Figure 4 .
[0148] 4.2, the electrical value is extracted by the multi-rhythm sampling point value average method, and the detection reservoir and electrical fitting precision of the nuclear magnetic logging is improved;
[0149] Figure 5 FIG. 4 is a correlation graph of microelectrode and irreducible water saturation established by the rhythm form extreme value method applied in the embodiment; Figure 6 FIG. 5 is a correlation graph of density and irreducible water saturation established by the rhythm form extreme value method applied in the embodiment; Figure 7 FIG. 6 is a correlation graph of spontaneous potential and irreducible water saturation established by the rhythm form extreme value method applied in the embodiment; Figure 8 FIG. 7 is a correlation graph of gamma and irreducible water saturation established by the rhythm form extreme value method applied in the embodiment;
[0150] Figure 9 FIG. 8 is a correlation graph of microelectrode and irreducible water saturation established by the data point electrical average method applied in the embodiment; Figure 10 FIG. 9 is a correlation graph of density and irreducible water saturation established by the data point electrical average method applied in the embodiment; Figure 11 FIG. 10 is a correlation graph of spontaneous potential and irreducible water saturation established by the data point electrical average method applied in the embodiment; Figure 12 FIG. 11 is a correlation graph of gamma and irreducible water saturation established by the data point electrical average method applied in the embodiment;
[0151] Nuclear magnetic resonance logging, also known as nuclear magnetic resonance logging while drilling, detects thick sandstone layers with multiple rhythmic combinations, which may also contain mudstone interlayers or calcareous interlayers of a certain thickness. The electrical properties of these layers differ significantly from those of sandstone, and neither the rhythmic morphology geometric mean method nor the extreme value method can fully reflect their electrical characteristics. Figures 5-8 This project innovated the data point electrical averaging method and selected NMR data from segments with fewer rhythms or interlayers within the detection layer for modeling, significantly improving the fitting accuracy. Figures 9-12 ).
[0152] Table 5 compares the fitting accuracy of univariate models using different electrical property values (fitting of bound water saturation and electrical property correlation). Taking the correlation between bound water saturation and electrical property as an example, the accuracy of the univariate fitting model is significantly improved.
[0153] Table 5
[0154]
[0155] Step 5: Determine the multiple regression coefficients using the goodness-of-fit weighting coefficient method;
[0156] Through the innovation of different value selection methods for the three data sources mentioned above, the accuracy of univariate fitting of reservoir parameters has been greatly improved. However, the key to improving the calculation accuracy of multivariate fitting models lies in the calculation method of multivariate regression coefficients.
[0157] This paper compares three common methods for calculating regression coefficients: mathematical equation solving, statistical least squares, and correction coefficient method, and innovates the goodness-of-fit weight coefficient method.
[0158] ① Mathematical Equation Method
[0159] Transform the univariate regression equation of each parameter into the form of x and y variables: The mathematical equation method uses matrices to obtain multivariate equations and substitutes the calculation results into the univariate regression equation.
[0160]
[0161] Find the multiple regression coefficient A by calculating y = (1 / 5) * (a1x1 + a2x21 + a3x3 + a4x41 + a5x5).
[0162] The equation for the multiple regression coefficients is A = a*(1 / n);
[0163] In the formula: a is the univariate correlation coefficient of each parameter, A is the multivariate regression coefficient, and n is the number of independent variables.
[0164] This method distributes the weights of the influence of each independent variable on the dependent variable equally, increasing the influence of parameters with weak correlation and decreasing the influence of parameters with strong correlation, resulting in a relatively high explanatory error.
[0165] ②Counting the least square method
[0166]
[0167] Then the multiple regression coefficient B = R * b, B is the multiple regression coefficient, and b is the one-parameter regression coefficient. This method accumulates the independent influence of each parameter without weight sum limitation, and the calculation result is very large.
[0168] ③Correction coefficient method
[0169] The correction coefficient method is the most commonly used method in the industry, that is, the fitting error is matched with the correction coefficient of each parameter of the model.
[0170] If the data source Y = 80 and the one-parameter model fitting result y = 100, then the multiple regression coefficient C ≈ Y / y ≈ 0.8, the correction coefficient method is high in data source fitting accuracy, but the prediction error is large in actual application.
[0171] ④Fitting goodness weight coefficient method
[0172] The above three methods have certain defects, and the fitting goodness weight coefficient method is created by the present application. The fitting goodness R 2 represents the influence degree of the independent variable on the dependent variable, and the one-parameter fitting goodness / fitting goodness sum is used as the multiple regression weight coefficient.
[0173] Regression coefficient formula:
[0174] D = (R 2 / Σ(R 2 );
[0175] In the formula, D is the multiple regression coefficient, R 2 is the one-parameter model fitting goodness.
[0176] Taking the irreducible water saturation as an example, the calculation of the multiple weight coefficient is shown in Table 6.
[0177] Table 6 Multiple model weight coefficient calculation table
[0178]
[0179] Step 6, the accuracy of the geological parameter electrical multiple fitting model is evaluated.
[0180] Taking the permeability as an example, the permeability is directly related to the low fitting accuracy of the electrical property, and the existing technology of the permeability model is commonly used with porosity as the independent variable, and the correlation between porosity and electrical property is established. Because the total porosity has low one-parameter correlation with the electrical property, the permeability interpretation accuracy is low, and the prediction accuracy is one order of magnitude.
[0181] The embodiment is based on core and nuclear magnetic logging data, and establishes a permeability electrical model by using bound water saturation as medium.
[0182] Figure 13 The model of the embodiment of the application is related to sound wave and bound water saturation. The bound water saturation electrical one-dimensional fitting equation ( Figures 9-13 ) is brought into the permeability and bound water saturation one-dimensional fitting model ( Figure 14 ), and a permeability electrical prediction mathematical model is obtained:
[0183] Swi=(0.5085 / (0.5085+0.5739+0.6217+0.3417+0.0522))*(82.543e 313*RM )
[0184] +(0.5739 / (0.5085+0.5739+0.6217+0.3417+0.0522))*(0.0241*DEN 9.3805 )+(0.6217 / (0.5085+0.5739+0.6217+0.3417+0.0522))*(81.508*e -0.024*SP ))+(0.3417 / (0.5085+0.5739+0.6217+0.3417+0.0522))*(0.0373*GR 1.6331 )+(0.0522 / (0.5085+0.5739+0.6217+0.3417+0.0522))*((-0.2522)*HAC+136.63);
[0185] K=(-540.7)*ln(Swi)+2337;
[0186] After integration:
[0187] K=(-540.7)*ln((20e 313*RM )+(0.0066DEN 9.3805 )+(24.15e -0.024*SP )
[0188] +(0.006GR 1.6331 )+(-0.006HAC+3.4))+2337;
[0189] In the formula, Swi is bound water saturation, and K is permeability.
[0190] The sensitive parameters are used to establish a single model by using the innovative method of electrical value selection (the shape geometric mean method is used in conventional logging, the shape extreme value method is used in sealed coring well, and the sampling point value average method is used in nuclear magnetic logging). Then, the statistical least square method, the mathematical equation method, the correction coefficient method and the fitting degree weight coefficient method are used to establish multiple models. The precision evaluation of the multiple models of the electrical property of permeability is shown in Table 7. According to the core evaluation of 1515 blocks in four wells, the relative error of the predicted value is 1339.2% when the multiple regression coefficients are calculated by the mathematical equation method; the relative error of the predicted value is 631% when the regression coefficients are calculated by the correction coefficient method; the relative error of the predicted value exceeds 2000% when the multiple regression coefficients are calculated by the statistical least square method, which is not compared here; and the relative error of the predicted value is reduced to less than 50% when the multiple regression coefficients are calculated by the fitting degree weight coefficient method, which realizes a major breakthrough in improving the precision of the permeability interpretation.
[0191] Table 7 Precision evaluation of the multiple models of the electrical property of permeability in the research area
[0192]
[0193] Those skilled in the art will understand that the embodiments described herein are for the purpose of helping the reader to understand the implementation method of the present application and should be understood as the protection scope of the present application not being limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspirations disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the protection scope of the present application.
Claims
1. A method for improving the prediction accuracy of a multi-parameter electrical model of a geological parameter, characterized in that: It comprises the following steps: S1: sealed coring in the target work area to detect the required geological parameters of the oil reservoir; S2: nuclear magnetic logging in the target work area to detect the required geological parameters of the oil reservoir; S3: conventional logging in the target work area to collect the required logging data of the oil reservoir; S4: based on the electrical response characteristics of step S3, standardizing the target layer of modeling and model prediction, establishing single rhythm subdivision layer standards and electrical property value standards of the subdivision layer; S5: based on the obtained geological parameters and logging data, matching different electrical property value methods through different data sources; S6: determining the weight coefficients of each parameter; S7: precision evaluation of the electrical property multi-element fitting model of the geological parameters.
2. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 1, characterized in that: The detected geological parameters include porosity, permeability and irreducible water saturation; the collected logging data include natural gamma ray logging curve GR, density logging curve DEN, micro-electrode logging curve RM, acoustic logging curve HAC and spontaneous potential logging curve SP.
3. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 1, characterized in that: The single rhythm subdivision layer standard of step S4 is that the micro-potential curve return of 14% is used as the standard of the subdivision layer.
4. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 1, characterized in that: In the establishment of the single rhythm electrical property value standard of step S4, the electrical property value method includes three methods of fluctuation peak value method, peak average method and morphological geometric average value method.
5. The method for improving the prediction accuracy of the electrical multi-parameter model of the geological parameter according to claim 4, characterized in that: The electrical property value method is preferably the morphological geometric average value method. When the electrical property value is in the obvious interlayer, the three methods of fluctuation peak value method, peak average method and morphological geometric average value method are compared, and the calculation results are similar. When the electrical property value is in the transition type rhythm caused by lithology change and the interlayer is not obvious, the morphological geometric average value method is selected, which is more consistent with the electrical property response characteristics in the logging principle and has higher prediction accuracy.
6. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 5, characterized in that: The feature that the interlayer is not obvious is that the micro-potential curve return is in the range of greater than 5% and less than 14%.
7. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 1, characterized in that: The step S5 of matching different electrical property value methods through different data sources comprises: S501: based on the core data of step S1, the single rhythm morphological extreme value method is used to extract the electrical property value to improve the fitting accuracy of the core detected geological parameters and electrical property; S502: based on the data of step S2, the multi-rhythm sampling point value average method is used to extract the electrical property value to improve the fitting accuracy of the nuclear magnetic logging detected geological parameters and electrical property.
8. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 7, characterized in that: The specific steps of the single rhythm morphological extreme value method for extracting the electrical property value comprise: Firstly, the core data of the coring well is classified into a single rhythm; Secondly, the core detected geological parameters are normalized calculated; Thirdly, the core representativeness is calculated; Fourthly, the reservoir where the core with high representativeness is located is optimized, the rhythm morphological geometric average value and the core standard value are extracted for electrical property fitting, the sensitive parameters are optimized, and a unary model is established.
9. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 8, characterized in that: The specific method of optimizing the sensitive parameters and establishing the unary model comprises: Optimizing the sensitive parameters and the geological parameters to make a scatter plot; Then, a trend line is added, a trend line regression mathematical formula is set according to the principle of maximum correlation coefficient, and thus a unary model is established; The formula can be linear, power function, exponential or logarithmic function; The preferred method for the sensitive parameters is as follows: Based on descriptive analysis of well logging interpretation principles, the microelectrode is preferably a permeability-sensitive parameter. The microelectrode responds to the permeability difference based on the difference in formation resistivity amplitude corresponding to different detection depths of microgradients and micropotentials; the larger the amplitude difference, the better the permeability. Spontaneous potential is preferably a sensitive parameter, as it affects formation permeability based on the difference in formation water and mud salinity. Density is preferably a sensitive parameter, as density responds to the size of formation pores, and gamma responds to the size of formation micropores; the two work together to indirectly reflect permeability. Acoustic logging is preferably a sensitive parameter, as acoustic logging responds to the density of the formation and can also indirectly reflect permeability.
10. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 7, characterized in that: The methods for extracting electrical values using the multi-rhythm sampling point numerical averaging method include: rhythmic morphological geometric average method, extreme value method, and data point electrical averaging method.
11. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 10, characterized in that: The preferred method for extracting electrical values using the multi-rhythm sampling point numerical averaging method is the data point electrical averaging method. The specific method of the data point electrical averaging method is as follows: read an electrical value at certain intervals from the logging curve, and read all the electrical values. Based on all the electrical values read, calculate the average value of all the above electrical values; And / or, If the interval between the logging curves is 0.05m, then the total number of electrical values read is: layer thickness / 0.
05.
12. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 1, characterized in that: Step S6 determines the weight coefficients of each parameter, and the methods include: mathematical equation solving method, statistical least squares method, correction coefficient method and goodness-of-fit weight coefficient method.
13. The method for improving the prediction accuracy of the electrical multi-parameter model of geological parameters according to claim 1, characterized in that: The method for evaluating the accuracy of the geological parameter electrical multivariate fitting model in step S7 includes: S701: Based on the univariate fitting model of each parameter, the coefficients are matched by the goodness-of-fit weighting coefficient method to establish a multivariate model. S702: Evaluate the multivariate model according to the model evaluation criteria.
14. The method for improving the prediction accuracy of the electrical multi-parameter model of the geological parameter according to claim 12 or 13, characterized in that: The goodness-of-fit weighting coefficient method specifically includes: Goodness of fit R 2 , the goodness of fit R 2 represents the degree of influence of the independent variable on the dependent variable; The ratio of the univariate goodness of fit of the independent variable to the sum of the goodness of fits is determined as the weight coefficient of the multiple regression.
15. The method for improving the prediction accuracy of the electrical multi-parameter model of the geological parameter according to claim 14, characterized in that: The regression formula for the goodness-of-fit weight coefficients is as follows: D = (R 2 / ∑(R 2 ); where: D is the multiple regression coefficient, R 2 is the goodness of fit of the one-parameter model for each parameter.
16. The method for improving the prediction accuracy of the electrical multi-parameter model of the geological parameter according to claim 13, characterized in that: Based on the univariate fitting model of each parameter, the specific methods for establishing a multivariate model by matching coefficients using the goodness-of-fit weighting coefficient method include: Assuming that each parameter univariate model is linear, power function, logarithm, exponential function, the relationship is respectively: y=a1x1+b1; y=a2x2 0.5507 ; y=a3ln(x)-b3; y=a4e b4x Multiple regression formula is: D1*(a1x1+b1)+D2*(a2x2 0.5507 )+D3*(a3ln(x)-b3)+D4*(a4e b4x ) In the formula: a1 is the coefficient of a univariate function, and b1 is the constant of a univariate function; a2 represents the coefficients of the power function, and b2 represents the constant of the power function. a3 is the coefficient of the logarithmic function, and b3 is the constant of the logarithmic function; a4 represents the coefficients of the exponential function, and b4 represents the constant of the exponential function. D1 is the univariate regression coefficient; D2 is the power function regression coefficient; D3 is the logarithmic function regression coefficient; D4 is the exponential function regression coefficient.
17. The method for improving the prediction accuracy of the electrical multi-parameter model of the geological parameter according to claim 13, characterized in that: The model evaluation criteria include: Core data used for model evaluation are not modeling data sources; The model is applied to a single prosodic subdivision layer; Core samples are taken from wells that are traversed for evaluation; core samples are not selected subjectively. Exclude error layers caused by small core samples or low representativeness, with an exclusion rate not exceeding 15% of all well samples.