Coal mine porosity calculation method for improving sequential Gaussian simulation under multi-source data fusion

Abstract

The invention discloses a coal mine porosity calculation method for improving sequential Gaussian simulation under multi-source data fusion. On the basis of logging data preprocessing, logging curves of natural gamma, density, three-lateral resistivity and the like are synthesized, the high-organic-matter stratum is accurately discriminated, a volume physical model is constructed to explain the porosity of the rock stratum, a linear regression model is constructed to explain the porosity of the high-organic-matter stratum, and the porosity of the high-organic-matter stratum is accurately discriminated. After the two results are spliced and combined, abnormal sudden change points on a curve are eliminated through a curve smoothing algorithm, then the porosity curve, rock core data and the water temperature experiment result are normalized and mapped into a geological grid model, variation function analysis and parameter selection work are carried out, improved sequential Gaussian simulation and post-processing are carried out on the basis, and the geological grid model is obtained. And the reliability of the porosity attribute model is improved.

Description

Technical Field

[0001] This invention relates to a method for calculating porosity in a three-dimensional geological model of a coal mine, specifically a method for calculating coal mine porosity using an improved sequential Gaussian simulation based on multi-source data fusion. Background Technology

[0002] In underground coal mining, porosity reflects the ability of coal and rock strata to store fluids and directly affects the occurrence and migration patterns of coalbed methane and the seepage characteristics of groundwater. Accurately describing the spatial variation characteristics of porosity in coal and rock strata has significant engineering and safety value. Data sources that can be used for coal mine porosity attribute modeling mainly include core test results, water injection test results, and coal mine geophysical logging data. Core tests can directly measure porosity with high accuracy, but sampling points are sparse and costs are high, making it difficult to cover areas with complex geological structures. Water injection tests can also indirectly measure the porosity of rock strata, but the results are discrete and sparse, and their reliability is lower than that of core test results. Coal mine geophysical logging technology can indirectly measure the porosity of coal and rock strata, with advantages such as simple measurement methods, no disturbance to the natural structure and stress state of the measured rock mass, and the ability to reflect the variation law of porosity in the vertical direction.

[0003] Traditional methods, when integrating multi-source data to construct porosity models, do not fully consider the differences in confidence levels among data from core experiments, water injection tests, and geophysical well logging. This leads to systematic biases and accumulated uncertainties in the porosity attribute model simulation process, making it difficult to accurately characterize the spatial heterogeneity of porosity under complex geological structures. Summary of the Invention

[0004] To address the shortcomings or deficiencies of existing technologies, this invention provides an improved sequential Gaussian simulation method for calculating coal mine porosity using multi-source data fusion.

[0005] Therefore, the coal mine porosity calculation method based on improved sequential Gaussian simulation under multi-source data fusion provided by this invention includes: Step 1: Based on the well logging data of the area to be tested, determine the rock strata and non-rock strata of the area to be tested; The porosity of the rock formation was calculated using well logging data, and then the porosity-depth curve L1 of the rock formation was plotted. Using the porosity of non-rock core experiments as the dependent variable and the density logging value at the corresponding location as the independent variable, regression analysis was performed to obtain the porosity-depth curve L2 of non-rock layers. By splicing curves L1 and L2, the initial porosity and depth curve L of the area to be measured is obtained. The porosity of the area to be tested was calculated using hydrological experimental data of the area to be tested. Step 2: Perform a normal transformation on curve L and the porosity data calculated using hydrological experimental data to obtain normally distributed multi-source data; Step 3: Map the normally distributed multi-source data to the three-dimensional geological grid model of the area to be measured based on geographic coordinates; Then, the theoretical variation function model of the multi-source data in the three-dimensional geological grid model is obtained; Step 4: Randomly generate simulation paths in the three-dimensional geological grid model obtained in Step 3. The simulation paths consist of blank grids without mapped multi-source data. Then, use Equation (1) to calculate the simulated porosity values ​​of each blank grid on the simulation path: (1) In formula (1): The porosity simulation value for the blank grid to be calculated; N To calculate the number of neighborhood points in the multi-source data of the blank grid to be calculated, N Take a natural number greater than or equal to 2; l n Let the kriging weights be the nth multi-source data neighborhood points in the blank grid to be calculated. n=1,2,3,…,N ; , c nn The variation function value between any two neighboring points of multi-source data; c n0 The value of the variation function between the blank grid and the nth neighboring point of the multi-source data is to be calculated. oh It is the Lagrange multiplicative constant; D n For the first n Correction coefficients for neighborhood points of multi-source data. n=1,2,3,…,N ; D n The value depends on the type of multi-source data for that neighborhood point; , D j for j Correction coefficients for multi-source data, j Choose a, b, or c, where a, b, and c represent multi-source data obtained from well logging, core experiments, and hydrological tests, respectively. data, m j for j The reliability weight for multi-source data is determined by experts based on common sense and experience, and ranges from 0 to 1. A higher value indicates greater reliability. This is the average of the reliability weights for the three types of data; Z ( x n ) is the first n Porosity of neighborhood points in multi-source data; ℓ m A random number that conforms to a standard normal distribution with a mean of 0 and a variance of 1. S Kriging variance; , ; Repeat step 4 until the simulated porosity values ​​of all blank grids in the three-dimensional geological grid model are obtained.

[0006] An alternative approach is to use an average smoothing algorithm to stitch curves L1 and L2 together.

[0007] An alternative approach is to use an arithmetic mean to determine the value of a grid when multiple values ​​appear in a grid during the mapping process in step 3.

[0008] An alternative approach is to use a spherical model or a Gaussian model to obtain the theoretical variation function model of the multi-source data in the three-dimensional geological grid model.

[0009] An alternative approach is to use a truncated wedge search method to obtain the multi-source data neighborhood points of each blank grid.

[0010] Based on well logging data preprocessing, this invention integrates well logging curves such as natural gamma, density, and three-lateral resistivity to accurately identify high organic matter formations. A volume physical model is constructed to explain the porosity of the rock formation, and a linear regression model is constructed to explain the porosity of high organic matter formations. After splicing and combining the two results, an abnormal abrupt change point on the curve is eliminated using a curve smoothing algorithm. Then, the porosity curve, core data, and water temperature experimental results are normalized and mapped into a geological grid model. Variation function analysis and parameter selection are carried out. Based on this, improved sequential Gaussian simulation and post-processing are performed to improve the reliability of the porosity attribute model.

[0011] This invention integrates confidence information from multiple data sources and constructs a porosity attribute model based on an improved sequential Gaussian simulation algorithm suitable for multiple data sources. This model is more faithful to the original data volume and can significantly improve the reliability of the model. Detailed Implementation

[0012] Unless otherwise specified, the scientific and technical terms used in this article are intended for understanding by those skilled in the art.

[0013] In a specific implementation plan, this invention utilizes a well logging interpretation method applicable to coal mines to construct well logging porosity curves. Combined with pumping test results and core test results, a coal mine porosity attribute modeling dataset is constructed. Data processing is performed on the dataset, primarily including normal transformation. Subsequently, the data is mapped onto a three-dimensional geological grid model, and a variogram function is fitted in three-dimensional space. For cases where the confidence levels of the three original data sets differ, the weight matrix is ​​adjusted when calculating the Kriging estimate of the simulated values ​​to integrate the confidence information of the original data, thereby improving the reliability of the porosity attribute model.

[0014] In this invention, well logging data is typically stored as continuous curves in electronic CAD drawings or paper drawings. The horizontal axis of the curve represents the measured values ​​of various well logging types, and the vertical axis represents the depth. Electronic data can export the coordinates of multiple line nodes, and simultaneously extract the endpoint coordinates of various well logging curve ranges, as well as the coordinates of the wellhead and bottom. Paper data requires digitization into raster image data, selection of the coordinate origin, and reading the coordinates of corresponding pixels using the same method. By performing coordinate transformation and scale correction on the curve, and determining a depth point sequence at intervals such as 0.125m, the curve is resampled to obtain a standard well logging curve, which can be used to calculate porosity.

[0015] The non-stratum identification method is as follows: First, the logging data is subjected to depth correction and environmental standardization to eliminate interference from well diameter changes or mud intrusion. The focus is on the abnormal uranium (U) content in the natural gamma spectrum (usually >120 API), because it is significantly positively correlated with the abundance of organic matter. Meanwhile, density logging shows that the organic matter density (approximately 1.0 g / cm³) is significantly lower than that of conventional rock skeletons (2.3-2.7 g / cm³). When the formation density value is below 2.5 g / cm³, it can indicate high organic matter enrichment, but it needs to be verified in conjunction with three lateral resistivity: due to the insulation and low invasion characteristics of kerogen, the organic matter strata show a positive amplitude difference (LLD / LLS>1.2) where the deep lateral resistivity is significantly higher than the shallow lateral resistivity (LLS), and the overall resistivity is higher than that of the surrounding rock. Further, the organic matter enrichment interval is identified by natural gamma-density cross plot (high GR, low DEN area) and resistivity-natural gamma cross plot (high GR accompanied by high resistivity). Finally, the boundary of the organic matter layer is determined by combining the regional geological model to ensure the reliability and geological adaptability of the discrimination results.

[0016] Methods for calculating porosity of rock strata and non-rock strata: First, a volume physical model of the rock strata is constructed, and the complex rock strata are decomposed into three parts: skeleton, mud, and pores. It is assumed that the pores are filled with water. Then, the overall density logging response of the rock strata is described by the linear combination of the three parts.

[0017] Using borehole density logging curves and natural gamma ray curves as data sources, the formation clay content is estimated using the natural gamma ray (GR) curve. Then, combining the density logging (DEN) curve, a response equation is established based on a rock volume physical model. Finally, by substituting response parameters calibrated from the mining area's strata (such as skeleton density), the response equation is calculated. r ma Fluid density r f (etc.) to calculate the porosity of the rock strata.

[0018] For non-rock strata, including low-density coal seams or organic-rich formations, the main rock-forming minerals are no longer quartz / feldspar, thus the volumetric physics model fails in these formations. Based on step two, using the porosity results from core experiments as the dependent variable and the density logging values ​​at the corresponding core locations as the independent variable, regression analysis is performed to construct a linear regression model, thereby calculating the porosity of the non-rock strata.

[0019] In this invention, the porosity results of rock strata and non-rock strata are spliced ​​together. For example, an average smoothing algorithm is used to achieve smooth splicing of curves, and an arithmetic mean of local data is taken by selecting a sliding window of appropriate size to reduce the influence of random noise and discontinuity of curves at the splicing point.

[0020] This invention employs an improved sequential Gaussian simulation, which is based on the Gaussian random field assumption and uses the Kriging method to calculate the local conditional probability distribution function. In this process, the original data is required to follow a normal distribution. For data that does not conform to the Gaussian distribution, a normal transformation is required. Its main purpose is to simplify the calculation by taking advantage of the mathematical characteristics of the Gaussian model: (1) the mean and variance of the conditional distribution can be obtained directly through the Kriging method without a complicated iterative process; (2) samples can be directly drawn from the normal distribution to obtain the closed solution of the local conditional probability distribution function.

[0021] In the specific scheme, the multi-source data after normal transformation are uniformly mapped to a three-dimensional geological grid model as sample points to participate in the simulation. When multiple values ​​appear in a grid during the mapping process, the value of the grid is determined by the arithmetic mean.

[0022] The variogram, based on regionalized variables, reflects the spatial variability of these variables as a function of distance between sample points. In practice, it is difficult for sample point pairs to strictly conform to directional and distance requirements. The truncated wedge search constructs a double-sector wedge search domain based on the primary and secondary directions of the geological structure, with the primary and secondary ranges as axes. A maximum search range is defined, and the wedge opening is limited by azimuth and tolerance angles. Only sample point pairs within this sector are collected, and a suitable hysteresis tolerance is defined; any point pair falling within the hysteresis tolerance range can participate in the calculation. In a specific scheme, the theoretical variogram model can be generated in GOCAD software. In the implementation plan, a spherical model or a Gaussian model can be used to fit the variogram scatter plot, where the independent variable h represents the distance between sample point pairs, and the dependent variable is the variogram value of the point pair.

[0023] This invention, based on the extraction and interpretation of porosity logging data from coal mines, fully explores the sources of porosity data in coal and rock strata, constructing a multi-source dataset for porosity attribute modeling, primarily composed of core data, porosity curves, and water injection experiment results. The dataset is converted to a standard normal distribution and mapped onto a three-dimensional geological grid. A variogram model is fitted to the dataset, and relevant parameters are calculated. Based on the variogram model, an improved sequential Gaussian simulation is conducted on unknown points. During this process, geological knowledge is incorporated to assess the reliability of data from various sources, and a confidence matrix is ​​constructed to digitally express the confidence level. Data fusion during the simulation is achieved through a modified Kriging coefficient matrix. Therefore, this invention not only provides an effective method for constructing coal mine porosity attribute datasets but also integrates reliability information from different data sources during the simulation process, playing a crucial role in improving the reliability of the construction and application of three-dimensional porosity attribute models for coal mines.

Claims

1. An improved sequential Gaussian simulation method for calculating coal mine porosity using multi-source data fusion, characterized in that, The methods include: Step 1: Based on the well logging data of the area to be tested, determine the rock strata and non-rock strata of the area to be tested; The porosity of the rock formation was calculated using well logging data, and then the porosity-depth curve L1 of the rock formation was plotted. Using the porosity of non-rock core experiments as the dependent variable and the density logging value at the corresponding location as the independent variable, regression analysis was performed to obtain the porosity-depth curve L2 of non-rock layers. By splicing curves L1 and L2, the initial porosity and depth curve L of the area to be measured is obtained. The porosity of the area to be tested was calculated using hydrological experimental data of the area to be tested. Step 2: Perform a normal transformation on curve L and the porosity data calculated using hydrological experimental data to obtain normally distributed multi-source data; Step 3: Map the normally distributed multi-source data to the three-dimensional geological grid model of the area to be measured based on geographic coordinates; Then, the theoretical variation function model of the multi-source data in the three-dimensional geological grid model is obtained; Step 4: Randomly generate simulation paths in the three-dimensional geological grid model obtained in Step 3. The simulation paths consist of blank grids without mapped multi-source data. Then, use Equation (1) to calculate the simulated porosity values ​​of each blank grid on the simulation path: （1） In formula (1): The simulated porosity value for the blank grid to be calculated; N To calculate the number of neighborhood points in the multi-source data of the blank grid to be calculated, N Take a natural number greater than or equal to 2; λ n Let the kriging weights be the nth multi-source data neighborhood points in the blank grid to be calculated. n=1,2,3,…,N ; , γ nn The variation function value between any two neighboring points of multi-source data; γ n0 The value of the variation function between the blank grid and the nth neighboring point of the multi-source data is to be calculated. ω It is the Lagrange multiplicative constant; Δ n For the first n Correction coefficients for neighborhood points of multi-source data. n=1,2,3,…,N ; Δ n The value depends on the type of multi-source data for that neighborhood point; , Δ j for j Correction coefficients for multi-source data, j Choose a, b, or c; a, b, and c are divided. This represents multi-source data obtained from well logging, core experiments, and hydrological tests. μ j for j The reliability weight for multi-source data ranges from 0 to 1. This represents the average of the reliability weights for the three types of multi-source data. Z ( x n ) is the first n Porosity of neighborhood points in multi-source data; ℓ m A random number that conforms to a standard normal distribution with a mean of 0 and a variance of 1. S Kriging variance; ， ； Repeat step 4 until the simulated porosity values ​​of all blank grids in the three-dimensional geological grid model are obtained.

2. The coal mine porosity calculation method based on improved sequential Gaussian simulation under multi-source data fusion as described in claim 1, characterized in that, The average value smoothing algorithm is used to stitch curves L1 and L2 together.

3. The coal mine porosity calculation method based on improved sequential Gaussian simulation under multi-source data fusion as described in claim 1, characterized in that, Step 3: When multiple values ​​appear in a grid during the mapping process, the value of that grid is determined by the arithmetic mean.

4. The coal mine porosity calculation method based on improved sequential Gaussian simulation under multi-source data fusion as described in claim 1, characterized in that, The theoretical variation function model of multi-source data in the three-dimensional geological grid model is obtained by using a spherical model or a Gaussian model.

5. The method for calculating coal mine porosity using improved sequential Gaussian simulation based on multi-source data fusion as described in claim 1, characterized in that, The truncated wedge search method is used to obtain the multi-source data neighborhood points of each blank grid.

Images