Fluid mobility prediction method and device based on sequential Gaussian simulation

Abstract

The invention provides a fluid mobility prediction method and device based on sequential Gaussian simulation. According to the technical scheme disclosed by the invention, on one hand, two kinds of data of logging fluidity and earthquake low frequency can be synthesized and calculated at the same time, on the other hand, the two problems of calculation errors caused by inaccurate low-frequency models in inversion and inversion calculation falling into a local optimal solution can be avoided, and the reservoir fluid fluidity can be directly obtained; and the intuition of the prediction result is obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of geophysical exploration in the petrochemical industry, and more particularly, to a method and apparatus for predicting fluid mobility based on sequential Gaussian simulation. Background Art

[0002] Reservoir fluid mobility has good oil and gas reservoir imaging capabilities, and this property provides important guidance for the prediction of high-quality reservoirs. Relying on existing reservoir prediction technologies, currently two methods can be used to obtain reservoir fluid mobility from seismic data. One is to extract this property from seismic low-frequency information using time-frequency analysis methods, and the other is to use inversion algorithms to obtain this property.

[0003] The first approach: Silin et al. (2004, 2006) derived the propagation equation of low-frequency harmonics in a saturated elastic porous medium in the asymptotic regime from the basic seepage theory, obtaining an asymptotic expression for the reflection coefficient of a fluid-saturated reservoir that depends on frequency, i.e., the low-frequency reflection coefficient is proportional to the square root of the product of fluid mobility, fluid density, and seismic signal frequency. Korneev et al. (2004) used this reflection coefficient to image the low-frequency components in actual seismic data to predict the oilfield productivity. Goloshubin et al. (2006) formally introduced the concept of low-frequency imaging attributes based on Silin's low-frequency asymptotic analysis theory and Korneev's research work, verified the relationship between the imaging attributes and actual oil production rate, and applied this attribute to the discrimination of oil-water interfaces and the prediction of reservoir oil production rate. According to the low-frequency asymptotic analysis theory, Goloshubin et al. (2008) further derived the low-frequency imaging attribute as a seismic analysis attribute related to seismic frequency by applying fluid flow properties and scattering mechanisms, and used this attribute to predict the fluid flow capacity and permeability in the reservoir. Dai Shuanghe et al. (2010) extracted the reservoir fluid mobility attribute from actual seismic data and used this attribute to predict the hydrocarbon-bearing property of the reservoir, achieving good results. Cai Hanpeng (2012) systematically discussed the relationships among imaging attributes, reservoir fluid mobility attributes, permeability, and reservoir oil production rate, and verified the direct use of low-frequency information to predict reservoir productivity by combining well logging data and production test data. Chen Xuehua et al. (2012) derived a calculation method for the reservoir fluid mobility attribute using the generalized S transform and provided a theoretical method for determining the dominant frequency using the low-frequency asymptotic analysis theory. Chen Xuehua et al. (2013) combined the "low-frequency shadow technique" with reservoir fluid mobility to identify reservoir fluids and reduced its non-uniqueness and uncertainty. Ren et al. (2013) estimated the reservoir fluid mobility based on frequency-dependent azimuthal AVO and also illustrated the important influence of frequency on fluid mobility. Zhang Shengqiang et al. (2015) derived a calculation method for reservoir fluid mobility based on high-resolution sparse inversion spectral decomposition, improving the resolution of reservoir fluid mobility imaging. Rusakov et al. (2016) applied the reservoir fluid mobility attribute to estimate formation permeability from 3D seismic data.

[0004] The second approach is that the seismic reflection coefficient is the key to inverting reservoir fluid information and reservoir fluid mobility. Since Biot's theory describing the propagation of elastic waves in fluid-saturated porous media (Biot, 1941, 1956, 1957, and 1962) was established, more and more research has been conducted on the derivation of the interface reflection coefficient. For example, Korneev et al. (2004) explained the frequency dependence of reflection in a fluid-saturated porous layer. They observed that in gas- and fluid-saturated porous reservoirs, seismic reflections vary greatly and show a decreasing trend with increasing frequency at low frequencies (15 - 50 Hz). Silin et al. (2006) derived the reflection of seismic waves from a planar interface between two elastic media and found that if one of the media is poroelastic and fluid-saturated, the reflection will become frequency-dependent. Zhao et al. (2014) studied the reflectivity in a diffusive viscous medium and demonstrated that the magnitude of the reflection coefficient in such a medium is not only related to the incident angle and the parameters of the media (such as velocity, density, diffusion, and viscous attenuation), but also closely related to frequency. Qin et al. (2018) studied the frequency dependence of the reflection coefficient of saturated gas-bearing P-waves. They found that the dispersion of the longitudinal wave reflection increases with increasing gas saturation, and the absolute value of the reflection coefficient decreases with increasing frequency. Yin Xingyao et al. and Zhou Dongyong et al. (2019) derived the equations for the reflection coefficient and transmission coefficient at the interface between a porous medium and a viscous fluid, and proposed that the magnitudes of the reflection coefficient and transmission coefficient are affected by frequency, fluid viscosity, and different pore fluid types.

[0005] The first approach mentioned above has the following defects: 1. Time-frequency analysis is limited by time-frequency resolution and it is difficult to accurately extract the fluid mobility information of thin-layer reservoirs; 2. Only relying on seismic low-frequency information, other data sources such as logging data cannot be fully utilized. The second approach has the following defects: 1. The inversion process is easily affected by low-frequency seismic model errors, resulting in calculation errors; 2. The inversion algorithm is prone to falling into local optimal solutions, affecting the calculation results; 3. The inversion does not directly obtain the fluid mobility and post-processing is required to obtain the fluid mobility distribution; 4. The computational amount is large and higher hardware requirements are needed.

[0006] In summary, both of these two existing technologies have certain limitations and are difficult to be satisfactory. Summary of the Invention

[0007] In view of this, the present invention proposes a technical solution for predicting fluid mobility by using the idea of sequential Gaussian algorithm-based stochastic simulation. According to the technical solution of the present invention, on the one hand, it can comprehensively calculate both logging fluid mobility and seismic low-frequency data, and on the other hand, it can avoid the two problems of calculation errors caused by inaccurate low-frequency models and the inversion calculation falling into local optimal solutions in the inversion, and can directly obtain the reservoir fluid mobility, improving the accuracy and intuitiveness of the prediction results.

[0008] According to one aspect of the present invention, a method for predicting fluid mobility based on sequential Gaussian simulation is proposed. The method includes:

[0009] Step 1: Perform a generalized S-transform on the original time-domain seismic signal to obtain seismic frequency data.

[0010] Step 2: Obtain well logging fluid mobility data of the reservoir from well logging data.

[0011] Step 3: Perform a normal transformation on the well logging fluid mobility data and the seismic frequency data respectively to make them both present a normal distribution.

[0012] Step 4: Use the sequential Gaussian simulation algorithm to divide the grid with the target area as the boundary, determine the access path, sequentially obtain the conditional probability distribution function at each grid node, and randomly extract the quantile from the conditional probability distribution function as the simulation value of the grid node.

[0013] Step 5: Perform an inverse normal transformation on the simulation values of all grid nodes to obtain the simulated fluid mobility field of the target area.

[0014] In some embodiments, in Step 1, the original time-domain seismic signal x(t) is subjected to a generalized transformation based on the following formula to obtain seismic frequency data GST(τ, f):

[0015]

[0016] where λ > 0, p > 0, and λ and p are the parameter factors of the generalized S-transform.

[0017] In some embodiments, the method further includes:

[0018] Before Step 3, first check the spatial distribution of the seismic frequency data and the well logging fluid mobility data. If the spatial distribution of the seismic frequency data or the well logging fluid mobility data is uneven, then perform aggregation reduction processing and / or smoothing extrapolation on the seismic frequency data or the well logging fluid mobility data first.

[0019] In some embodiments, in Step 4, the co-Kriging algorithm is applied to obtain the conditional probability distribution function at each grid node.

[0020] In some embodiments, in Step 4, the access path is randomly determined.

[0021] In some embodiments, when calculating the simulation value of each grid node, the method specifically includes:

[0022] a. Reserve a preset number of domain condition data for each grid node, where the domain condition data includes seismic frequency data, logging fluid mobility data, and the simulated values of the grid nodes that have been simulated;

[0023] b. For each grid node, apply the co-Kriging algorithm to determine the parameters of the conditional probability function at this grid node to determine the conditional probability function at this grid node;

[0024] c. Randomly extract a quantile from the conditional probability function at the grid node as the simulated value of this grid node;

[0025] d. Load the newly obtained simulated value of the grid node into the array of the simulated values of the grid nodes that have been simulated;

[0026] e. Along the access path, perform simulation on the next grid node, and repeat the above steps a to d during the simulation process.

[0027] According to another aspect of the present invention, a fluid mobility prediction device based on sequential Gaussian simulation is also proposed. The device includes:

[0028] A seismic frequency data acquisition unit, configured to perform a generalized S transform on the original time-domain seismic signal to obtain seismic frequency data;

[0029] A logging mobility data acquisition unit, configured to obtain the logging fluid mobility data of the reservoir from logging data;

[0030] A normal transformation unit, configured to perform a normal transformation on the logging fluid mobility data and the seismic frequency data respectively to make them both present a normal distribution;

[0031] A sequential Gaussian simulation unit, configured to use the sequential Gaussian simulation algorithm, divide the target area as the boundary into grids, determine the access path, sequentially obtain the conditional probability distribution function at each grid node, and randomly extract a quantile from the conditional probability distribution function as the simulated value of this grid node;

[0032] An inverse normal transformation unit, configured to perform an inverse normal transformation on the simulated values of all grid nodes to obtain the simulated fluid mobility field of the target area.

[0033] In some embodiments, the seismic frequency data acquisition unit performs a generalized transform on the original time-domain seismic signal x(t) based on the following formula to obtain the seismic frequency data GST(τ, f):

[0034]

[0035] where λ > 0, p > 0, and λ and p are the parameter factors of the generalized S transform.

[0036] In some embodiments, the device further includes a spatial distribution preprocessing unit, which is specifically configured to, before performing the normal transformation, first check the spatial distributions of the seismic frequency data and the logging fluid mobility data. If the spatial distribution of the seismic frequency data or the logging fluid mobility data is uneven, perform aggregation reduction processing and / or smooth extrapolation on the seismic frequency data or the logging fluid mobility data first.

[0037] In some embodiments, the co-Kriging algorithm is applied to obtain the conditional probability distribution function of each grid node.

[0038] In some embodiments, the access path is randomly determined.

[0039] In some embodiments, when calculating the simulated value of each grid node, the sequential Gaussian simulation unit is specifically configured to:

[0040] a. Reserve a preset number of neighborhood conditional data for each grid node, where the neighborhood conditional data includes seismic frequency data, logging fluid mobility data, and the simulated values of the grid nodes that have been simulated;

[0041] b. For each grid node, apply the co-Kriging algorithm to determine the parameters of the conditional probability function at this grid node to determine the conditional probability function at this grid node;

[0042] c. Randomly extract a quantile from the conditional probability function at the grid node as the simulated value of this grid node;

[0043] d. Load the newly obtained simulated value of the grid node into the array of the simulated values of the grid nodes that have been simulated;

[0044] e. Perform simulation on the next grid node along the access path, and repeat the above steps a to d in the simulation process.

[0045] According to another aspect of the present invention, an electronic device is further provided, and the electronic device includes:

[0046] A memory storing executable instructions;

[0047] A processor that runs the executable instructions in the memory to implement the fluid mobility prediction method based on sequential Gaussian simulation as described above.

[0048] According to another aspect of the present invention, a computer-readable storage medium is further provided, and the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the fluid mobility prediction method based on sequential Gaussian simulation as described above.

[0049] The technical solution proposed by the present invention has at least the following beneficial effects:

[0050] (1) The low-frequency seismic information and well logging flow rate data can be utilized simultaneously, and fluid mobility can be predicted by integrating the information.

[0051] (2) The influence of model error in the inversion process is avoided through simulation.

[0052] (3) The problem that the inversion algorithm falls into a local optimal solution is also avoided.

[0053] (4) The fluid mobility distribution is directly output, and the prediction result is more intuitive and economical.

[0054] (5) The calculation process is simple and intuitive, with a small amount of calculation and low requirements for hardware.

[0055] (6) Sequential Gaussian simulation takes more random factors into account, and the result is more representative.

[0056] (7) The extraction effect of seismic frequency data can be optimized by adjusting the parameters of the generalized S transform.

[0057] (8) It can be widely applied to the prediction and description of oil and gas reservoirs.

[0058] (9) It helps to improve the efficiency of oil and gas exploration and reduce the development risk.

[0059] The method and device of the present invention have other characteristics and advantages, which will be obvious in the accompanying drawings and the subsequent detailed implementation manners incorporated herein, or will be described in detail in the accompanying drawings and the subsequent detailed implementation manners incorporated herein. These accompanying drawings and detailed implementation manners are used together to explain the specific principles of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] By describing the exemplary embodiments of the present invention in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present invention will become more obvious. Among them, in the exemplary embodiments of the present invention, the same reference numerals generally represent the same components.

[0061] Figure 1 The flowchart of a fluid mobility prediction method based on sequential Gaussian simulation according to an embodiment of the present invention is shown.

[0062] Figure 2 The flowchart of a fluid mobility prediction method based on sequential Gaussian simulation according to an exemplary embodiment of the present invention is shown.

[0063] Figure 3 The structural block diagram of a fluid mobility prediction device based on sequential Gaussian simulation according to an embodiment of the present invention is shown.

[0064] Figure 4Shows a schematic diagram of the channel fluid mobility distribution obtained according to random sequential Gaussian simulation according to an exemplary embodiment of the present invention. Detailed implementation mode

[0065] The preferred embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0066] Example 1

[0067] Figure 1 Shows a flowchart of a fluid mobility prediction method based on sequential Gaussian simulation according to an embodiment of the present invention. As shown in the figure, the method includes steps 1 to 5.

[0068] Step 1, perform a generalized S transform on the original time-domain seismic signal to obtain seismic frequency data.

[0069] In some embodiments, the original time-domain seismic signal x(t) is subjected to a generalized transform based on the following formula to obtain seismic frequency data GST(τ, f):

[0070]

[0071] where λ > 0, p > 0, and λ and p are parameter factors of the generalized S transform.

[0072] Let the short-time Fourier transform of the time-domain seismic signal x(t) be:

[0073]

[0074] It is known from experiments that the window function used directly determines the time-frequency resolution of the STFT. According to the characteristics of the time-frequency uncertainty principle, the time-frequency area of the Gaussian-like window function is the smallest. Then, the window function in the above formula is defined as follows:

[0075]

[0076] where δ is a scale factor that determines the time width of the window function. Let

[0077]

[0078] where λ > 0, p > 0, are parameter factors.

[0079] Then, the scale factor δ is a function of the frequency f. It not only has the ability to adaptively adjust the window width in the time-frequency plane with respect to the frequency, but also has the characteristic of multi-resolution. Combining the above derivations, the expression of the generalized S-transform is obtained as follows:

[0080]

[0081] where λ > 0, p > 0, and λ is also called the adjustment factor.

[0082] The generalized S-transform according to the present invention discusses a new generalized S-transform by introducing two parameter factors λ and p, and modifies the window function of the standard S-transform. It can flexibly adjust the window function according to the frequency distribution characteristics of non-stationary signals, and can meet different time-frequency analysis purposes. It can not only flexibly change the width of the time window according to the characteristics of actual seismic signals, but also make the connection line of the peak points of the window function show diversity by adjusting parameters.

[0083] According to the present embodiment, the extraction effect of seismic frequency data can be optimized by adjusting the parameters of the generalized S-transform.

[0084] When λ = 1, p = 1, the generalized S-transform degenerates into the standard S-transform:

[0085]

[0086] Step 2: Obtain the logging fluid mobility data of the reservoir from logging data.

[0087] In some embodiments, the logging fluid mobility data F of the reservoir can be calculated based on the following formula:

[0088]

[0089] where K is the permeability and η is the viscosity coefficient. The permeability K and the viscosity coefficient η can be directly obtained from logging data.

[0090] Step 3: Perform normal transformation on the logging fluid mobility data and the seismic frequency data respectively to make them both present a normal distribution.

[0091] The theory of sequential Gaussian simulation is based on a Gaussian random field, which is the most classical random function. The biggest characteristic of this model is that the random variables conform to a normal distribution (Gaussian distribution). Before performing sequential Gaussian simulation, it is necessary to test the normality of the original data. If the original data does not conform to a normal distribution, it is necessary to perform a normal transformation on this data to transform it into data with a normal distribution.

[0092] Normal transformation (Normal Transformation) refers to the process of converting non-normally distributed data into normally distributed data. Its main ideas and steps include:

[0093] (1) Check the distribution of the original data. If it does not conform to the normal distribution, a normal transformation is required.

[0094] (2) Calculate the statistical parameters of the original data, such as the mean, variance, etc., to obtain the probability distribution function of the original data and describe its statistical distribution.

[0095] (3) According to the distribution type of the original data, select an appropriate transformation method, such as logarithmic transformation, exponential transformation, Box-Cox transformation, etc.

[0096] (4) Apply the formula of the selected transformation method to calculate the parameter changes required to map the original data to the normal distribution.

[0097] (5) Perform a parameter transformation on the original data so that its statistical distribution satisfies the properties of the normal distribution.

[0098] (6) Plot the histogram and probability density curve of the transformed data to check whether it conforms to the normal distribution.

[0099] Subsequently, the data with a normal distribution can be mapped back to the original values according to the inverse of the transformation parameters.

[0100] In some embodiments, in step (1) above, when checking the spatial distribution of the seismic frequency data and the logging fluid mobility data, if the spatial distribution of the seismic frequency data or the logging fluid mobility data is uneven, the seismic frequency data or the logging fluid mobility data is first subjected to de-clustering processing and / or smoothing extrapolation, so that the spatial distribution of the data tends to be uniform, and then the logging fluid mobility data / seismic frequency data with a uniform distribution is subjected to a normal transformation.

[0101] A brief introduction to de-clustering processing and smoothing extrapolation is given below.

[0102] The basic idea and steps of de-clustering processing are as follows:

[0103] (1) Calculate the statistical parameters of the original data, such as the maximum value, minimum value, mean, variance, etc.

[0104] (2) According to the aggregation situation of the original data distribution, select an appropriate de-clustering method. Common methods include distance de-clustering method and block de-clustering method, etc.

[0105] (3) The distance de-clustering method assigns weights according to the proximity of data points. Points far from the aggregation center obtain greater weights.

[0106] (4) The block de-clustering method divides the space into blocks and assigns weights according to the number of data points within the blocks.

[0107] (5) Calculate the new weights of each data point according to the selected method, and substitute them into the formula for the calculation of data value aggregation reduction.

[0108] (6) After processing, plot the distribution map of the new data and check whether its aggregation degree has been reduced.

[0109] (7) Different parameters can be repeatedly tested to select the optimal effect of aggregation reduction processing.

[0110] In summary, the aggregation reduction processing reduces the extreme value area of the data distribution by the method of weight assignment, making its statistical distribution more uniform.

[0111] Smoothing Interpolation usually refers to a technique that uses the method of function fitting to extrapolate known discrete data to obtain a continuous distribution between data. Its basic principle and steps are as follows:

[0112] (1) Collect the known discrete data points in the target area.

[0113] (2) According to the data distribution, select a suitable function model. Commonly used ones include polynomial functions, cubic spline functions, etc.

[0114] (3) Substitute the selected function model into the known data points for function parameter fitting.

[0115] (4) In the target area, use the fitted function model for data extrapolation.

[0116] (5) The extrapolation result is a smooth continuous distribution field that connects the discrete data points.

[0117] (6) According to actual needs, the function type and parameters can be adjusted to optimize the extrapolation effect.

[0118] (7) The extrapolation result usually generates a certain error and needs to be verified according to the actual situation.

[0119] (8) Smoothing Interpolation is suitable for data distributions with a certain degree of correlation and regularity.

[0120] In summary, Smoothing Interpolation uses function fitting to expand discrete data to obtain a continuous distribution.

[0121] According to this embodiment, through aggregation reduction processing and / or Smoothing Interpolation, the spatially unevenly distributed data can be made to tend to be uniform, making the results of subsequent normal transformation more statistically significant and representative.

[0122] The following briefly introduces the relevant theory of the normality of random functions.

[0123] If the continuous spatial random function Z(u), u ∈ A is the sum of a relatively small number of random functions {Y k (u), u ∈ A, k = 1, Λ, K} with similar spatial distributions, then the spatial distribution of such a random function can be modeled using a multivariate Gaussian random function model. This principle does not lie in the components having the same distribution and number, but in the independence of the components from each other. A is the entire space to be predicted. When normalizing well logging fluid mobility data, {Y k (u), u ∈ A, k = 1, Λ, K} can be regarded as all known well logging fluid mobility data; when normalizing seismic frequency data, {Y k (u), u ∈ A, k = 1, Λ, K} can be regarded as all known seismic frequency data.

[0124] The random function Z(u), u ∈ A follows a multivariate normal distribution, that is, the well logging mobility data of all wells should conform to a multivariate normal distribution if and only if:

[0125] (l) All subsets Z(u) of the random function, are also multivariate normal distributions;

[0126] (2) All linear combinations of the random variable components of Z(u) are univariate normal distributions;

[0127] (3) The covariance function or correlation coefficient is 0, ensuring complete independence;

[0128] (4) Given any one subset, the conditional distribution of any other subset of the random function Z(u) is a multivariate normal distribution.

[0129] After the above normal transformation, for n known data points, the conditional probability distribution function (ccdf) of the random function Z(u 0 ) is normal and has the following characteristics:

[0130] (1) The mean or conditional expectation is consistent with its Kriging estimate;

[0131] (2) The variance or conditional variance is the simple Kriging variance.

[0132] Therefore, for Gaussian simulation, the conditional probability distribution function can be obtained using the Kriging method. Due to the normality of the conditional probability distribution function, the entire simulation process according to the present invention is greatly simplified, and sequentially determining a series of conditional probability distribution functions is simplified to solving a series of Kriging equations.

[0133] Step 4: Using the sequential Gaussian simulation algorithm, divide the target area as the boundary into grids, determine the access path, sequentially obtain the conditional probability distribution function at each grid node, and randomly extract quantiles from the conditional probability distribution function as the simulation value of the grid node.

[0134] In sequential Gaussian simulation, a grid node refers to the grid points into which the entire target area to be simulated and predicted is divided.

[0135] Grid division can be performed by the following method: According to the range of the target area, determine the specific boundary of the simulation and prediction; inside the target area, divide the target area into grids according to the required prediction fineness. The grids can be square grids, triangular grids, etc.; each node of the grid is the position of a discrete point, and the set of these nodes constitutes a continuous grid field.

[0136] A grid node refers to the position of each discrete point that constitutes the simulation grid. It is generated by dividing the entire area, and its distribution density determines the simulation fineness. Those skilled in the art can set the distribution density of the grid nodes according to needs. During sequential Gaussian simulation prediction, each grid node will be accessed according to a certain access path, and the simulation value of the grid node will be extracted. After traversing all the grid nodes, a normal inverse transformation is performed on the entire simulated grid field, and the simulated distribution of the fluid mobility in the target area can be obtained.

[0137] In some embodiments, the access path is randomly determined. The random access path can improve the randomness of the sequential Gaussian simulation and make the results more representative.

[0138] In some embodiments, the co-Kriging algorithm is applied to obtain the conditional probability distribution function of each grid node.

[0139] In sequential Gaussian simulation, different Kriging methods can be used to obtain the conditional probability distribution function according to the different characteristics of the actual data. In the present invention, the inventor adopts the sequential Gaussian co-simulation of multiple variables. Therefore, the sequential Gaussian simulation of sub-variables can be adopted, and the sequential Gaussian co-simulation configured with co-Kriging is selected, and the co-Kriging algorithm is applied to obtain the conditional probability distribution function of each grid node.

[0140] Specifically, in some embodiments, when calculating the simulation value of each grid node:

[0141] a. Reserve a preset number of domain conditional data for each grid node. The domain conditional data includes seismic frequency data, well logging fluid mobility data, and the simulation values of the grid nodes that have been simulated.

[0142] b. For each grid node, apply the co-Kriging algorithm to determine the parameters of the conditional probability function at this grid node, so as to determine the conditional probability function at this grid node;

[0143] c. Randomly extract a quantile from the conditional probability function at the grid node as the simulated value of this grid node;

[0144] d. Load the simulated value of the newly obtained grid node into the array of the simulated values of the already simulated grid nodes;

[0145] e. Simulate the next grid node along the access path, and repeat the above steps a to d in the simulation process.

[0146] According to the embodiment, the simulation results of the entire grid field can be obtained, sequential Gaussian simulation is realized with a small amount of calculation, and a variety of data sources are fully utilized to further improve the randomness of sequential Gaussian simulation.

[0147] Step 5. Perform an inverse normal transformation on the simulated values of all grid nodes to obtain the simulated fluid mobility field of the target area.

[0148] This embodiment uses the idea of random simulation based on the sequential Gaussian algorithm to predict fluid mobility. On the one hand, it can comprehensively calculate the well logging mobility and seismic low-frequency data at the same time. On the other hand, it can avoid the two problems of calculation error caused by inaccurate low-frequency models and the inversion calculation falling into local optimal solutions in the inversion, and can directly obtain the reservoir fluid mobility, and significantly improve the intuitiveness of the prediction results.

[0149] Example 2

[0150] Figure 2The flowchart of a fluid mobility prediction method based on sequential Gaussian simulation according to an exemplary embodiment of the present invention is shown. As shown, in step 201, the generalized S transform is used to extract seismic frequency data; in step 202, well logging fluid mobility data is obtained from well logging data. In step S203, the probability distribution function of the seismic frequency data is obtained. In step 204, the seismic frequency data is subjected to a normal transformation based on this probability distribution function. And in step 205, the Kriging algorithm is used to obtain the conditional probability distribution function of the seismic frequency data after the normal transformation. Similarly, in step 206, the probability distribution function of the well logging fluid mobility data is obtained. In step 207, the well logging fluid mobility data is subjected to a normal transformation based on this probability distribution function. And in step 208, the Kriging algorithm is used to obtain the conditional probability distribution function of the well logging fluid mobility data after the normal transformation. Before the normal transformation, if the fluid mobility data obtained from well logging data is insufficient, the fluid mobility data obtained from seismic data can be used for supplementation. In step 209, the access path of the sequential Gaussian simulation is randomly determined. In step 210, the co-Kriging estimation is used to sequentially obtain the conditional probability distribution function at each grid node, and a quantile is randomly extracted from the conditional probability distribution function as the simulation value of this grid node. In step 211, after it is judged that the simulation value extraction of all grid nodes has been traversed, step 212 is entered, and the normal inverse transformation is performed on the simulated grid field to obtain the simulated distribution of the fluid mobility in the target area.

[0151] For the detailed description of this embodiment, reference can be made to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0152] Example 3

[0153] Figure 3 The structural block diagram of a fluid mobility prediction device based on sequential Gaussian simulation according to an embodiment of the present invention is shown. The device includes a seismic frequency data acquisition unit 301, a well logging mobility data acquisition unit 302, a normal transformation unit 303, a sequential Gaussian simulation unit 304, and an inverse normal transformation unit 305.

[0154] The seismic frequency data acquisition unit 301 is used to perform a generalized S transform on the original time-domain seismic signal to obtain seismic frequency data.

[0155] The well logging mobility data acquisition unit 302 is used to obtain the well logging fluid mobility data of the reservoir from well logging data.

[0156] The normal transformation unit 303 is used to perform normal transformations on the well logging fluid mobility data and the seismic frequency data respectively to make them both present a normal distribution.

[0157] The sequential Gaussian simulation unit 304 is used to perform grid division with the target area as the boundary by using the sequential Gaussian simulation algorithm, determine the access path, sequentially obtain the conditional probability distribution function at each grid node, and randomly extract a quantile from the conditional probability distribution function as the simulation value of the grid node.

[0158] The inverse normal transformation unit 305 is used to perform inverse normal transformation on the simulation values of all grid nodes to obtain the simulated fluid mobility field of the target area.

[0159] In some embodiments, the seismic frequency data acquisition unit 301 performs a generalized transformation on the original time-domain seismic signal x(t) based on the following formula to obtain the seismic frequency data GST(τ, f):

[0160]

[0161] where λ > 0, p > 0, and λ and p are the parameter factors of the generalized S transformation.

[0162] In some embodiments, the device further includes a spatial distribution preprocessing unit, which is specifically configured to, before performing the normal transformation, first check the spatial distributions of the seismic frequency data and the well logging fluid mobility data. If the spatial distribution of the seismic frequency data or the well logging fluid mobility data is uneven, then perform decimation processing and / or smooth extrapolation on the seismic frequency data or the well logging fluid mobility data first.

[0163] In some embodiments, the co-Kriging algorithm is used to obtain the conditional probability distribution function at each grid node.

[0164] In some embodiments, the access path is randomly determined.

[0165] In some embodiments, when calculating the simulation value of each grid node, the sequential Gaussian simulation unit 304 is specifically configured to:

[0166] a. Reserve a preset number of domain conditional data for each grid node, where the domain conditional data includes seismic frequency data, well logging fluid mobility data, and the simulation values of the grid nodes that have been simulated;

[0167] b. For each grid node, apply the co-Kriging algorithm to determine the parameters of the conditional probability function at the grid node to determine the conditional probability function at the grid node;

[0168] c. Randomly extract a quantile from the conditional probability function at the grid node as the simulation value of the grid node;

[0169] d. Load the newly obtained simulation value of the grid node into the array of the simulation values of the grid nodes that have been simulated;

[0170] e. Simulate the next grid node along the access path, and repeat the above steps a to d during the simulation process.

[0171] For the detailed description of this embodiment, reference may be made to the corresponding descriptions in the foregoing embodiments, which will not be elaborated herein.

[0172] Example 4

[0173] According to another aspect of the present invention, an electronic device is further provided. The electronic device includes:

[0174] A memory storing executable instructions;

[0175] A processor that runs the executable instructions in the memory to implement a fluid mobility prediction method based on sequential Gaussian simulation.

[0176] The method includes the following steps:

[0177] Step 1: Perform a generalized S transform on the original time-domain seismic signal to obtain seismic frequency data;

[0178] Step 2: Obtain the well logging fluid mobility data of the reservoir from well logging data;

[0179] Step 3: Perform a normal transformation on the well logging fluid mobility data and the seismic frequency data respectively to make them both present a normal distribution;

[0180] Step 4: Use the sequential Gaussian simulation algorithm to divide the target area as the boundary into grids, determine the access path, sequentially obtain the conditional probability distribution function at each grid node, and randomly extract the quantile from the conditional probability distribution function as the simulation value of the grid node;

[0181] Step 5: Perform an inverse normal transformation on the simulation values of all grid nodes to obtain the simulated fluid mobility field of the target area.

[0182] In some embodiments, in Step 1, the original time-domain seismic signal x(t) is subjected to a generalized transform based on the following formula to obtain the seismic frequency data GST(τ, f):

[0183]

[0184] where λ > 0, p > 0, and λ and p are the parameter factors of the generalized S transform.

[0185] In some embodiments, the method further includes:

[0186] Before step 3, first check the spatial distribution of the seismic frequency data and the logging fluid mobility data. If the spatial distribution of the seismic frequency data or the logging fluid mobility data is uneven, first perform declustering processing and / or smooth extrapolation on the seismic frequency data or the logging fluid mobility data.

[0187] In some embodiments, in step 4, the cokriging algorithm is applied to obtain the conditional probability distribution function of each grid node.

[0188] In some embodiments, in step 4, the access path is randomly determined.

[0189] In some embodiments, when calculating the simulated value of each grid node, the method specifically includes:

[0190] a. Reserve a preset number of domain condition data for each grid node. The domain condition data includes seismic frequency data, logging fluid mobility data, and the simulated values of the grid nodes that have been simulated.

[0191] b. For each grid node, apply the cokriging algorithm to determine the parameters of the conditional probability function at this grid node to determine the conditional probability function at this grid node.

[0192] c. Randomly extract a quantile from the conditional probability function at the grid node as the simulated value of this grid node.

[0193] d. Load the newly obtained simulated value of the grid node into the array of the simulated values of the grid nodes that have been simulated.

[0194] e. Simulate the next grid node along the access path, and repeat the above steps a to d in the simulation process.

[0195] Specifically, the memory may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0196] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform the desired functions. In an embodiment of the present invention, the processor is used to run the computer-readable instructions stored in the memory.

[0197] For a detailed description of this embodiment, reference may be made to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0198] Example 5

[0199] According to another aspect of the present invention, there is also provided a computer-readable storage medium storing a computer program, which when executed by a processor implements a fluid mobility prediction method based on sequential Gaussian simulation.

[0200] The method includes the following steps:

[0201] Step 1: Perform a generalized S-transform on the original time-domain seismic signal to obtain seismic frequency data;

[0202] Step 2: Obtain well logging fluid mobility data of the reservoir from well logging data;

[0203] Step 3: Perform a normal transformation on the well logging fluid mobility data and the seismic frequency data respectively to make them both present a normal distribution;

[0204] Step 4: Use the sequential Gaussian simulation algorithm to divide the grid with the target area as the boundary, determine the access path, sequentially obtain the conditional probability distribution function at each grid node, and randomly extract the quantile from the conditional probability distribution function as the simulation value of the grid node;

[0205] Step 5: Perform an inverse normal transformation on the simulation values of all grid nodes to obtain the simulated fluid mobility field of the target area.

[0206] In some embodiments, in step 1, the original time-domain seismic signal x(t) is subjected to a generalized transformation based on the following formula to obtain seismic frequency data GST(τ, f):

[0207]

[0208] where λ > 0, p > 0, and λ and p are parameter factors of the generalized S-transform.

[0209] In some embodiments, the method further includes:

[0210] Before step 3, first check the spatial distribution of the seismic frequency data and the well logging fluid mobility data. If the spatial distribution of the seismic frequency data or the well logging fluid mobility data is uneven, then perform clustering reduction processing and / or smoothing extrapolation on the seismic frequency data or the well logging fluid mobility data first.

[0211] In some embodiments, in step 4, the co-Kriging algorithm is applied to obtain the conditional probability distribution function at each grid node.

[0212] In some embodiments, in step 4, the access path is determined randomly.

[0213] In some embodiments, when calculating the simulation value of each grid node, the method specifically includes:

[0214] a. Reserve a preset number of domain condition data for each grid node, where the domain condition data includes seismic frequency data, logging fluid mobility data, and the simulation values of the grid nodes that have been simulated;

[0215] b. For each grid node, apply the co - Kriging algorithm to determine the parameters of the conditional probability function at this grid node to determine the conditional probability function at this grid node;

[0216] c. Randomly extract a quantile from the conditional probability function at the grid node as the simulation value of this grid node;

[0217] d. Load the newly obtained simulation value of the grid node into the array of the simulation values of the grid nodes that have been simulated;

[0218] e. Perform simulation on the next grid node along the access path, and repeat the above steps a to d in the simulation process.

[0219] According to the computer - readable storage medium of the embodiments of the present invention, non - transitory computer - readable instructions are stored thereon. When the non - transitory computer - readable instructions are run by a processor, all or part of the steps of the methods of the various embodiments of the present invention described above are executed.

[0220] The above - mentioned computer - readable storage medium includes but is not limited to: optical storage media (such as CD - ROM and DVD), magneto - optical storage media (such as MO), magnetic storage media (such as magnetic tapes or external hard drives), media with built - in rewritable non - volatile memories (such as memory cards), and media with built - in ROM (such as ROM cartridges).

[0221] Those skilled in the art should understand that in order to solve the technical problem of how to obtain good user experience effects, this embodiment may also include well - known structures such as communication buses and interfaces, and these well - known structures should also be included in the protection scope of the present invention.

[0222] For the detailed description of this embodiment, reference may be made to the corresponding descriptions in the foregoing embodiments, and details will not be repeated here.

[0223] Example 6

[0224] Figure 4The figure shows a schematic diagram of the channel fluid mobility distribution obtained from random sequential Gaussian simulation according to an exemplary embodiment of the present invention, where the red part is the predicted channel distribution. From Figure 4 It can be seen that by using the method of the present invention, the fluid sensitive area can be accurately predicted. The red part is typical channel sandstone deposition. By combining well and seismic data, it not only avoids the calculation error caused by inaccurate low-frequency models in inversion, improves the accuracy of the prediction results, but also can directly obtain the reservoir fluid mobility to improve the intuitiveness of the prediction results.

[0225] In summary, the present invention discloses a fluid mobility prediction method based on sequential Gaussian simulation, which mainly includes the following steps: performing a generalized S transform on the original time-domain seismic signal to obtain seismic frequency data; obtaining the logging fluid mobility data of the reservoir from logging data; using the co-Kriging algorithm, taking the logging fluid mobility data as the main variable and the seismic frequency data as the secondary variable, constructing a sequential Gaussian simulation equation for estimating fluid mobility, and determining the coefficient terms in the sequential Gaussian simulation equation; predicting the fluid mobility distribution in the study area based on the constructed sequential Gaussian simulation equation.

[0226] The present invention has at least the following beneficial effects:

[0227] 1. Effectively comprehensively utilizes two information sources of logging mobility data and seismic frequency data, significantly improving the accuracy of the prediction results;

[0228] 2. By considering the correlation between data through the co-Kriging algorithm, the prediction deviation can be reduced;

[0229] 3. Compared with directly inverting fluid mobility, the solution of the present invention has higher calculation efficiency and is more intuitive;

[0230] 4. Avoids the trouble of time-frequency analysis being limited by time-frequency resolution;

[0231] 5. Can directly obtain the fluid mobility distribution, improving the intuitiveness of the results;

[0232] 6. The flow calculation process is simplified, and the calculation efficiency is significantly improved;

[0233] 7. Can be widely applied to the prediction and description of oil and gas reservoirs;

[0234] 8. Helps to improve the efficiency of oil and gas exploration and reduce the development risk.

[0235] The embodiments of the present invention have been described above. The above description is exemplary and not exhaustive, and is also not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, the practical application, or the technical improvements to the technology in the market, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

Claims

1. A fluid mobility prediction method based on sequential Gaussian simulation, characterized in that: The method comprises: Step 1, performing generalized S transform on the original time-domain seismic signal to obtain seismic frequency data; Step 2, obtaining reservoir logging fluid mobility data from logging data; Step 3, performing normal transformation on the logging fluid mobility data and seismic frequency data respectively so that they both present a normal distribution; Step 4: Use the sequential Gaussian simulation algorithm to divide the target area into grids as the boundary, determine the access path, obtain the conditional probability distribution function at each grid node in turn, and randomly extract quantiles from the conditional probability distribution function as the simulation value of the grid node; Step 5: Perform an inverse normal transformation on the simulated values ​​of all grid nodes to obtain the simulated fluid flow field in the target area.

2. The method according to claim 1, characterized in that: In step 1, the original time-domain seismic signal x(t) is generalized transformed based on the following formula to obtain seismic frequency data GST(τ, f): Among them, λ>0,p>0, λ and p are parameter factors of generalized S transform.

3. The method according to claim 1, characterized in that The method further comprises: Before step 3, first check the spatial distribution of the seismic frequency data and the logging fluid mobility data. If the spatial distribution of the seismic frequency data or the logging fluid mobility data is uneven, first perform de-aggregation processing and / or smoothing extrapolation on the seismic frequency data or the logging fluid mobility data.

4. The method according to claim 1, characterized in that In step 4, the co-kriging algorithm is applied to obtain the conditional probability distribution function of each grid node.

5. The method according to claim 1, characterized in that In step 4, the access path is determined randomly.

6. The method according to claim 1, characterized in that When calculating the simulation value of each grid node, the method specifically includes: a. retaining a preset amount of domain condition data for each grid node, wherein the domain condition data includes seismic frequency data, well logging fluid mobility data, and simulated values ​​of the simulated grid nodes; b. For each grid node, applying the cokriging algorithm to determine the parameters of the conditional probability function at the grid node to determine the conditional probability function at the grid node; c, randomly extract a quantile from the conditional probability function at the grid node as the simulation value of the grid node; d, loading the newly obtained simulation value of the grid node into the array of simulation values ​​of the simulated grid nodes; e. Simulate the next grid node along the access path, and repeat the above steps a to d during the simulation process.

7. A fluid flow prediction device based on sequential Gaussian simulation, characterized in that: The device comprises: A seismic frequency data acquisition unit, used for performing a generalized S transform on the original time-domain seismic signal to acquire seismic frequency data; A well logging fluidity data acquisition unit, used for acquiring well logging fluid fluidity data of the reservoir from the well logging data; A normal transformation unit is used to perform normal transformation on the logging fluid mobility data and the seismic frequency data respectively so that both of them present a normal distribution; A sequential Gaussian simulation unit is used to adopt a sequential Gaussian simulation algorithm, divide the target area into grids as the boundary, determine the access path, sequentially obtain the conditional probability distribution function at each grid node, and randomly extract quantiles from the conditional probability distribution function as the simulation value of the grid node; The inverse normal transformation unit is used to perform inverse normal transformation on the simulation values ​​of all grid nodes to obtain the simulated fluid flow field in the target area.

8. The device according to claim 7, characterized in that When calculating the simulation value of each grid node, the sequential Gaussian simulation unit is specifically used to: a. retaining a preset amount of domain condition data for each grid node, wherein the domain condition data includes seismic frequency data, well logging fluid mobility data, and simulated values ​​of the simulated grid nodes; b. For each grid node, applying the cokriging algorithm to determine the parameters of the conditional probability function at the grid node to determine the conditional probability function at the grid node; c, randomly extract a quantile from the conditional probability function at the grid node as the simulation value of the grid node; d, loading the newly obtained simulation value of the grid node into the array of simulation values ​​of the simulated grid nodes; e. Simulate the next grid node along the access path, and repeat the above steps a to d during the simulation process.

9. An electronic device, characterized in that: The electronic device comprises: A memory storing executable instructions; A processor, wherein the processor runs the executable instructions in the memory to implement the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images