Method and system for mining subterranean reservoir
The method addresses the inefficiencies in determining hydrocarbon reservoir properties by using uncertainty distributions and optimization to enhance simulation and exploitation, improving well operations and production rates.
Patent Information
- Application Number
- CN202280095502.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-02
- Publication Date
- 2024-12-06
AI Technical Summary
Existing methods for determining properties of hydrocarbon underground reservoirs are costly, time-consuming, and inaccurate, particularly when dealing with small data challenges, leading to incomplete or inaccurate simulation and exploitation of these reservoirs.
A method involving determining a first attribute's value using a first uncertainty distribution, a predetermined second attribute value, and a relationship with a third attribute, through optimization operations, allowing for improved simulation and exploitation of reservoir characteristics.
Enables accurate and cost-effective determination of reservoir properties, enhancing exploitation by improving well placement and production rates, and integrating with geomechanical modeling and 4D seismic drive history matching.
Smart Images

Figure CN119096166A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for simulating at least one property of a hydrocarbon subterranean reservoir for use in exploiting the reservoir. The present invention also relates to a system for simulating at least one property of a hydrocarbon subterranean reservoir for use in exploiting the reservoir. Background Art
[0002] Hydrocarbons occur naturally in geological formations beneath the Earth's surface (i.e., hydrocarbon underground reservoirs) and represent a primary energy source. These hydrocarbons can be further processed into a suitable energy form to meet the energy needs of the Earth's population.
[0003] Hydrocarbon underground reservoirs are usually rocks characterized or described by several characteristics or attributes. These characteristics or attributes include seismic, well logging and rock physical attributes, which vary in space (e.g., different spatial locations along the reservoir) and scale (e.g., between coarse and fine scales), and they may be highly dependent on other reservoir attributes. For example, the pressure in the reservoir may depend on the temperature of the reservoir. Other examples of reservoir attributes include the porosity (e.g., non-solid material space) of the reservoir that appears in the geological medium, the permeability (e.g., fluid flow capacity) of the reservoir, and the hydrocarbon saturation (e.g., the volume of hydrocarbons relative to other fluids) of the reservoir. The values of these and other attributes usually must meet certain criteria in order to exploit the reservoir, for example, by well logging. In particular, knowing the value of the attribute helps in decision-making for, for example, improved well location and well operation, and therefore provides cost-effective and economic advantages. Therefore, it is most important to be aware of and know one or more values of these attributes. Knowing one or more values of these attributes at future moments, such as knowing their development over time, has the potential to further improve decision-making.
[0004] In order to determine one or more values of an attribute, logging or sonic logging measurements can be performed, however, these measurements are expensive and limited due to test procedures or operating time constraints and / or subject to inaccuracies inherently associated with the measurements. Therefore, the measured data needs to be corrected when processing the data, which is an exhaustive task. Often, time and / or cost constraints are usually the dominant factor in giving up the acquisition, correction and / or processing of the values of the attributes. In addition, the interpretation of the values of the attributes representing more complex underground reservoirs is difficult and may not be determined. In addition, not all attributes can be captured by measurement, and their values cannot be measured at a future time. For example, dry rock properties are usually measured in core samples or depend on reference descriptions. The best description should come from direct testing, but direct testing is often omitted due to cost.
[0005] It is known to perform reservoir simulation to predict the properties of a reservoir. Specifically, reservoir simulation is a branch of fluid mechanics, and is dedicated to the numerical simulation of multiphase flow through a non-uniform porous medium in the form of an underground hydrocarbon (e.g., oil or natural gas) reservoir. The production of oil and natural gas by drilling one or more wells in these reservoirs is usually accompanied by the co-production of water, which is always naturally present in any hydrocarbon reservoir. In order to increase oil production and ensure sufficient reservoir pressure, it is usually necessary to drill one or more injection wells. Therefore, gas, water or chemicals (e.g., polymers) can be injected into the reservoir to increase the viscosity of water. The flow through one or more injection wells and one or more production wells is controlled by the flow rate, pressure and valve settings in the specified well, and the flow rate, pressure and valve settings can vary over time. In this context, the optimal operation of the wells of the reservoir depends on a variety of properties, which further depends on time and space, such as the different spatial positions of the reservoir.
[0006] However, there are still a large number of attributes for which one or more values cannot be determined at all, or cannot be determined with sufficient accuracy or in a fast and cost-effective manner. Therefore, this raises what can be called the small data challenge, i.e., some data is known and some of it is unknown in multidisciplinary systems. In other words, the small data challenge can be described as a situation where there is an excess of some data or data types compared to other data or data types that are non-existent or limited. This requires data augmentation methods to facilitate the determination of such missing data.
[0007]
[0006] Therefore, there is a need to provide a method of determining one or more values of one or more properties of a subterranean hydrocarbon reservoir in a cost-effective and rapid manner.
[0008] Efforts have been made in US2021 / 0149070 A1, which relates to a subsurface reservoir modeling method for applying elastic property constraints to model the distribution of pseudo-components. The proposed method is limited to a metaheuristic optimization process, in particular to a variable neighborhood particle swarm optimization (VNPSO) with an external penalty function (EPF). In addition, it includes nested optimization methods and is therefore limited to specific workflows. In addition, the data-augmented VNPSO+EPF method is time and computationally intensive.
[0009] Further attempts have been made to provide corresponding data enhancement methods for exploiting reservoirs using numerical models. Exemplary reference is made to the contributions of Travis Ramsay and Halliburton: "Small Data Driven Metaheuristic Petro-Elastic Model Development Using Seismic Inversion", SPE-195491-MS, presented at the EAGE Conference and Exhibition (2019) held in London, UK, and Brandon Thibodeaux et al.'s contribution: "Closed-Loop Integrated Time-Lapse Seismic Feasibility in Amberjack Field-Deepwater Offshore Gulf of Mexico", SPE-196670-MS, presented at the Reservoir Characterization and Simulation Conference and Exhibition (2019) held in Abu Dhabi, UAE. The techniques of these two contributions have limitations and defects similar to US2021 / 0149070 A1.
[0010] Against this background, the object of the present invention is to improve the deficiencies of the prior art. In particular, the object is to provide an improved method for enhancing unknown data in fields characterized by small data challenges. The object is to provide a method for improving and facilitating the simulation of the properties of hydrocarbon subsurface reservoirs for exploitation of the reservoirs. In general, the object is to improve the exploitation of subsurface reservoirs.
[0011] These and other objects which become apparent from the following description are solved by a method according to claim 1 , a system according to claim 15 and a computer program according to claim 16 . Summary of the invention
[0012] One aspect of the present invention relates to a method for simulating at least one characteristic of a hydrocarbon underground reservoir for exploitation of the reservoir. The method can be used to determine and predict the characteristics of the reservoir to facilitate, for example, optimal production or exploitation of the reservoir. In one example, the production rate can be advantageously adjusted based on the simulated characteristics, which improves exploitation in an economical manner. In addition, the method can help determine characteristics that cannot be measured with high accuracy in a cost-effective manner. Thus, ultimately, the exploitation of the reservoir is improved.
[0013] The one or more characteristics of a hydrocarbon underground reservoir may be understood to include any parameter associated with the reservoir and that may be used to describe and / or exploit the reservoir. In one example, this may include reservoir pressure, wellbore damage or obstruction, and permeability changes. The one or more characteristics of a reservoir may also describe the behavior or condition of the reservoir.
[0014] The reservoir is characterized by at least a first attribute, a second attribute, and a third attribute, wherein the first attribute, the second attribute, and the third attribute are different from each other and are associated with each other according to a relationship. The reservoir can be characterized by several attributes. Exemplarily, the reservoir can be characterized by static and / or dynamic reservoir attributes such as petrophysical, rock physical, and / or geomechanical attributes. Other examples are primary wave velocity (Vp) and secondary wave velocity (Vs). In addition, a relationship between at least the first attribute, the second attribute, and the third attribute is provided. The relationship can be a physical relationship, i.e., the attributes can be related to each other according to one or more physical laws, and one or more physical laws can be expressed by mathematical formulas. The relationship can also be established by one or more empirical rules, wherein the behavior of the first attribute, the second attribute, and the third attribute has been previously observed to determine the relationship between them. As an example, it can be known that the primary wave velocity (Vp) is sensitive to the fluid attributes in the rock pores. This correlation of the two attributes can be described empirically. Similarly, it can be known that the secondary wave velocity (Vs) is insensitive to the fluid attributes. In addition, the ratio Vp / Vs can be used to classify the lithology of the reservoir to establish a relationship. The relationship may also be provided by a combination of one or more physical laws and one or more empirical rules.One property of the reservoir may be defined or described by other properties of the reservoir, where the definition or description may be provided by a relationship.
[0015] The method includes the step of determining a value of a first property of a reservoir, the step including the steps of: providing a predetermined value of a second property of the reservoir; providing an input value of a third property of the reservoir; providing a first uncertainty distribution of the first property of the reservoir; and performing an optimization operation at least in part using the first uncertainty distribution, the predetermined value of the second property, the relationship between the properties, and the input value of the third property when determining the value of the first property. A property of the reservoir (e.g., pressure) can have values. The property can depend on time and space and can vary along the spatial position of the reservoir. Thus, the property can also have two or more or more values, such as any tensor (an algebraic object that describes a multilinear condition between a group of algebraic objects related to a vector space), i.e., permeability or connected porosity. The uncertainty distribution can also be considered to include a range of values, such as a range of reasonable values.
[0016] Determine a value for a first attribute that may be initially unknown.
[0017] A predetermined value of the second property is provided. The predetermined value of the second property may be previously derived using suitable measurement techniques or based on one or more previous applications of the method according to embodiments of the invention, or estimated based on, for example, engineering expertise.
[0018] In addition, an input value of a third attribute is provided. For example, the input value of the third attribute may be previously derived using a suitable measurement technique or based on one or more previous applications of a method according to an embodiment of the invention, or estimated based on, for example, engineering expertise or based on random guesswork about a distribution of values or a reasonable range of values.
[0019] A first uncertainty distribution of a first attribute is provided. Preferably, the distribution has a lower end and an upper end of a value based on an applicable value of the first attribute (i.e., a value that can be considered to be physically applicable). However, different lower and upper end values of the distribution may also be applied, for example, based on engineering expertise. Therefore, the first uncertainty distribution of the first attribute of the reservoir may be based at least in part on a predetermined range of values associated with the first attribute. This may be advantageous when it is desired to limit one or more values of the first attribute to be determined to physically applicable values. In addition, a suitable shape of the uncertainty distribution (i.e., the shape of each function describing the uncertainty distribution when represented by a graph, or the shape of the uncertainty distribution when represented by a graph) may include a uniform uncertainty distribution, a normal uncertainty distribution or a Gaussian uncertainty distribution, a u-shaped uncertainty distribution, a triangular uncertainty distribution, a quadratic uncertainty distribution, a lognormal uncertainty distribution, a negatively skewed uncertainty distribution, or a positively skewed uncertainty distribution. A combination of such distributions may also be used. Specifically, the shape of the uncertainty distribution may be designed for the first attribute. Providing such a first uncertainty distribution advantageously facilitates consideration of any changes that are generally dominant in attributes such as the first attribute. For example, the attribute may be affected by structural changes or uncertainties (e.g., changes or uncertainties in the total rock volume as a first attribute or changes or uncertainties in pore pressure as a first attribute). Changes or uncertainties may also be classified according to uncertainty maps, which may be obtained from wavelength and / or amplitude maps associated with the attributes of the reservoir. This is advantageous because providing changes in the attribute to be determined may result in improved determination of the value of the attribute. Such improved determination may help to draw conclusions for improved reservoir exploitation. Typically, changes or alterations in the basic attributes of a reservoir always exist at least to some extent. For example, the first attribute may be the permeability of the reservoir, where the corresponding value may be estimated to be within a range between the lower end x1 and the upper end x2. The distribution range may be between x1 and x2. In addition, the distribution may be described by a normal shape or a Gaussian shape, so that values close to but greater than the lower end x1 and values close to but less than the upper end x2 are less likely to occur than the average values of x1 and x2. The distribution may also be described by a uniform shape, so that each value appears with the same probability. For example, an uncertainty distribution may also be considered as a range of values or a range of reasonable values. Preferably, the distribution has an uneven shape. An inhomogeneous shape of the distribution may be understood to mean that values of the distribution within the range of x1 and x2 occur with different probabilities and / or that at least one value has a different probability of occurrence than one or more of the other values.
[0020] Determining the value of the first attribute also includes performing an optimization operation using, at least in part, the first uncertainty distribution, the predetermined value of the second attribute, the relationship between the attributes, and the input value of the third attribute when determining the value of the first attribute. The optimization operation can be understood as selecting the best or optimal element(s) or the best or optimal value(s) from some available elements or values according to one or more criteria. The criteria can be based on one or more attributes or values associated with the optimization operation. For example, the optimization operation can include maximizing or minimizing an objective function, wherein the objective function can include one or more criteria.
[0021] The optimization operation may include minimizing the difference between an input value and a calculated value of the third attribute, wherein the calculated value may be determined using the relationship between the attributes, the first uncertainty distribution, and a predetermined value of the second attribute. Thus, the value of the first attribute may be determined with high accuracy in a fast and cost-effective manner. Consideration of the uncertainty distribution even further improves the accuracy and reliability of the determination of the value of the first attribute.
[0022] In another step, the method comprises applying the determined value of the first attribute when simulating at least one characteristic of a hydrocarbon underground reservoir for exploitation of the reservoir. Thus, the value of the first attribute previously determined by performing an optimization operation is now applied in the context of simulating at least one characteristic of the reservoir. In other words, the value accurately determined by taking into account the first uncertainty distribution now enables improved exploitation of the reservoir.
[0023] The method of the present invention can be applied to simulate the behavior of a reservoir with high accuracy over a potentially long period of time, and thereby improve the exploitation of the reservoir. Improved exploitation may exemplarily require placing well logging based at least in part on one or more characteristics of the simulated reservoir. Another example is to adjust the parameters associated with the placed well logging to increase the production of the reservoir. In addition, the method improves existing integrated workflows, such as geomechanical modeling, virtual cable logging, geosteering, 4D seismic drive history matching and other workflows. In particular, the method provides consideration of statistical evaluations of determined data (e.g., first attributes), and thereby improves the exploitation of the reservoir.
[0024] The application of the method can be extended to other technical fields and across disciplines, including data and / or attributes of multiple dimensions characterized by limited or missing data. In one example, the method can be a computer-implemented method. The entire method can be computer-implemented or only some method steps.
[0025] In the above, one or more of the first attribute, the second attribute and / or the third attribute may be an N-dimensional structure, where N is a number greater than 0, 1, 2, or greater than or equal to 3. In addition, one or more of the first attribute, the second attribute and / or the third attribute may also each depend on space and time and / or include multiple sub-attributes.
[0026] In a preferred embodiment, performing the optimization operation includes: calculating a sampled value of the first attribute based at least in part on a first uncertainty distribution of the first attribute; calculating an output value of a third attribute based at least in part on a relationship between the calculated sampled value of the first attribute and the attribute; selecting a value of the calculated output value of the third attribute that is closest to an input value of the third attribute; and determining a corresponding value of the first attribute according to the calculated sampled value of the first attribute corresponding to the value of the selected output value of the calculated third attribute that is closest to the input value of the third attribute. The method includes calculating a sampled value of the first attribute, which is an unknown attribute of the reservoir that needs to be determined. The calculation is performed at least in part using a first uncertainty distribution of the first attribute. For example, the sampled value is calculated based on a probability of occurrence of the sampled value according to the first uncertainty distribution. Such changes or changes in values that are generally inherent to the attribute can be advantageously considered together with the uncertainty distribution. The output value of the third attribute is calculated based on a relationship between the attributes (e.g., the first attribute, the second attribute, and the third attribute). In this calculation step, the sampled value is used as the value of the first attribute. In addition, the output value is calculated using a predetermined value of the second attribute of the provided reservoir in the relationship between the attributes. Attributes may have multiple dimensions, such as relative to space or relative to a spatial axis of the reservoir and relative to time, and may therefore provide a relationship that cannot be solved in a straightforward manner for unknown values.
[0027] In one example, an unknown value of an attribute (e.g., a first attribute) associated with another attribute or multiple attributes (e.g., a second attribute and a third attribute) through a relationship can be calculated by an explicit calculation scheme. In another example, this may not be the case, so that the unknown value may not be explicitly calculated by a simple reorganization of the relationship. Preferably, all sampled values of the first attribute are used as the value of the first attribute in the relationship between the first attribute, the second attribute, and the third attribute to calculate the output value of the third attribute. Then, a value in the calculated output value of the third attribute is selected, which is preferably the value closest to the input value of the third attribute. In this regard, based on the absolute difference between the input value and the output value of the third attribute, the closest value can be the value of the input value closest to the third attribute. In addition, in the case of multidimensional attributes, the value of the calculated output value can be performed using the L1 norm, i.e., by forming the sum of the absolute differences between the calculated output value in each dimension and the input value in each dimension. In another example, the value of the calculated output value can be performed using the L2 norm, i.e., by forming the sum of the squares of the differences between the calculated output value in each dimension and the input value in each dimension. Furthermore, the method includes determining a corresponding value of the first attribute based on the calculated sampled value, the calculated sampled value corresponding to the selected value of the calculated output value of the third attribute. This means that a value of the first attribute is determined that, when used in a relationship between the attributes, produces the calculated value of the selected calculated output value of the third attribute. The method can help determine the value of the first attribute in a cost-effective and statistically robust manner by taking into account potential changes in the value of the first attribute.
[0028] Preferably, the value of one or more of the first attribute, the second attribute, and the third attribute has a first data type, and the value of one or more of the first attribute, the second attribute, and the third attribute has a second data type different from the first data type. The data type of a value can be understood as a way of describing how a value can be represented, used, and / or processed. For example, an integer can be a data type representing a number, such as without a floating point. As another example, a real value can be represented by a floating point. Other examples of data types include characters, strings, or Booleans. For example, the data type of one or more values of the first attribute can be different from the data type of one or more values of the second attribute. Therefore, the method can handle different data types. The values of reservoir attributes typically have different data types, and reflecting these can improve the accuracy of determining missing values of attributes. This can be achieved by integer programming, which can be understood as a technique in which the values of attributes can be restricted to integers. Therefore, the exploitation of reservoirs can be improved.
[0029] Also preferably, the first data type is an integer value data type, and the first data type is saved during the step of performing the optimization operation. Also preferably, the first data type is saved during any or all steps of the method. One or more attributes can be an integer value data type, which can be saved or maintained during the execution of the optimization operation. Therefore, the data type is not changed. For example, preferably throughout the method, the integer remains an integer. Therefore, the method can advantageously not require pre-processing and / or post-processing to change the data type before and / or after the optimization operation or during any other step of the method. Such pre-processing and / or post-processing may adversely affect the determination of the first attribute, because, for example, rounding errors may be introduced in this way, and the determination may be slowed down by additional processing steps. Therefore, the method can further promote the improved determination of the value of the first attribute.
[0030] For example, an attribute may be a rock type, where integer values 1, 2, 3, and 4 represent the rock types "Siltstone (high porosity)", "Siltstone (low porosity)", "Shale", and "Dolomite", respectively. Other attributes may also be classified as integer values. Using and storing integer-valued data types representing values of one or more attributes may be implemented by integer programming, which offers the advantage of consuming even less computer memory when the method is applied to a computer. Integer programming may be understood as a technique in which the values of an attribute may be restricted to integers. In particular, integers may require less memory than floating point data types or other data types.
[0031] In addition, preferably, the second data type is different from the integer value data type, preferably a real value data type. The second data type can be saved during the step of performing the optimization operation. In addition, preferably, the second data type is saved during any or all steps of the method. For example, the first attribute is an integer value data type, and the second attribute is a real value data type. Therefore, the method preferably allows the attributes of integer value data types and real value data types to be processed simultaneously. This may be advantageous because in any or all steps of the entire optimization operation or method, an attribute (such as a displacement in one direction) can be saved or maintained as a real value data type, while another attribute or other attributes can be saved as an integer value data type. Therefore, the method can, for example, process two data types simultaneously during the optimization operation. Therefore, before and / or after the optimization operation or during any other step of the method, no pre-processing and / or post-processing is required to change the data type.
[0032] In another preferred embodiment, performing the optimization operation includes: providing an initial generative adversarial network (GAN); optionally using predetermined data associated with the reservoir, at least partially using a first uncertainty distribution, a predetermined value of the second attribute, a relationship between the attributes, and / or an input value of a third attribute to train the initial GAN to obtain a trained GAN; and using the trained GAN when determining the value of the first attribute. Thus, the GAN is used to determine the value of the first attribute in a fast and cost-effective manner. An initial GAN is provided and then trained. Advantageously, this may require less training data than a typical neural network because the initial GAN and the trained GAN each include two neural networks. The training data for the initial GAN may be a first uncertainty distribution, a predetermined value of the second attribute, a relationship, and / or an input value of the third attribute. In addition, additional predetermined data associated with the reservoir may be used. The additional predetermined data may be referred to as values of the first attribute, the second attribute, and / or the third attribute, or in another example, as values of other attributes of the reservoir. The GAN may be provided with a generator as a first neural network, which generates one or more potential values, and a discriminator as a second neural network, which evaluates and / or corrects the generated one or more potential values. Therefore, the algorithm structure of the GAN may use two neural networks to generate improved one or more potential values. This may include improved potential values for the first attribute. The application of the obtained or trained GAN enables the calculation time to be reduced when the value of the first attribute is determined by using a generator and a discriminator (i.e., two neural networks). Therefore, the method may advantageously solve the overall data determination by using the GAN, rather than optimizing a single data point. Therefore, the calculation of the control variable (e.g., the third attribute) may be solved at the highest level, and the unknown value of the attribute (e.g., the first attribute) may be determined simultaneously. By using the GAN, the determination of the unknown attribute may be improved and the performance may be better. For example, compared with particle swarm optimization (PSO), the GAN may enable the determination of attributes simultaneously over the entire domain, rather than determining a specific data type one value at a time.
[0033] In another preferred embodiment, performing an optimization operation includes: applying at least one proxy function based at least in part on a subset of predetermined values of the first attribute; and using at least one proxy function and an input value of a third attribute at least in part when determining the value of the first attribute. For example, the optimization operation includes applying Bayesian optimization. It may be the case that some values of an attribute are known, but other values of the attribute are unknown. In an example, multiple values of the first attribute are required, but only a small amount of data is given. In another example, the relationship may be at least partially ambiguous or unknown. In another example, multiple values of the first attribute are required, but only a small amount of data is given, and in addition the relationship is at least partially ambiguous or unknown. In at least one of the above examples, the method facilitates providing one or more proxy functions. These one or more proxy functions may be similar to a given small amount of data, i.e., some values of the first attribute. Then, one or more proxy functions may be advantageously updated to improve accuracy, such as between a given small amount of data and the proxy function, to improve accuracy in the remaining data range. Therefore, this embodiment may advantageously handle ambiguous or undefined small data relationships, such as physical or empirical relationships (one or more). This embodiment of the method of the invention may also be performant because it does not require the evaluation of one or more derivatives to improve accuracy.
[0034] In another preferred embodiment, the method further comprises: preferably before applying the value of the determined first attribute in simulating at least one characteristic of the hydrocarbon underground reservoir, at least partially based on a predetermined value of the first attribute, verifying the value of the determined first attribute. The verification can be used to determine the accuracy of the value of the determined first attribute. The first attribute can be a static or dynamic attribute. One or more static attributes of the reservoir are typically obtained from static tests associated with core samples of the reservoir. One or more dynamic attributes can be obtained from cable measurements of compression, shear velocity and / or rock density. Both static attributes and dynamic attributes can refer to attributes such as Young's modulus, but they may be subject to different aspects of the attribute, such as time. Verification of static attributes can include determining statistical confidence intervals for realizations. For example, if the previously unknown values of the attribute may only be constrained, i.e., some values but not all values are known, verification of static attributes can also include minimizing the difference of the value of the determined attribute (e.g., the first attribute) relative to one or more known values of the attribute. Verification of dynamic attributes can include determining at least one statistical confidence interval for realizations. The interval of realization can include realizations generated from previously unknown determined values by the method, optionally by involving realizations generated from determined values. For example, if the previously unknown values of the attribute may only be restricted, i.e., some values but not all values are known, the verification of the dynamic attribute can also include minimizing the difference of the value of the determined attribute (e.g., the first attribute) relative to one or more known values of the attribute. In this regard, a portion of the value of the first attribute can be known, such as the portion of the value of the first attribute can be observed or measured. In another example, the portion of the value of the first attribute can be retrieved from a simulation. In this article, "realization" can be understood as data or values generated, determined or estimated for unknown data or attributes. Verification can be not limited to the applicability of the method of the present invention as described herein, but can also be applied to meta-heuristic optimization techniques, such as variable neighborhood particle swarm optimization (VNPSO) using an external penalty function (EPF).
[0035] In a preferred embodiment, the method further comprises: preferably before the step of performing the optimization operation, preferably based on a predetermined value range associated with the first attribute, the second attribute and / or the third attribute, converting the first uncertainty distribution of the first attribute, the predetermined value of the second attribute and / or the input value of the third attribute into one or more quantum states; wherein performing the optimization operation comprises at least partially using one or more quantum states when determining the value of the first attribute. For example, this is achieved by applying quantum computing. Quantum computing enables further reduction of the computation time when determining the value of the first attribute. The value of the attribute can be calculated taking into account, for example, quantum states applicable to one or more quantum bits (qubits) or quantum bits (qbits). As an example, once measured, a quantum bit (qubit) or quantum bit (qbit) can be in any of the defined states |0> or |1>, where |0> or |1> can be an orthogonal basis of a two-dimensional vector space. Quantum superposition can ensure that n quantum bits are in 2 n(2 to the power of n) states exist in superposition. This can make it possible to exponentially explore the complex interactions between physically reasonable numerical solutions for a single given unknown number in a data set. In addition, due to quantum entanglement, measurements of multiple unknowns in a data set can be performed simultaneously. These values can be defined by quantum states according to the laws of quantum physics. Quantum computing provides superposition and entanglement. Superposition can be regarded as a linear combination of other different quantum states. Superimposed quantum states can form new effective quantum states. Quantum states can be described by quantum particles. In the case of multiple unknown values of multiple attributes in any and / or across multiple dimensions (e.g., across different spatial dimensions), the measurement of superimposed quantum particles can determine the unknown value of an attribute (e.g., a first attribute) or the state solution of the unknown value of one or more attributes. Classical computers using one or more bits cannot provide superposition. Therefore, quantum computing is high-performance and provides a rapid determination of the value of the first attribute. In addition, in quantum mechanics, when the quantum state of each particle cannot be described independently of the quantum state of other particles, the quantum particles may be entangled. However, the quantum state of the system as a whole can be described, wherein the system can include all quantum particles. Thus, compared to a classical bit, a quantum state can include additional values between 0 and 1. This reduces computation time. The quantum state can be advantageously scaled according to the range of physically reasonable or applicable solutions for expected unsolved or unknown data in space. Scaling can be understood by the following example: a quantum computer can be employed to determine the (previously unknown) value of the parameter h that satisfies the function g(h) as a minimum. In addition, the created superposition state can be utilized so that g(h) can be simultaneously applied to a series of physically reasonable or applicable solutions. In addition, quantum interference effects that can be associated with a pseudo-objective function can be applied to bias the measurement of quantum bits (qubits) or quantum bits (qbits) to a desired set of states. Therefore, quantum computing further improves the simulation of at least one characteristic of the reservoir by determining the value of the first attribute more quickly.
[0036] In another preferred embodiment, performing the optimization operation includes: preferably before applying the determined value of the first attribute in simulating at least one characteristic of the hydrocarbon subsurface reservoir (e.g., step b)), providing an updated first uncertainty distribution for the first attribute based at least in part on the determined value of the first attribute; and repeating the steps of performing the optimization operation using at least in part the updated first uncertainty distribution in place of the first uncertainty distribution to determine an updated value of the first attribute to be used in place of the determined value of the first attribute for step b). In this preferred embodiment, a Stochastic Simplex Approximate Gradient technique is applied, preferably in combination with applying a bootstrapping method. The Stochastic Simplex Approximate Gradient technique is a gradient-based optimization technique such that it generally determines a derivative to reach or approach a target. In this embodiment, after determining the value of the first attribute, a bootstrap technique is employed to exploit the overall connectivity of the solution space at the solution (i.e., the value of the first attribute). The bootstrap method advantageously generates a range of different values of the first attribute by varying the first solution and calculating, for example, an L1 norm or an L2 norm. The self-help method can be applied to any optimization operation applied in the context of the method according to the present invention and in particular in the context of step a4) of the method according to the present invention. For example, when minimizing the difference between the input value and the calculated value of the third attribute, the self-help method can be applied, wherein the calculated value can be determined using the relationship between the attributes, the first uncertainty distribution and the predetermined value of the second attribute. The self-help method can also be applied when performing an optimization operation including providing an initial and / or trained GAN. As will be understood by those skilled in the art, the self-help method can generally be applied to any optimization operation. This can be repeated until a specified allowed change of the L1 norm or the L2 norm occurs. This embodiment helps to efficiently handle the uncertainty behind the attributes of the method. Therefore, this embodiment further improves the determination of the value of the first attribute and enhances the exploitation of the reservoir. In particular, it is advantageous to apply this embodiment to data enhancement in reservoir characteristics and determination of reservoir attributes because the reservoir can be described more accurately.
[0037] Preferably, the method step of providing a first uncertainty distribution for a first property of the reservoir further comprises providing a second uncertainty distribution for a second property of the reservoir, wherein performing the optimization operation further comprises at least partially using the second uncertainty distribution when determining the value of the first property. Thus, the method also provides for the uncertainty of a known property. For example, the value of the second property may be known, however, these values may still be subject to variation under actual conditions. These variations that are typically inherent to such properties can advantageously be taken into account using the second uncertainty distribution.
[0038] Preferably, the property of the reservoir is a static and / or dynamic property of the reservoir, and / or the property of the reservoir is at least one of the following list: P impedance, S impedance, density, resistivity, gamma ray, ratio of primary wave velocity (Vp) to secondary wave velocity (Vs), porosity, Poisson's ratio, effective stress, displacement such as horizontal displacement, vertical displacement or depth displacement, pressure, saturation, porosity, permeability. The property of the reservoir may also be one of the following: capillary pressure, phase saturation, relative permeability of any phase of the fluid of the reservoir and / or fluid properties (e.g. oil viscosity).
[0039] Preferably, providing the predetermined value of the second property and / or providing the input value of the third property comprises measuring the respective value, preferably by logging and / or sonic logging the respective value.
[0040] In a preferred embodiment, the method further comprises applying the determined value of the first attribute to: optimizing exploitation of the reservoir, such as hydrocarbon production, production time and / or production rate of the reservoir; monitoring connectivity and / or compartments of the reservoir; and / or determining at least one well location for exploiting the reservoir. The method can be used in a variety of technical applications and can improve exploitation of hydrocarbon underground reservoirs. For example, the determined value of the first attribute can be used to identify optimal hydrocarbon reservoir production time and / or rate. In particular, if multiple values of the first attribute were previously missing or the values of several attributes were unknown and potentially connected according to a relationship, the method provides progress in exploiting the reservoir.
[0041] The method may not be limited to application to reservoir production optimization, but may be generally applied to determine or generate data to help characterize a reservoir. Applying the value of the determined first attribute may also improve or enable decision making, or may be utilized to perform other workflows that rely on data, data types, or attributes that are ambiguous or unresolved or whose values are unknown. In addition, a series of relationships may be considered that may depend on attributes of a reservoir having values of different data types. These attributes may be related to wireline logging, sections, production, or rock physics models. Other examples of attributes may include rock type, implementation scheme, path, and sort type answer, which may be described as only integer valued data types.
[0042] Another aspect of the present invention relates to a system for simulating at least one characteristic of a hydrocarbon underground reservoir for the purpose of exploiting the reservoir, wherein the reservoir is characterized by at least a first attribute, a second attribute and a third attribute, wherein the first attribute, the second attribute and the third attribute are different from each other and are related to each other according to a relationship, and the system comprises: a device for determining a value of a first attribute of the reservoir, comprising: a device for providing a predetermined value of the second attribute of the reservoir, a device for providing an input value of the third attribute of the reservoir; a device for providing a first uncertainty distribution of the first attribute of the reservoir; a device for determining the value of the first attribute using at least in part the first uncertainty distribution, the predetermined value of the second attribute, the relationship between the attributes and the input value of the third attribute; and a device for applying the determined value of the first attribute when simulating at least one characteristic of a hydrocarbon underground reservoir for the purpose of exploiting the reservoir.
[0043] The system can be used to determine the value of the first attribute and simulate at least one reservoir characteristic. The system can determine the value and / or simulate at least one characteristic automatically or by user interaction. For example, the simulation by the system can be started and / or stopped by the user. The device for determining, providing and / or applying can be a physical device, or can be represented by, for example, a computer or a computer program. It should be understood that the advantages and / or features of the method can also be applied to the system, and the advantages and / or features of the system can also be applied to the method. In addition, the system according to the present disclosure can be used to perform one or all of the method steps described above. In one example, the first uncertainty distribution can be described by a uniform shape so that each value appears with the same probability. For example, the first uncertainty distribution can also be considered as a range of values or a range of reasonable values. Preferably, the distribution has an uneven shape. The uneven shape of the distribution can be understood as making the values of the distribution within the range of x1 and x2 appear with different probabilities and / or at least one value has a different probability of occurrence compared to one or more of the other values.
[0044] The means for determining a value of a first property of the reservoir may be configured such that the value of the first property is calculated, derived or otherwise retrieved, and may include one or more of the following means.
[0045] The means for providing a predetermined value of a second property of the reservoir may be configured such that the value of the second property is provided, for example, by measurements or previously known values of the second property.
[0046] The means for providing an input value of a third property of the reservoir may be configured such that a value of the third property is provided, for example by measurement or a previously known value of the third property and then used as the input value of the third property.
[0047] The means for providing a first uncertainty distribution for the first attribute may be configured such that the uncertainty distribution is calculated, determined or otherwise retrieved, wherein the uncertainty distribution is associated with the first attribute.
[0048] The means for determining the value of the first attribute may be configured such that the value of the first attribute is calculated, derived or otherwise retrieved such that the value is known.
[0049] The apparatus for applying the determined value of the first property in modeling at least one characteristic of the reservoir may be configured such that at least one characteristic of the reservoir may be analyzed, evaluated, examined or otherwise assessed to derive information about the reservoir that was not available prior to the beneficial contribution of the system.
[0050] The system can be used to determine the value of the first attribute in a rapid and cost-effective manner, which can further enable simulation of at least one characteristic of the reservoir. Thus, the system facilitates the exploitation of the reservoir. In addition, the system enhances decision making, such as for improved well location and well operation, and thus provides cost-effectiveness and economic advantages.
[0051] A means for providing something may also be one device or the same device.
[0052] Another aspect of the present invention relates to a system for simulating at least one characteristic of a hydrocarbon underground reservoir for the purpose of exploiting the reservoir, wherein the reservoir is characterized by at least a first attribute, a second attribute and a third attribute, wherein the first attribute, the second attribute and the third attribute are different from each other and are related to each other according to a relationship, and the system includes: a computing unit for determining a value of a first attribute of the reservoir, the computing unit including: a first data acquisition unit for providing a predetermined value of the second attribute of the reservoir; a second data acquisition unit for providing an input value of the third attribute of the reservoir; a third data acquisition unit for providing a first uncertainty distribution of the first attribute of the reservoir; an optimizer unit for determining the value of the first attribute at least partially using the first uncertainty distribution, the predetermined value of the second attribute, the relationship between the attributes and the input value of the third attribute; and a processing unit for applying the determined value of the first attribute when simulating at least one characteristic of the hydrocarbon underground reservoir for the purpose of exploiting the reservoir.
[0053] The first data acquisition unit, the second data acquisition unit and the third data acquisition unit may also be one data acquisition unit.
[0054] The system or different apparatus may advantageously be configured to perform the method as described above.
[0055] Another aspect of the invention relates to a computer program comprising instructions which, when executed by a computer, cause the computer to carry out the method described above. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] In the following, the drawings are briefly described:
[0057] Figure 1 A flow chart showing a method for simulating at least one property of a hydrocarbon underground reservoir for exploitation of the reservoir according to an embodiment,
[0058] Figure 2 A system for simulating at least one property of a hydrocarbon underground reservoir for exploitation of the reservoir according to another embodiment of the present invention is shown, and
[0059] Figure 3 A conceptual example of an implementation according to an aspect of the invention is illustrated by a Geertsma shift. DETAILED DESCRIPTION
[0060] Hereinafter, the present invention is described in more detail with reference to the accompanying drawings.
[0061] Figure 1 A flow chart of a method 1 for simulating at least one characteristic of an underground hydrocarbon reservoir for exploitation of the reservoir according to an embodiment of the present invention is shown. Exploitation of a reservoir may also include exploration, production and recovery of hydrocarbons. Simulating at least one characteristic of a reservoir is also important for strategic decision making, for example to improve economic and also environmental aspects associated with the earth's natural resources. Method 1 is implemented as a process for determining a value of an attribute of a reservoir, wherein the value of the attribute is unknown, ambiguous or too complex to be measured or otherwise retrieved. An unknown attribute value may be, for example, the porosity of a rock with respect to space and time. Knowing this parameter may improve exploitation of a reservoir. The method may be performed entirely on a computer.
[0062] In step 100, the value of a first attribute associated with a reservoir is determined, including the following sub-steps. In step 110, a predetermined value of a second attribute of the reservoir is provided. In step 120, an input value of a third attribute of the reservoir is provided. These values are ensured to be provided by measured data, observed data or previously known data. In step 130, a first uncertainty distribution is provided, which first uncertainty distribution may include a range of values of the first attribute of the reservoir. The uncertainty distribution enables consideration of statistical variations associated with the value of the first attribute within a suitable physical range. The first attribute, the second attribute and the third attribute are associated with each other according to a relationship. The method may also enable provision of a second uncertainty distribution, which second uncertainty distribution may include a range of values of the second attribute. Therefore, the method may also consider its variation. The attributes may also be multidimensional and dependent on space and time. Therefore, the relationship is inherently complex and cannot be easily solved for unknown data. In step 140, an optimization operation is performed to determine the value of the first attribute. This step may include using a generative adversarial network GAN to beneficially use two neural networks, preferably trained using attributes associated with the reservoir, to determine the value of the first attribute. The data types of the attributes may be different. For example, "lithology" can be classified according to integer values, where the attributes "displacement" and "density" are described using real values. Both data types are reflected and saved during method 1. Therefore, no pre-processing and / or post-processing is required, and the accuracy of the value of the determined first attribute is improved, while the computational cost is reduced. Steps 130 and 140 can also be repeated by applying a random simplex approximation gradient technique combined with a bootstrap method, thereby even further improving the accuracy of the value of the determined first attribute. In the case where the relationship is unknown, inaccurately known, ambiguous or otherwise unavailable, a Bayesian optimization can be performed in the optimization operation of step 140. Therefore, the third attribute can be regarded as a black box, and the surrogate function is similar to at least a portion of the value of the first attribute that needs to be known in order to determine the remaining value of the first attribute. As an example, it is sufficient to know 10%, 20%, 30% or more of the value of the first attribute.
[0063] In step 200, a validation is performed to determine the achieved statistical confidence interval, thereby evaluating the accuracy of the determined attribute value. To do this, it may be necessary to know at least some data of the determined attribute. This validation step may be optional, but may provide a valuable assessment of the determined attribute value.
[0064] In step 300, the determined value of the first attribute or the determined and previously unknown value is applied to simulate at least one characteristic of the hydrocarbon subsurface reservoir for exploitation of the reservoir. This includes multiple tasks associated with the reservoir and extends and prolongs the entire reservoir and production life cycle. For example, method step 300 or method 1 can be applied even before production begins.
[0065] Figure 2 A system 1000 for simulating at least one characteristic of a hydrocarbon underground reservoir for exploitation of the reservoir according to another embodiment of the present invention is shown. The reservoir is characterized by at least a first attribute, a second attribute, and a third attribute, wherein the first attribute, the second attribute, and the third attribute are different from each other and are related to each other according to a relationship. These attributes are multidimensional, for example, depending on three-dimensional space and time. The system 1000 includes a computing unit 1100 for determining one or more values of a first attribute of the reservoir. The computing unit 1100 also includes a data acquisition unit 1200, which is used to provide one or more predetermined values of the second attribute of the reservoir, one or more input values of the third attribute of the reservoir, and a first uncertainty distribution of the first attribute of the reservoir. The computing unit 1100 also includes an optimizer unit 1300, which is used to determine the value of the first attribute using, at least in part, the first uncertainty distribution, the predetermined value of the second attribute, the relationship between the attributes, and the input value of the third attribute. The system 1000 further comprises a processing unit 1400 for applying the determined value of the first property in simulating at least one characteristic of a hydrocarbon subterranean reservoir for exploiting said reservoir.
[0066] Figure 3 A conceptual example of an embodiment according to an aspect of the invention is shown by the perpendicular free surface displacement from the Geertsma displacement of the strain core.
[0067] The mathematical problem associated with the Geertsma vertical free surface displacement from the strain kernel can be described as follows. The extraction of liquid or gas from the reservoir 500 may cause compaction of the reservoir 500 and may change the stress and deformation field in the surrounding rock. As an example, the reservoir 500 may be described as a cylinder. This exemplary structure has been analyzed theoretically by Geertsma, providing a relationship between displacement and stress. The third attribute u according to the present invention may be z is the vertical displacement and is related to the other properties according to the following relation (1). where r is the radius and z is the vertical axis; for illustration purposes Figure 3 These two are depicted in , along with another axis, to describe the space of the reservoir 500. In this case, c m , v, Δp, H, R, integral and u z The relationship between (symbols explained below) is a mathematical equation.
[0068]
[0069] In this embodiment, the uniaxial compression coefficient (c m) and Poisson's ratio (v) may be unknown. The values of the pressure difference (Δp), the thickness of the production reservoir 500 (H), and the radius (R) of the reservoir 500 may be known. The attributes Δp, H, and R, whose values may be known, may also be referred to as one or more second attributes. The terms J0 and J1 are defined as zero-order Bessel functions and first-order Bessel functions, respectively. The solution to the integral in relation (1) may be determined analytically.
[0070] The vertical displacement can be given, is known, observed, recorded or otherwise determined. This can be achieved by appropriate measurements and / or by sonic logging data of well logging data and / or by seismic data associated with the hydrocarbon underground reservoir 500. Furthermore, a given distribution may be known from historical data.
[0071] Uncertainty distributions are given for those attributes whose values are unknown within a defined range of values for one or all of the attributes. A separate uncertainty distribution may be provided for each unknown attribute. Advantageously, the defined range of values is designed individually for each attribute. For example, Poisson's ratio (v) may be provided with an uncertainty distribution, wherein in one example the values are in the range from 0.1 to 0.5. In another example, such as in fluid-saturated carbonates, the values range from 0.2 to 0.4 or 0.36. For saturated sands, the values may range from 0.07 to 0.5, and for saturated soft clays, the values may range from 0.4 to 0.5. Based on these distributions and the known attributes, u is determined using relation (1) z These estimates are expressed as Because they represent best guesses or even optimal solutions, as described below.
[0072] This step may also involve the uncertainties associated with the known properties, ie the known properties (Δp, H and R) may also be provided with an uncertainty distribution within a defined range of values.
[0073] Determine z The estimate of (r, 0) (expressed as ) step can also be called sampling u z The estimate of (r, 0) (expressed as ). The method may include sequentially or simultaneously z The estimate of (r, 0) (expressed as ) steps for sampling. For simultaneous sampling, quantum computing and / or generative adversarial networks (GANs) can be used.
[0074] Determine or calculate the recorded With estimation To find the deviation between In this example, we use the L1 norm to find In another example, the L2 norm is used to find In another example, finding the best guess may be achieved by the following equation (2), which may be performed, for example, on a node-by-node or geographic cell-by-geographic cell basis according to the structure of the attribute:
[0075]
[0076] Therefore, the attribute (c m and v) that yields the best guess or optimal solution Thus, the method advantageously provides for determining the value of an attribute whose value was previously unknown in a statistically robust manner by taking into account the uncertainty of the unknown attribute.
[0077] Reference numerals list
[0078] 1 Methods
[0079] 100 Method steps: Determine
[0080] 110 Methods and Steps: Provide
[0081] 120 Another method step: providing
[0082] 130 Another method step: providing
[0083] 140 Method steps: Execution
[0084] 200 Verification
[0085] 300 Applications
[0086] 500 Reservoirs
[0087] 1000 Systems
[0088] 1100 computing units
[0089] 1200 Data Acquisition Unit
[0090] 1300 Optimizer Unit
[0091] 1400 Processing Units
Claims
1. A method (1) for simulating at least one property of an underground hydrocarbon reservoir for the purpose of exploiting said reservoir, in, The reservoir is characterized by at least a first attribute, a second attribute, and a third attribute, wherein the first attribute, the second attribute, and the third attribute are different from each other and are related to each other according to a relationship, and the method comprises the following steps: a) determining (100) a value of a first property of the reservoir, comprising the following steps: a1) providing (110) a predetermined value of a second property of the reservoir; a2) providing (120) an input value of a third property of the reservoir; a3) providing (130) a first uncertainty distribution for a first property of the reservoir; a4) performing (140) an optimization operation using, at least in part, the first uncertainty distribution, the predetermined value of the second attribute, the relationship between the attributes, and the input value of the third attribute in determining the value of the first attribute; b) applying (300) the determined value of the first property in simulating at least one characteristic of the hydrocarbon underground reservoir for exploiting the reservoir.
2. The method according to the preceding claim, in, Executing (140) the optimization operation includes: - calculating a sampled value of the first property based at least in part on a first uncertainty distribution of the first property; - calculating an output value of the third attribute based at least in part on the relationship between the attributes and the calculated sample value of the first attribute; - selecting the value of the calculated output value of the third attribute that is closest to the input value of the third attribute; and - determining a corresponding value of the first attribute based on the calculated sampled value of the first attribute corresponding to the value of the selected calculated output value of the third attribute that is closest to the input value of the third attribute.
3. The method according to any one of the preceding claims, in, The values of one or more of the first attribute, the second attribute, and the third attribute have a first data type, and the values of one or more other attributes of the first attribute, the second attribute, and the third attribute have a second data type different from the first data type.
4. The method according to claim 3, in, the first data type is an integer valued data type, and wherein the first data type is saved during the step of performing (140) the optimization operation, Wherein, preferably, the first data type is preserved during any step of the method.
5. The method according to any one of claims 3 or 4, in, The second data type is different from an integer valued data type, preferably wherein the second data type is a real valued data type, and wherein the second data type is stored during the step of performing (140) the optimization operation, Preferably, the second data type is stored during any step of the method.
6. The method according to any one of the preceding claims, in, Executing (140) the optimization operation includes: - Provide an initial Generative Adversarial Network (GAN); - optionally using predetermined data associated with the reservoir, training the initial GAN at least in part using the first uncertainty distribution, the predetermined value of the second attribute, the relationship between the attributes and / or the input value of the third attribute to obtain a trained GAN; and - using the trained GAN in determining the value of the first attribute.
7. The method according to any one of the preceding claims, in, Executing (140) the optimization operation includes: - applying at least one proxy function based at least in part on a subset of predetermined values of said first attribute; and - using at least partly said at least one proxy function and an input value of said third attribute in determining the value of said first attribute.
8. The method according to any one of the preceding claims, in, The method further comprises: - preferably before step b), verifying (200) the determined value of the first property based at least in part on a predetermined value of the first property.
9. The method according to any one of the preceding claims, in, The method further comprises: - preferably before step a4), converting a first uncertainty distribution of the first property, a predetermined value of the second property and / or an input value of the third property into one or more quantum states, preferably based on a predetermined value range associated with the first property, the second property and / or the third property, Wherein performing (140) the optimization operation comprises: using at least in part the one or more quantum states in determining the value of the first property.
10. The method according to any one of the preceding claims, in, Executing (140) the optimization operation includes: - preferably before step b), providing an updated first uncertainty distribution for said first property based at least in part on said determined value of said first property; and - repeating step a4) at least partially using said updated first uncertainty distribution instead of said first uncertainty distribution to determine an updated value of said first property to be used in place of the value of said first property determined for step b).
11. The method according to any one of the preceding claims, in, Providing (130) a first uncertainty distribution for a first property of the reservoir includes: - providing a second uncertainty distribution for a second property of the reservoir; and wherein performing (140) the optimization operation further comprises: - using at least partly said second uncertainty distribution when determining the value of said first property.
12. The method according to any one of the preceding claims, in, The properties of the reservoir are static and / or dynamic properties of the reservoir, and / or wherein, The properties of the reservoir are at least one of the following list: P impedance, S impedance, density, resistivity, gamma ray, ratio of primary wave velocity (Vp) to secondary wave velocity (Vs), porosity, Poisson's ratio, effective stress, displacement such as horizontal displacement, vertical displacement or depth displacement, pressure, saturation, porosity, permeability.
13. The method according to any one of the preceding claims, in, Providing a predetermined value for the second property and / or providing an input value for the third property comprises measuring the respective values, preferably by logging and / or sonic logging the respective values.
14. The method according to any of the preceding claims, further comprising applying (300) the determined value of the first property to: - optimizing the exploitation of said reservoir, e.g. the production, production time and / or production rate of hydrocarbons from said reservoir; - monitoring connectivity and / or compartments of the reservoir, and / or - determining at least one well location for exploiting said reservoir.
15. A system (1000) for simulating at least one property of a hydrocarbon subterranean reservoir for use in exploiting the reservoir, in, The reservoir is characterized by at least a first attribute, a second attribute, and a third attribute, wherein the first attribute, the second attribute, and the third attribute are different from each other and are related to each other according to a relationship, and the system (1000) comprises: a) means (1100) for determining a value of a first property of the reservoir, the means comprising: a1) means (1200) for providing a predetermined value of a second property of the reservoir; a2) means (1200) for providing an input value of a third property of the reservoir; a3) means (1200) for providing a first uncertainty distribution for a first property of said reservoir; a4) means (1300) for determining a value of the first attribute using, at least in part, the first uncertainty distribution, a predetermined value of the second attribute, a relationship between the attributes, and an input value of the third attribute; b) means (1400) for applying the determined value of the first property in simulating at least one characteristic of the hydrocarbon underground reservoir for exploiting the reservoir.
16. A computer program comprising instructions, which, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 14.
