Maximum entropy application methods and systems
AssigneeSchlumberger Technology Corporation
InventorMichael John Williams
Life of this patent
Estimating the term…
Abstract
Techniques for a maximum entropy approach to assigning probabilities, including those used in multiple realization history matching systems and methods, are disclosed. For example, in one embodiment, a method includes obtaining one or more available sample values Yi associated with a reservoir; computing a maximum entropy assignment λ of one or more discrete probabilities Pi(Yi) associated with each of the one or more available sample values Yi, where each discrete probability Pi(Yi) represents a probability that one or more variables Y will take a set of particular values Yi; and performing at least one determination regarding the reservoir using the maximum entropy assignment λ, including approximating a continuous probability distribution P(Y) using a sum of probability distributions Pi(Yi)+/−λ.
Claims16
A method, comprising:
- obtaining available sample values Yi associated with a reservoir;
- computing, using a computing device, a maximum entropy assignment λ of one or more discrete probabilities Pi(Yi) each of the discrete probabilities Pi(Yi) associated with one or more of the available sample values Yi, where each of the discrete probabilities Pi(Yi) represents a probability that one or more variables Y will take a set of particular values Yi;
- and
- performing at least one determination regarding the reservoir using the maximum entropy assignment λ and the computing device, including approximating a continuous probability distribution P(Y) using a sum of probability distributions Pi(Yi)+/−λ.
One or more computer-readable non-transitory storage media containing instructions that, when executed by a computer, perform a method comprising:
- determining a maximum entropy assignment λ of probabilities Pi(Yi) associated with one or more reservoir operation variables, including assigning a likelihood by combining one or more of the probabilities Pi(Yi) with a constraint range associated with at least one of the one or more reservoir operation variables;
- for each of the probabilities Pi(Yi), determining a corresponding probability distribution Pi(Yi)+/−λ;
- and
- performing at least one determination regarding one or more reservoir operation variables using the maximum entropy assignment λ wherein performing at least one determination regarding one or more reservoir operation variables using the maximum entropy assignment λ includes approximating a continuous probability distribution P(Y) using a sum of the probability distributions Pi(Yi)+/−λ.
One or more computer-readable non-transitory storage media containing instructions that, when executed by a computer, perform a method comprising:
- reducing a multiple realization history matching model of a hydrocarbon reservoir having a plurality of variables to a set of separate one-dimensional maximizations, each one-dimensional maximization corresponding to a discrete observation of a particular variable at a particular time;
- and
- performing an estimation of at least one variable of the plurality of variables of the hydrocarbon reservoir using the set of separate one-dimensional maximizations.
Dependencies are inferred from the claim text.