Inverse Rock Physics Modeling with Probability Distribution Functions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional inverse rock physics modeling (IRPM) methods for predicting geological reservoir parameters face challenges in handling uncertainty and evaluating model performance, leading to non-unique solutions and inadequate reliability in reservoir characterization.

Innovation Solution

The method incorporates probability distribution functions (PDFs) to calculate model probabilities, allowing for improved evaluation of predicted model parameters and characterization of geological formations by integrating uncertainty analysis and Bayesian probabilities, enabling more reliable and reproducible predictions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional IRPM methods are used to predict model parameters, then the prediction process can be completed, but the reliability and reproducibility of the predictions are inadequate due to non-unique solutions and poor uncertainty handling

Engineering Contradiction:
Improvereliability of model parameter predictionsVSAvoiduncertainty information in reservoir parameters
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent transforms the conventional deterministic IRPM approach into a probabilistic framework by changing the parameter representation from single values to probability distribution functions. This allows the model to capture uncertainty in reservoir parameters (porosity, lithology, fluid saturation) and provides a more reliable prediction by quantifying the confidence levels associated with each predicted parameter.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent implements a feedback mechanism where the inverse model solver iteratively adjusts model parameters to minimize the misfit between observed and modeled data. The probabilistic framework provides feedback on the quality of the fit through model probabilities, allowing the system to evaluate and compare different rock physics models based on their ability to reproduce the observed seismic data.

Inventive Principle:
Principle #23Feedback

2Adaptability or versatility

If multiple rock physics models are applied to solve the underdetermined problem, then various solutions can be obtained, but the solutions cannot be evaluated against each other due to lack of performance assessment measures

Engineering Contradiction:
Improveability to apply different rock physics modelsVSAvoidprecision of model evaluation
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent introduces a model evaluation framework that provides feedback on the performance of different rock physics models. By calculating model probabilities based on the fit between observed and predicted data, the system enables quantitative comparison of multiple models, allowing users to select the most appropriate model for their specific geological context.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent changes the evaluation metric from qualitative assessment to quantitative probabilistic measurement. By expressing model performance in terms of probability distributions and goodness-of-fit statistics, the system enables precise comparison between different rock physics models and their predicted reservoir parameters.

Inventive Principle:
Principle #35Parameter changes

3Ease of operation

If the proximity factor is used to handle uncertainty in the inverse model solver, then the search can be performed, but the uncertainty handling is unrefined and does not well reflect information about uncertainty in reservoir parameters

Engineering Contradiction:
Improveease of performing inverse model searchVSAvoiduncertainty information in reservoir parameters
Core Design Contradiction:
Ease of operationVSLoss of information

Solution Approach 1:

The patent replaces the simplified proximity factor approach with a comprehensive probabilistic uncertainty framework. Instead of using a single scalar factor to control the search, the system employs probability distribution functions that capture the full range of uncertainty in reservoir parameters. This provides a more refined and informative representation of uncertainty while maintaining ease of operation through automated probabilistic inference.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3121622B1Method of predicting parameters of a geological formation
Publication Date: 2021.06.16 BERGEN TEKNOLOGIOVERFOERING AS
  • EP3121622B1 patent drawingFigure 1
  • EP3121622B1 patent drawingFigure 2
  • EP3121622B1 patent drawingFigure 3

AI summary

A method of predicting model parameters (R1, R2, R3, ...) of a geological formation under investigation, wherein said geological formation is distinguished by reservoir parameters including observable data parameters and the model parameters (R1, R2, R3, ...) to be predicted, comprises the steps of calculating at least one model constraint (M1, M2, M3, ...) of the model parameters (R1, R2, R3, ...) by applying at least one rock physics model (f1, f2, f3, ...) on the model parameters (R1, R2, R3, ...), said at least one model constraint (M1, M2, M3, ...) including modelled data of at least one of the data parameters, and applying an inverse model solver process on observable input data (d1, d2, d3, ...) of at least one of the data parameters, including calculating predicted model parameters, which comprise values of the model parameters (R1, R2, R3, ...) which give a mutual matching of the input data and the modelled data, wherein at least one of the input data (d1, d2, d3, ...) and the modelled data are provided with probability distribution functions, the inverse model solver process is conducted based on the probability distribution functions, wherein multiple predicted model parameters are obtained comprising values of the model parameters (R1, R2, R3, ...) which give the mutual matching of the input data and the modelled data, and model probabilities of the predicted model parameters are calculated in dependency on the probability distribution functions.