Iterative Clustering for Geosteering Inversion Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional electromagnetic resistivity inversion techniques for directional drilling and wellbore placement in hydrocarbon exploration often fail to accurately account for noise in model data, leading to suboptimal clustering of inversion models and reduced accuracy in estimating formation properties.
Innovation Solution
The implementation of a modified K-means clustering technique that iteratively re-clusters inversion models using average models as centroids, enhancing the accuracy and convergence of clustering results by reducing noise and improving the selection of optimal inversion models for geosteering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional clustering techniques with random selection are used, then the process is simple, but the clustering accuracy is reduced due to noise in model data
Solution Approach 1:
The patent applies preliminary action by performing an initial clustering step to generate average models before the main clustering process. These average models serve as informed centroids for subsequent clustering iterations, preventing random selection and reducing noise impact from the outset.
Solution Approach 2:
The patent implements feedback through iterative re-clustering where clustering results generate new average models that become centroids for the next clustering round. This feedback loop continuously refines cluster assignments and reduces noise impact, improving clustering accuracy progressively.
2Adaptability or versatility
If multiple inversion models are generated to account for formation heterogeneity, then the coverage of formation properties is improved, but the noise in model data increases
Solution Approach 1:
The patent applies merging by combining multiple inversion models into clustered groups and generating average models for each cluster. This consolidation reduces the impact of individual model noise while preserving the diversity needed to cover formation heterogeneity, achieving both goals simultaneously.
3Reliability
If iterative re-clustering with average models is performed, then the clustering convergence is improved, but the computational time increases
Solution Approach 1:
The patent performs preliminary clustering to generate average models before the main iterative process. This preliminary action provides better initial centroids that guide subsequent iterations toward convergence more efficiently, reducing the total computational time despite the iterative nature of the method.
Solution Approach 2:
The patent extracts key representative models (average models) from each cluster to serve as centroids for subsequent iterations. This extraction reduces computational complexity by working with condensed representations rather than all individual models, maintaining convergence reliability while reducing time loss.
Data Source
AI summary
System and methods for geosteering inversion are provided. Downhole tool responses are predicted for different points along a planned path of a wellbore during a downhole operation, based on each of a plurality of inversion models. Measurements of the downhole tool's actual responses are obtained as the wellbore is drilled over the different points during a current stage of the operation. The inversion models are clustered based on a comparison between the actual and predicted tool responses and a randomly selected centroid for each cluster. The inversion models are re-clustered using an average inversion model determined for each cluster as the centroid for that cluster. At least one of the re-m clustered inversion models is used to perform inversion for one or more subsequent stages of the downhole operation along the planned wellbore path. The planned wellbore path is adjusted for the subsequent stage(s) of the downhole operation.