Quality Control of LWD Resistivity Data Inversion Results

Drilling operators very often perform reservoir navigation and mapping using extra-deep resistivity tools. Tool responses depend on formation properties tens of meters away from the wellbore and require sophisticated processing by inversion to provide operators with a multilayer resistivity model. The accuracy and reliability of inversion results are very important and need thorough assessment. We present two new methods of inversion quality control, validate their applicability, and provide a comparative analysis with existing methods on several synthetic and field cases. Deterministic and statistical methods of estimation of resistivity, tool detection, and resolution capabilities are applied to evaluate the quality of inversion results. We discuss tool ability to detect single boundary, depth-of-detection (DOD) and depth-of-reliable-detection (DRD) concepts based on covariance matrix analysis, and introduce a new method of DOD estimation based on resistivity model perturbations, with posterior tool response monitoring. We propose a new statistical resolution analysis method related to response-surface technique and compare its results with other approaches. The applicability of the methods considered is validated by guided inversion for typical job stages (pre-well, real-time, post-well) and applications (landing, reservoir navigation, mapping). Inversion results for extra-deep logging-while-drilling (LWD) resistivity tools are usually shown as a multi-layer resistivity distribution map or picture, without a clear indication of the uncertainty of the structures presented on the picture. The uncertainty of inversion results depend not only on tool specifications (i.e., frequency range, electronic noise level and antennae spacings), but on the complexity of surrounding formations as well. The new method for DOD estimation deals with model complexity and gives several estimates based on different subsets of measurements. Common approaches to inversion result quality control only provide partial reliability indicators, usually around the final inverted model. The suggested resolution analysis method generates a statistic from models assessed during inversion execution, analyses it, and eventually provides the resolution accuracy of formation parameters. The method enables identification and quantification of disconnected uncertainty regions, when they exist, thus ensuring an exhaustive analysis of the parameter space. Based on synthetic and field cases considered, we conclude that understanding of uncertainties associated with reservoir navigation requires the application of several data analysis techniques. Complementary use of data inversion, DOD estimation and resolution analysis yield a comprehensive evaluation of the environment and show the realistic capabilities of the tool. The developed methods enabled the implementation of scenario-oriented workflows that deliver not only the final resistivity model but also its reliability indicators. The paper will show how to interpret and evaluate the quality of inversion results provided by vendors. Two new methods to evaluate the result model extend the capability to analyze uncertainty from several different perspectives. Better understanding of the inversion deliverables with the reliability indicators will help the operators to make more confident decisions during reservoir navigation, or posterior oil field development.