Complex geometry effects observed on log responses in high-angle and horizontal (HaHz) wells can create challenges in achieving accurate formation evaluation. The situation is made even more difficult in thinly bedded reservoirs where measurements may respond to multiple layers within their volume of investigation. Recent publications have outlined techniques in which a layered earth model is used to define the geometry of layering relative to a wellbore. A model-compareupdate workflow is then used to solve for layer properties. Although these techniques are efficient in horizontal wells, they require good geological understanding to manually create the formation model and can be time consuming if there are many thin layers. This paper presents a semiautomatic method to construct a layered earth model in the immediate proximity of the borehole and to solve for the formation properties and geometry of the layers locally. The approach is particularly useful for shallow-reading measurements and complements the more extensive layered earth models commonly used for modeling the response of deeper reading measurements.
In this method, high-resolution density and volumetric photoelectric factor (PEF) images, acquired by logging while-drilling (LWD) tools while rotating, are analyzed to define the boundary position, dip, and properties of layers as thin as 2 inch. The computed dips, layer boundaries, and log values are then used to automatically create a local layer model (LLM) and provide an initial estimate of the density and PEF layer properties in the high-angle intervals. In the horizontal intervals where layer boundaries do not cross the wellbore axis, the user completes the LLM manually using the density image projected onto the well trajectory as an aid.
Once the LLM has been created and populated with formation properties, the corresponding logging tool measurement response is simulated by a fast-forward modeling algorithm. Through a model-compare-update workflow, the user adjusts the formation model so that the forward modeled logs and images reasonably match the measured responses. In many cases, especially in beds thicker than 6 inch, little to no manual adjustment is required.
When modeling log responses in HaHz wells there is not always a unique solution to the problem. There are two main unknowns in the layer-cake formation model:
the layer geometry, that is to say the boundary positions and dips, and
the layer petrophysical properties.
In this new workflow, the layer geometry is clearly defined by the wellbore images, leaving the layer properties as the main unknown. In many cases, the layer properties can be read directly from high resolution measurements such as density images, but this is not always the case in thin beds or nonplanar layers or when lower resolution non-azimuthal measurements are interpreted (e.g., neutron porosity). By enforcing a common formation geometry and matching simulated logs that take into account both geometry and formation properties with measured logs, the described workflow significantly increases the confidence in the computed layer geometries and properties.
The methodology is demonstrated on two high angle/horizontal wells, one from offshore West Africa and the other from the North Sea. The paper shows how the LLM is quickly created and updated to provide a formation model proximal to the wellbore. The rapidly created LLM provides information about formation geometry, which facilitates determination of the true properties of thin layers, free from the geometry effects that are observed on the measured logs. The true layer properties enable more accurate formation evaluation than use of the geometrically uncorrected measured logs.