Integration of open and cased hole log data provides valuable insights

In this article, I present a case history that shows how integrating open hole log analysis with Production Log data can add valuable insights into permeability prediction.

As always, the analysis (both open hole and cased hole) was performed in Excel, and Python scripts were used to create the graphics.

I downloaded the log data from the UK's National Data Repository (UK NDR - National Data Repository (nstauthority.co.uk)). This is an excellent source of a wide variety of data types from wells drilled in the UK Offshore sectors.

Overview

The well I used was drilled in the UK Central North Sea in the early 2000's. I suspect it is an appraisal of a prior discovery, as the borehole is deviated by up to 34 degrees across the zone of interest. It found a thick oil column, and the Operator decided to run a production test. The interval flowed 43 degAPI oil at 4,660 b/d, with 3.61 MMscf/d of gas (GOR 800 stb/scf) and no water. The test operations included running a Production Logging tool that included spinners, fluid density, pressure and temperature sensors.

Open hole logs and log analysis

The Operator ran an extensive suite of open hole logs, using both LWD and wireline conveyance. They also took numerous wireline formation tester pressure tests and samples, but did not take any cores, either whole core or sidewall core.

This log display shows LWD triple combo curves (GR, resistivity, density and neutron) in the first three tracks from the left, a volumetric log analysis I performed in Excel in track 4, total and effective porosity in track 5, total Sw in track 6 and some permeability estimates in track 7.

The seal in this formation is a tight carbonate, and there are some carbonate streaks within the reservoir. Porosity varies between 15% and 25%, while oil saturation is between 70% and 90%.

I used two different methods to estimate permeability. The first uses Vsh as a proxy for Flow Zone Indicator (FZI). The FZI curve is then inverted and used with porosity to estimate permeability. This method works well for clastic reservoirs in which grain size controls permeability. It is not as reliable in clastics with cementation and other diagenetic effects.

The second is the Coates-Timur method, which combines porosity with free and bound fluid volumes. It is commonly applied to NMR log data, but that's not available in this well. Instead, I assumed that the whole interval was at irreducible water saturation, and used log-derived porosity and Sw instead.

Both methods give comparable results, and are broadly consistent with the formation tester drawdown mobility data too. So which do we use for our reservoir model? That's where the Production Log data can help.

Production Log data and analysis

I won't go into a lot of detail on how Production Logs work and how they are interpreted, but a little discussion of the data is helpful here.

The key measurement in this study is the fullbore spinner, or flowmeter. An example of this type of tool is shown in the image at the start of the article. Spinners measure the relative fluid velocity inside the borehole. Just like a pinwheel on a bike, when the tool is moving they rotate even if the fluid column is static. The rotation frequency (in revolutions per second) depends on a number of factors aside from the tool and fluid column velocity. These include fluid viscosity, spinner blade design and friction in the tool's bearings. That's why we can't use the spinner to directly determine fluid velocity, imstead we have to calibrate the response in each well. This is done by making a number of passes, both up and down, at varying logging speeds. The passes are recorded under both shut-in and flowing conditions. The next chart shows some of the data recorded in this well.

The first two tracks from the left show the spinner passes at varying speeds up and down, and the fluid density with the well shut in. The black shading shows the perforated interval. The spinners show a lot of noise, probably due to irregular movement in the hole. The fluid density data shows a clear oil/brine interface in the wellbore at 11,825ft. This is not a formation oil/water contact, just the point where the cumulative well inflow was strong enough to lift out the completion brine. This is seen in most PL surveys, although typically the interface is at the base of the perforations. While the spinner data is noisy, it can still be used to calibrate the tool response.

The next two tracks show the same data with the well flowing. There are some very interesting features in this data. First, note that below about 11,850ft, the spinner and density response while flowing is the same as the shut-in. This means that there is no inflow. Next, above 11,700 ft the density is the same as the shut-in, which means that the well is not flowing any appreciable amount of water. This is confirmed by the test separator data. The interval between is a little more complex, however. From the density data we can see that the wellbore oil/water interface has moved up hole, and is much more smeared-out, with no clean break from water to oil. In the same zone, we see that the spinner passes initially move to the left, which indicates a net downwards flow. They then move to the right, showing a net upwards flow (or production).

So what's happening here? We have water moving, but not produced to surface, and the spinners showing that fluid is actually going downwards over a short interval in a producing well. My interpretation is that there is oil production from below the shut-in oil/water interface, but not enough to lift out the brine. Instead, the brine is recirculated inside the wellbore, being lifted up by the oil and then dropping back down. This effect is enhanced by the hole deviation, which allows the oil to rise to the high side of the casing under buoyancy. The spinners see the down flow of the brine but not the up-flow of the oil, because it is riding along the top of the pipe. I have seen this effect in other Production Logs from deviated wells.

For the purposes of this study, I assumed that the net downflow of brine seen by the spinners was at the same rate as the up flow of oil. This allowed me to "flip" the spinner response in the brine recirculation zone, giving a continuous oil flow profile.

Open and cased hole integration

The Production Log data gives us the actual inflow profile for the perforated zone, but we can also predict the profile from our permeability analysis, by simply summing the curve over the perforated zone. In my experience the log and PL profiles match in wells drilled and tested before field production has started, and when they do it gives me confidence that the log curve is a good measure of permeability heterogeneity. You can also compare permeability thickness from Pressure Transient Analysis with the integrated log curve, but that's outside the scope of this study.

This plot shows the measured inflow profile from the Production Log data with the inflow predicted from the two permeability models I generated.

The last track on the right shows the two different permeability models (Vsh and Coates-Timur) with the observed profile from the Production Logs. To me, it seems that the Coates-Timur model (in red) is a better match to the Production Log (in green) than the Vsh model (in black). The Vsh model predicts a relatively uniform inflow, while the Coates-Timur model identifies some hot-spots, which the Production Log confirms.

I think it's fair to say that this analysis shows the Coates-Timur method to be more reliable, and it should be used for geocellular and flow simulation models in preference to the Vsh method. Without the Production Logs we would not have known this.

Thanks for reading this to the end! I hope you found it interesting and informative. If you have similar data sets that you would like to discuss with me, or other related topics, please feel free to connect on LI. I'm available to provide contract petrophysical services on a wide range of data types in addition to what I presented here.

My thanks to Neil Hopkins and Geoff Whiteley, who helped me decode the PL data.