Digging through Commingled Data

Recent thunderstorms in our area had caused a power outage but we managed to get back electricity after a couple of days ! Of course this typically happens only when the temperatures are soaring. As the electricity came back on, it was really striking to see how different the temperatures in each bedroom were. I was trying to cool the entire house using just one single temperature controller in the hallway. How efficient ?!

I couldn't help but reflect on the strange similarity with the way waterfloods in mature fields are managed especially with commingled production and multi-layer injection. It is common to see fields managed by using just fluid rates measured at the surface?- production and injection rates?(your hallway controller) !

Current status

You could argue, that's not entirely true. We do have Production Logging Tools (PLT - the equivalent of temperature sensors in rooms) to get us an estimate of layerwise production. Regulatory mandates in some regions require PLT for injectors at regular intervals to get a good estimate of how much is going into individual layers. Doing the same for producer wells means taking them offline (try convincing the asset manager !). In many cases where PLT data is unreliable or not most recent, assuming production in some proportion of horizontal and vertical permeability is also common.

Wouldn’t it be great if we knew how much is being produced from individual layers for every producer well? Are there specific layers that are watering out more than others so that we can selectively perhaps plug those layers? Surely, we can do without the needless water production at the surface !

What needs to be done

Capacitance resistance model (CRM) can be adapted to a multilayer reservoir to know the distribution of oil and water coming from individual layers for every single producer well. In order to adapt it to a multilayer reservoir, one transforms current injector and producer wells into multiple wells - one additional well for each layer it is producing or injecting. The fluid rates measured at the surface becomes an additional constraint. Of course the variables for nonlinear optimization blows up, but here is where you can use current advances in computations (GPUs, you listening) as well as combine with qualitative analysis (Yadav et al 2019 2020 for consistent solutions !

A multilayered CRM model helps determine fluid production coming from individual layers for every single production well.?This helps understand which layer is producing more water than the other producing layers. This also provides optimizing layerwise injection in the CRM framework (see figure). Throw in the residual oil saturation into the analysis and you can quickly see which layers have oil that may need a better sweep (next injector well location!).

Figure: Results of field implementation of multilayer CRM for a mature field (Dark shade bar is field data while light shade bar in the figure is optimized value from multilayer CRM). The layerwise WOR (water oil ratio) and oil production were obtained from the multilayer CRM model. Layer specific optimization suggests an increase in injection for layer 8 that reduces WOR from that layer by 10% and also gets more oil. This is adapted from our upcoming presentation in SPE ATCE Dubai 2021 (Davud et al 2021)

The next time you want to decrease water production from your mature fields producing in a?commingled fashion?– turn to your data ! We can do so much better than relying on a single hallway controller - in this context just the surface measurements of rates !