Elements of an Accurate Production Forecast

May 11, 2016 W. John Lee, PhD, Research Fellow
Decline Analysis, Oil and Gas Forecasting, Reservoir Engineering

LinkedIn twitter facebook Email

Extensive studies show that production forecasts, the basis for reserves estimates, are more reliable when they are based on certain simple elements. In unconventional resources, such as shale plays, these elements include the following:

  • Identify flow regimes in historical production data, such as early “transient” flow unaffected by reservoir boundaries and later “boundary-influenced” flow.
  • Apply appropriate decline models, such as Arps’ hyperbolic decline model, separately to each flow regime.
  • For forecasting, give dominant consideration to the most recent flow regime – we should not try to match all historical production data with a single decline model.
  • If changes in flow regime are expected in the future (e.g., when reservoir boundaries become important), it is essential to estimate when the flow regime will change and what the characteristics of the new flow regime will be.

There’s been a lot of work done in the field of decline curve analysis, particularly in the last ten years with the increased importance of unconventionals. I will examine some of the findings of this work that show the elements of accurate forecasting.

It is very important to identify distinct flow regimes in production data. The major flow regime at the beginning for wells with really low permeability (i.e., horizontal wells with multiple fractures), is going to be transient flow – flow that has not been affected yet by reservoir boundaries. Then later, the second major flow regime is flow that is influenced by the boundary. The decline models for these different flow regimes will be different. We can’t use the same decline model for both the early transient flow and the later flow influenced by the boundary if we want to be accurate.

What we have to do is find parameters in decline models (e.g., Arps hyperbolic) separately for each flow regime. Then, weight the most recent data (and flow regime) most heavily. If early on we saw transient flow and then we saw the influence of boundaries later, it helps our understanding to know that we saw the transient flow early, but we don’t want to give it much, if any, weight in matching later data influenced by the boundaries because that’s what we need for forecasting. We need to model the later data accurately.

If we expect the flow regime to change in the future from what it is now, then we need to do the best we can to figure out when that change is going to occur and include that in our forecasting model.

Finally, you need to ensure that the decline curve analysis software that you use includes all these elements.

How Do We Identify Flow Regimes?

If the first step in an accurate product forecast is to identify flow regimes, how is that done? Even though it’s an extra step in decline analysis, it’s worth spending the time to do it.

In the figure below, the different flow regimes are identified with characteristic signatures on this kind of plot. The transient flow early is almost always linear flow, a straight line with a slope of one-half. So if we just make this plot and identify this, we can find when we’re in transient flow and model that appropriately.

The flow regime that’s most important is the boundary-dominated flow — the characteristic shape once the full effect of the boundaries have been felt. On this kind of plot, we get a unit-slope: a straight line with a slope of one, double the slope we saw earlier. This is pretty-straight forward; an extra step that in my judgement is well worth taking.

What’s important to remember about these flow regimes is that what happened at the beginning really doesn’t matter. We need to project the later trend into the future. So if you’re in the middle, where there’s still some linear flow, you need to decide the time to switch from it to the boundary-dominated flow and use that for forecasting. If we don’t, our forecasts won’t be as accurate.

Fitting Final History, Forecasting

This leads us to the final step, forecasting. In this example, we’ve begun to see the effects of the boundaries, but only just.

We’ve got to make a judgement. In the example, there are two different forecasts: a homogeneous forecast (in red) with an Arps b of 0.5 and a heterogeneous forecast (dotted pink) to use if we know the geology, using an Arps b of 0.8. It’s a judgement call, but we need to be conscious of the fact that it will affect the final answer.


To ensure more accurate production forecasts, we need to:

  • Identify flow regimes – this is the single most important step.
  • Fit final regime (the most recent history). What’s happened earlier really doesn’t matter that much.
  • While forecasting:
    • In boundary dominated flow, select most appropriate final ‘b’ value, which will come with experienced judgement and analogy.
    • In transient flow, choose end of flow regime, and use your best judgement as to what the final b value will be.
    • Use analogs (preferred) or models to assist with choices.

These are extra steps that are necessary to ensure the most accurate forecast possible. To best achieve these elements, you want to make sure that your decline curve software can help you do these things quickly and accurately. If you don’t have the necessary tools to help you in this area, the benefits of accurate forecasting suggest that you should consider getting them.


About the author: A world-renowned expert in the field of petroleum reservoir engineering, John Lee, holds BS, MS and PhD degrees in Chemical Engineering from The Georgia Institute of Technology and has been a leader in our industry for over five decades. He is best known for his recent publications and presentations on oil and gas reserves regulations and estimation, and production forecasting in unconventional gas reservoirs. He served as an Academic Engineering Fellow with the U.S. Securities & Exchange Commission (SEC) and was a principal architect of the modernized SEC rules for reporting oil and gas reserves.


comments powered by Disqus