Introduction
Within the Predico software, there are two types of declines generated using data science and/or augmented by Outlier Detection Techniques and/or Unsupervised Machine Learning. These forecasts are:
-
Profile Forecast:
-
Forecasts generated based on more global trends observed in the data.
-
-
Operational Forecast:
-
Forecasts generated by looking at the more recent data trends.
-
Limitations of Decline Methods
In early life of a well, decline analysis can be highly non-unique (I.e., high uncertainty) hence the low, medium, and high predictive capability of AFA. In the case of unconventionals, it has often been recommended that at least 18 months of data is required.
Nonetheless even with a reasonable decline forecast, a change in operating conditions means the existing forecast is now erroneous.
Obviously, if there is no decline in the rate data, then decline analysis cannot be easily performed at all, and perhaps other methods such as Rate Transient Analysis or PSS Modelling should be undertaken.