-
Recent data is more important than older data
and should be more heavily weighted in the curve fitting algorithm because
the near term portion of the production forecast is more important (from a
present worth perspective) than the far-in-the-future portion of the
forecast due to time-value of money considerations.
-
Production history may be "segmented"
meaning that the production history can be divided into distinct time
periods when the well produces under a consistent set of conditions,
producing a period of related (and stable) performance
characteristics. The set of well conditions may change abruptly
creating a new recognizable period of related well performance behavior.
-
Production data scatter may contain
artificially low points such as periods of market curtailment, etc.
unrelated to the well's physical ability to produce. Such non-natural
low data points must be excluded from the fitted data set since they
(generally) have no offsetting artificially high production points (except
for cases of flush production). In the cases of artificially low
production where reservoir recharge can be observed (i.e. constrained (low)
production followed by artificially high flush production rates), curve
fitting algorithm should consider retaining the low points in the fitted
data set to counter balance the artificially high flush production points.
-
Oil sales may be sometimes be reported in lieu
of production. For low volume oil leases, this type of volume
reporting often results in monthly volumes that follow a repeating pattern
of months of no oil pickups (i.e. zero sales volume), followed by months
with sales volumes equal to approximately the lease tank volumes. Such
a pattern should be recognized as "sales data" where the fitting
algorithm retains the zero monthly values and uses them in the curve fit
data set.
-
Production decline functions exhibiting decline
rate decay over time (i.e. reduction in decline rate over time) generally
decay to some limiting minimum decline rate (i.e. projection rate method
changes to constant percentage decline). The point of this transition
should be automatically determinable (if it has occurred).
-
Must be able to detect whether the decline rate
is constant or is decaying (i.e. whether the curve exhibits exponential
decline or hyperbolic behavior). This determination applies to a
single segment and may include detection of a transition to constant decline
(at the minimum decline rate).
-
Projection from short term data segment (where
duration is too short to accurately predict future performance from that
data alone) should use longer term "curve shape" from historical
segment(s) as guide to defining the "shape" of the projection from
the new short term segment.
