American Oil and Gas Reporter - February 2019 - 41

to t1 are assigned a weight of W3. Finally,
historical data are assigned the lowest
weight of W4. The sum of the weights
equals 1. The t1 and t2 times for selecting
the points are not fixed, and their values
depend on the total number of data points
collected at the current time instant.
One of the main states is the drilling
time, which is calculated utilizing an
equation that includes factors such as
total depth (combination of individual
depths to be drilled by slide and rotary
modes based on the well plan), and formation-specific drilling data from offset
wells. The final predicted time to TD at
any given point is estimated by the sum
of the predicted times spent in each of
the various rig states.
Bakken Test Application
The proposed algorithm was tested on
data from a four-well Bakken pad that
consisted of one-second surface and downhole sensor data, formation top information,
the well plan, and BHA/bit run summaries
from daily morning reports. For these
wells, the time to TD prediction algorithm
was implemented on the horizontal sections,
all of which were drilled in the same formation. Three of the wells were selected
as offsets and historical data from the
wells were used to make initial predictions.
Descriptive statistical analysis was performed to formulate the initial conditions
for running the prediction routine.
The fourth well then was simulated as
a real-time well, and the data at every
time step (one second per data point)
were plugged into the algorithm in one
iteration of the conditional loop. Figure 1
demonstrates the process by which datadriven models for initialization were built
using the data from the three offset wells,
and then predictions of the time to TD
were made based on continuously updated
real-time data from the fourth well.
The rig-state-detection algorithm run
on data from the three historical wells
used basic "channels" typically available
in any drilling dataset, such as hook load,
block height, RPM, ROP, surface torque,
etc. Then, using the determined rig states
and certain predetermined data channels,
statistical estimations were made about
the time spent in the different states.
Figure 2 shows a scatter plot of all the
survey and connection times in one of
the offset wells. Based on the times in
each state, the mean, maximum and minimum times were estimated for all wells.
The next step was initializing the algorithm with the time estimates and
planned TD. To begin implementing the
algorithm, the only additional information
required was the starting and ending
depths of the section, and the historical

data-based statistics.
After initialization, the real-time data
for the fourth well were used to make
predictions. Based on these data channels,
rig states were identified and the next
steps were performed to estimate the individual times for the different well delivery tasks: on-bottom drilling, tripping,
making/breaking connections, etc. The
total time to TD at any point then was
calculated as the sum of the times predicted
for the individual well delivery tasks.
Figure 3A shows the times that were
estimated to make the predictions. Figure
3B illustrates the final plots generated

for the time-to-TD predictions.
The total time at any point is the sum
of the predicted times for the individual
well delivery tasks at that point. The
figure shows three individual runs and
two trips. The tripping time was projected,
based on the tripping times per foot estimated from the historical data and the
depth at which the trip was identified by
the algorithm. Therefore, tripping times
depend on the depth at which a tripping
operation is initiated.
The figure also illustrates that the plot
of the predicted time is not smooth. This
is because of the different weights assigned

A New Day
for Completion Optimization
Simple | Accurate | Affordable
IMAGE Frac™ pressure-based fracture maps
FracEYE SM frac hit analysis service
Analytics - well, stage data insights

info@reveal-energy.com
1.888.738.3250
reveal-energy.com/aogr
Copyright © 2019 Reveal Energy Services. All rights reserved.

FEBRUARY 2019 41

http://www.reveal-energy.com/aogr

American Oil and Gas Reporter - February 2019

Table of Contents for the Digital Edition of American Oil and Gas Reporter - February 2019