Analytical Approach for Drilling Optimization in Unconventional Assets
A ROP optimization methodology is presented, using offset drilled wells data. Data is used to train a neural network model for different rock properties which is presented in the vertical section of the reservoir. By changing the drilling parameters like Weight on Bit, Rate of Penetration (ROP) for the target well is optimized. Offset drilled wells data is used to train neural network; the inputs to the model are RPM,WOB (Weight on Bit) and Mud Flow Rate as well as ROP of the offset wells; the output is the new setpoints for input parameters to get the best ROP on the target well. Using offset well drilling data, rock properties of different formations drilled will be calculated explicitly. This will be used for bucketing data in the NN such that for each rock property, the best drilling parameters will be picked; these drilling parameters are then used on target well for best ROP on that specific formation. The technology is implemented in nine wells in Montney formation North of Alberta, Canada. The results are presented in this paper, among them the fastest well drilled to the lateral Kick of Point when compared with other offset wells on the same pad. The other outstanding result was the well with record low tripping for bit change to KOP. In total, based on the stakeholder feedback, the project had more than 5X ROI. The novelty of the analytic approach to ROP optimization is the capacity to use Azure Machin Learning Studio for real-time drilling operations optimization by changing the setpoints when the formation of rock property changes.
All content contained within this webinar is copyrighted by Stephen Park and Keivan Khademi and its use and/or reproduction outside the portal requires express permission from Stephen Park and Keivan Khademi.