|Hybrid data driven drilling and rate of penetration optimization
|Year of Publication
|Alali, A. M., M. F. Abughaban, B. M. Aman, and S. Ravela
|Journal of Petroleum Science and Engineering
|Data-driven, Drilling mechanics, Machine learning, Optimization, Real time adjustments, ROP
Optimizing the drilling process for cost and efficiency requires faster drilling with a higher rate of penetration (ROP). A high ROP usually indicates fast and cost-efficient drilling operations; however, high ROPs may induce unintended events, which may lead to nonproductive time (NPT). In this paper, the proposed hybrid, data-driven optimization system aims to improve the drilling process, maximize the ROP, and minimize NPT. This two-phase system leverages existing drilling data (Phase One) and real-time adjustments (Phase Two) to optimize drilling efficiency through the adjustment of controllable dynamic drilling parameters; weight on bit (WOB), pump flow rate (GPM), and rotary speed (RPM). In the first phase, geological data is queried to generate a model of drilling best practices in a certain field with attention to ROP and NPT. After the top-rated wells are identified, the model recommends adjustments to the controllable dynamic drilling parameters (WOB, GPM, RPM) and produces a cross-plot for each ROP. The optimum ROP parameter for each well is subsequently calculated as a conditioned mean based on proximity to the well as an Inverse Distance Weighting (IDW) interpolation. This phase is completed prior to any drilling activity. The second phase considers the model produced in phase one to run an automated drill-off test. The model directs the driller to deviate for controllable parameters WOB and RPM by a small percentage (0–5%), in a Constrained Random Search (CRS) way. These deviations allow the driller to explore new drilling conditions in small increments; the data is then compiled into a heat-map function of ROP and the model is updated to reflect optimum drilling parameters. The adjustments resulted in faster, safer ROP with a potential 22%+ improvement in efficiency and 15–20% cost-reduction. Lastly, to characterize the relationship between the controllable dynamic parameters (WOB, RPM, GPM) to the ROP, an ANN based empirical relationship was derived.