Joel Gordon and Steven Greene, PetroBench
Jaime Hecht and Ferdinand Hingerl, Ambyint
Ambyint and PetroBench have each independently developed proprietary downhole equations that provide a better representation of deviated wells compared to the vertical downhole card's prediction model. The deviated well downhole card's prediction model can:
Consider the well geometry to calculate the rod state (velocities, accelerations, forces and stresses) along the well trajectory. These specific forces are represented as axial, normal and binormal.
Calculate the dog severity and predict possible rod/tubing contact along the well trajectory.
Calculate the lateral load on the rod due to rod/tubing contact. This load is a combination of the normal and binormal forces expressed in vector forces.
Improve the axial force calculation by considering the friction force (Coulomb force) calculation due to rod/tubing contact.
Identify maximum and minimum axial, normal or binormal forces positions during the rod cycle.
Both companies have successfully deployed their proprietary downhole equations into commercially available software. Ambyint has been deployed on thousands of wells over 15 years, while PetroBench's has been deployed on hundreds of wells for 5 years.
In an effort to validate the outputs of their respective models, the companies have both implemented their models on 35 common wells under a number of different scenarios. Despite being developed completely independently, the outputs of the two downhole models are very similar in some parameters with some slight discrepancies in other parameters. This paper aims to highlight the similarities and differences between the two models.
In summary, both models had qualitative similar outputs on Side Loading, Dynamometer cards and absolute values of the surface card. The models demonstrated a slight difference in the load harmonics near the top of the stroke, most likely due to low friction assumptions in the PetroBench's simulation. One noticeable difference between the two models was in the pump fillage calculation, where PetroBench's graphed pump fillage visually showed a lower value than what was reported.
Although the ideal validation exercise would include comparison of results to downhole sensors capable of generating a real-world downhole dynamometer card, the validation exercise described above serves as a reasonable substitute. The nearly identical model outputs observed between the two downhole models should provide users of the respective models with some confidence in the accuracy of those models.