Interview with an Orinoco Oil Belt Engineer on AI Implementation in Artificial Lift Systems

Interviewer: How do you evaluate the implementation of intelligent artificial lift systems in the Orinoco Oil Belt, and what is your opinion regarding the Hydrog solution with artificial intelligence?

Engineer: Your concept is excellent and addresses very relevant aspects. To understand the true value of this technology, it is essential to compare it with conventional hydraulic pumping systems from the past.

In traditional hydraulic pumping systems, although it was possible to adjust parameters such as stroke speed and stroke length — both on the upstroke and downstroke — these values were set in a fixed manner without continuous supervision. Once the parameters were established, the machine operated invariably, without adaptation capability.

The incorporation of artificial intelligence in these systems represents a radical change. The main problem with conventional hydraulic systems is their vulnerability to failures: if there are no measurements enabling timely decision-making, the equipment is destroyed. This has been our field experience.

We have had equipment completely destroyed due to the inability to anticipate changes in critical parameters such as mixture density, temperature, plugging, or misalignments. Without predictive capability regarding downhole conditions, the system continues operating despite failures, resulting in bent, collapsed, or severely damaged units.

For example, a failure in the diluent system causes the fluid to become denser. The conventional hydraulic system does not detect this anomaly and continues attempting to pump, or it lowers the sucker rod string believing it is executing the operation normally, when in reality the rod string is already floating or stuck in a fixed position.

This type of failure was impossible for conventional units to detect. Frequently, when the sucker rod was floating and the unit initiated the downstroke without detecting that the rod was descending simultaneously, upon attempting the upstroke, the wireline would coil, generating catastrophic damage.

The incorporation of automation through real-time readings and parameters not only enables efficient production but also significantly reduces downtime associated with failures in high-cost equipment. This is a significant technological contribution.

Just as operators require a constant crude oil production rate, they also need to minimize failure times. A shutdown of two units represents considerable economic losses: due to deferred production and material damage to equipment.

Therefore, it is essential that these units have information capture systems enabling preventive decision-making, such as stopping the installation before a catastrophic failure occurs. This protects assets, reduces service times through preventive shutdowns, and allows for timely corrective actions.

This problem is not exclusive to hydraulic pumping; it also occurs with sucker rod pumping systems and progressive cavity pumps (PCP). All these artificial lift methods must be linked to monitored parameters. However, historically we have not had the capability to automate and control downhole variables to compensate for failures and reduce response times that prevent equipment destruction.

For example, a failure in the PCP or in the diluent system immediately generates an excessively heavy fluid column. This causes pump sticking, and if the motor does not have adequate protection, belt slippage occurs, circuit breaker trips, and immediate destructive failures: sucker rod parting, stator detachment, damage to internal components, belt slippage or rupture, and in systems with gearboxes, electrical system trips.

These events generate significant lost time: surface equipment repair, production loss, downhole repair, and system restart. In the past we constantly faced these problems, and they still persist today in non-automated wells.

Solutions based on artificial intelligence are cutting-edge technologies. Although they require considerable upfront investment, they are amortized over time through reduced service times, prevention of catastrophic failures, and maintenance of stable and continuous production.

During the first decades of PCP implementation in the Orinoco Oil Belt in the 1990s, the sucker rod string connected through threaded connections worked in the direction of thread tightening. However, when the pump stuck, the rod continued rotating with the motor. Although there was a system at the wellhead designed to prevent reverse rotation, cascading failures were inevitable, generating considerable lost time.

In sucker rod pumping systems, the situation was relatively simpler because the rod string was not subjected to torsion, only to tension and compression. However, when a failure occurred, generally in the beam pumping unit, the consequences were equally destructive. The pumping unit continued operating in an uncontrolled manner, generating violent impacts against the wellhead, until it was completely destroyed.

In hydraulic pumping systems, it was common to find units collapsed on the ground, still running, without the ability to detect actual downhole conditions. The failures were totally destructive, leaving the equipment unusable.

The hydraulic units we operated, Rotoflex type, used an elastomeric belt wound on an upper drum, with a metallic element at the end connected to the sucker rod string. Although they allowed control of stroke length and operating speed, they lacked the supervision and dynamic adjustment capability that intelligent systems now offer.

In conclusion, the Hydrog solution with artificial intelligence represents a fundamental advancement for operations in the Orinoco Oil Belt, addressing the historical limitations of conventional systems and offering a technically and economically superior alternative.

Publicado originalmente en Medium: https://medium.com/@francisco.l.r.a/interview-with-an-orinoco-oil-belt-engineer-on-ai-implementation-in-artificial-lift-systems-3c48741cfee0