33 hours optimization work in 10 seconds with 96% accuracy

Physics-informed AI, the SLB technology that combines the rigor of physics-based modeling with the speed of machine learning. The model, trained on more than 19,000 high-fidelity simulations and delivers optimization results in under 10 seconds—compared to 33 hours using conventional simulators—while maintaining over 96% accuracy.

Maximizing revenue under regulatory pressure and fluctuating prices

In a complex onshore gas network comprising 80 well pads and four delivery points, operators faced the challenge of maximizing revenue while navigating fluctuating gas prices, operational constraints, and regulatory pressures. Traditional physics-based simulators, though accurate, were too slow for daily decision-making, while pure machine learning models struggled with data quality and completeness.

33 hours work in 10 seconds with 96% accuracy

To overcome these limitations, an approach that combined the rigor of physics-based modeling with the speed and adaptability of machine learning (ML), was developed. This surrogate physics-informed AI (PI-AI) model, trained on more than 19,000 high-fidelity simulations, delivered optimization results in under 10 seconds—compared to 33 hours using conventional simulators—while maintaining over 96% accuracy.

PI-AI delivering fast, reliable, and scalable solutions

The solution enabled real-time optimization of gas distribution across multiple scenarios, including compressor failures, reduced sales point capacities, fluctuating gas prices, and the integration of new well pads. Operators could rapidly evaluate “what-if” scenarios and adjust choke settings to meet production commitments and pressure constraints while maximizing revenue. The intuitive interface and robust model execution facilitated seamless integration into daily workflows, transforming how operational decisions are made.