Johan Sverdrup’s subsea production system relies on multiphase pipelines transporting varying mixtures of oil, water, and gas from subsea wells to the topside processing facility. In this environment, flow behavior can shift quickly—particularly during transient operations such as choke adjustments, well startups, and shutdowns—making it difficult to distinguish a true leak from normal process variability.
Equinor required a dependable, highly sensitive leak detection capability that could meet strict Norwegian requirements for rapid detection, including small leaks, and operate continuously. Traditional approaches that depend on limited sensor signals or mass-balance calculations can be vulnerable to noisy measurements and inevitable differences between predicted and actual multiphase behavior, especially during transients. These limitations created operational risk: Missed events, unnecessary alarms, or avoidable shutdowns that disrupt production.
SLB implemented a multidimensional LDS built around a real-time transient model for each individual pipeline. The system uses a broad range of measurements—pressure, temperature, and flow—from both subsea and topside sensors. Before data is used by the transient model, all incoming measurements pass through an automated validation layer to detect anomalies and ensure the model receives reliable inputs.
The LDS combines validated measurements into multiple leak signatures that characterize how the pipeline should behave under normal conditions. These signatures are organized into expected operating regions (EORs) that adapt to operating state. When conditions shift from steady-state to fast transient periods, the system transitions from a normal operating EOR to a transient EOR, which helps reduce avoidable alarms without compromising detection performance.
Additional features support robust performance across a wide range of scenarios, including automatic enabling and disabling of leak signatures based on valve positions or data quality, moving-average filtering to dampen measurement noise, and virtual flow metering at the wellhead and pipeline outlet. Virtual metering helps maintain situational awareness when physical flowmeters are unavailable.
With the multidimensional LDS in place, Equinor achieved improved sensitivity and reliability across all four pipelines. The system maintained more than 99% availability; interruptions were limited to brief model restarts or temporary communication issues. Avoidable false alarms were reduced to fewer than two per pipeline per year, minimizing the likelihood of unnecessary operational responses.
Regular leak detection tests using live and replayed historical field data demonstrated consistent performance across leak-size categories. Leaks greater than 30% of average pipeline flow were detected within 30 minutes, while 20% and 15% leaks were detected within one and two hours, respectively. During steady-state operation, leaks of approximately 10% were detected within two hours. Ongoing refinements are intended to further improve sensitivity toward reliable detection of leaks below 5% within two hours in multiphase production pipelines.
Equinor now benefits from a stable and trustworthy leak detection process for multiphase pipelines that supports compliance with stringent environmental regulations and increases operator confidence in day-to-day decision-making. Because the LDS adapts to operating state and uses continuous validation of inputs, it can evolve alongside changing field conditions.
Reference: Technical paper ISAVFT-2025-533.
