Hybrid AI: Trusted Insight for Pipeline Performance and Efficiency
Pipeline operators are sitting on vast amounts of data, yet much of it goes underutilised. As the sector faces growing pressure to improve performance, cut costs, and reduce energy use, hybrid AI offers a compelling way forward.
Energy Efficiency: A Critical Opportunity
Pipelines are among the most efficient means of fluid transport, but their pumps are heavy energy consumers. In fact, energy use can account for up to 85% of a pump’s lifecycle cost (Sulzer, 2020). Improving energy efficiency represents a major opportunity, but it’s not straightforward.
Operators face the challenge of extracting insight from complex, often siloed SCADA data. These systems were never designed for advanced analytics, leading many to turn to AI in search of deeper operational insights. AI can suggest real-time changes that drive meaningful improvements, but only if operators trust the output.
Why Hybrid AI?
Trust is central to AI adoption in safety-critical environments. That’s why Klarian focuses on hybrid AI - a method that blends statistical learning with deterministic, physics-based models.
Statistical models, such as machine learning, are flexible and powerful, particularly in recognising patterns across large datasets. But they’re often seen as “black boxes”, difficult to explain and harder to trust. Deterministic models, by contrast, are grounded in physical laws and deliver transparent results, but they lack predictive agility.
Hybrid AI combines the best of both worlds: predictive power with reliability. Operators can act confidently on powerful recommendations that are grounded in physical reality.
Working with Real-World Data
Real-world pipeline data is rarely perfect. Gaps in SCADA datasets are common, and critical variables may be missing. Sometimes, missing information, such as fluid density, can be inferred through analytical models. Other times, more complex hydraulic or statistical modelling is required.
But greater model complexity introduces potential for greater uncertainty. That’s why Klarian incorporates error propagation throughout the data pipeline. Every result is delivered with confidence intervals, ensuring transparency and reliability.
Sensor reliability is also a key consideration. All sensors carry some level of error and risk of malfunction. Building robust AI solutions requires acknowledging this, and designing with fault tolerance in mind.
Interoperability, Security and Adoption
Another real-world challenge is data integration. Pulling together datasets from across a network, whether via API or simple exports, is essential. Interoperability underpins hybrid AI, alongside robust data security and system integrity.
But successful AI adoption isn’t just a technical challenge, it’s operational. Solutions must be explainable and intuitive for control room operators, or they risk being sidelined. Many AI tools fail not due to poor performance, but because they weren’t embraced by frontline users.
Hybrid AI eases this transition. By combining familiar, trusted models with new predictive capabilities, it creates a smoother path to adoption and delivers results faster.
Building the Future of Pipeline Operations
Introducing AI into pipeline operations brings technical and operational complexity, but the payoff is significant: increased uptime, lower energy costs, and smarter maintenance strategies.
To get there, collaboration is key.
Klarian’s approach ensures that AI tools are developed with operators, not just for them. The result is technology that fits seamlessly into existing workflows, and earns the trust needed to deliver lasting value.
Based on the paper by Dr Liam Trimby, Head of Research and Development, and Clare Acreman, Applied Scientist, at Klarian.