New AI Project Tackles Costly Measurement Errors in the Green Gas Transition
Delft, Saturday, 22 August 2026.
TNO is deploying machine learning to correct a 35% underestimation in current gas metering, securing a reliable transition to hydrogen and biomethane.
Bridging the Metering Gap in Renewable Gases
As the European Commission targets a fully carbon-neutral energy system by the year 2050, the European energy grid is undergoing a rapid evolution [1]. Transitioning from traditional natural gas to renewable alternatives, such as biomethane and hydrogen, introduces significant physical challenges [1]. These green gases have highly variable compositions, fluctuating supply and demand patterns, and unpredictable flow rates [1]. Compounding this issue, current fiscal metering practices are estimated to underestimate measurement uncertainty by approximately 35% [1]. This discrepancy poses a serious financial and operational risk for grid operators who require absolute precision in billing and distribution [GPT].
Deploying AI to the Front Lines of Energy Transition
To address this critical measurement gap, the Netherlands Organisation for Applied Scientific Research (TNO) initiated a new machine learning project on August 21, 2026, in Delft to optimize smart gas flow metering [2]. Working in tandem with these operational plans, TNO recruiter Ryvo Octaviano announced on August 17, 2026, an MSc internship and thesis opportunity based at TNO’s Rijswijk location [3]. This research project aims to leverage advanced data-driven models to predict and analyze complex gas flow dynamics in real-time, establishing the vital data infrastructure needed for Dutch grid modernization [GPT].
The Technical Mechanics of Smart Flow Modeling
The project will combine experimental laboratory data with High-Fidelity Computational Fluid Dynamics (CFD) simulations [1][3]. By utilizing these datasets, researchers plan to train and validate diverse machine learning architectures, including classical machine learning, Physics-Informed Neural Networks (PINNs), hybrid models, and probabilistic machine learning [1]. This hybrid approach ensures that the resulting AI models do not just recognize statistical patterns but also respect the fundamental laws of fluid dynamics [2]. Crucially, the engineered models must comply with strict international metrological standards, including ISO/IEC 17025, OIML R140, ISO 15112, and the ISO/IEC Guide 98 [1].
Project Parameters and Future Scaling
The recruitment process for this initiative is currently underway, with the application deadline set for September 9, 2026 [1]. The selected Master of Science student will work between 32 and 40 hours per week [3]. For a student working the maximum weekly commitment, this equates to 160 hours of research over a standard four-week period. TNO provides a monthly allowance of €615 for full-time work, alongside up to 8 hours of leave per internship month [1]. Looking beyond the initial research phase, TNO plans to scale the pilot program to additional Dutch municipalities by January 1, 2027 [2].