Leveraging AI and Machine Learning to build confidence in electrolyser performance
Endua engineers analysing EIS data
The impact of AI-interpreted EIS data on the future of hydrogen electrolysis
As the hydrogen economy scales rapidly, electrolyser manufacturers face a critical challenge: how do we provide customers with reliable long-term cost projections when our most advanced PEM technologies are constantly evolving? Unlike mature industrial equipment with decades of field data, next-generation electrolysers incorporate novel materials, higher current densities, and optimised membrane assemblies—all of which have the potential to deliver transformative cost and performance benefits to customers.
However, this innovation creates uncertainty for project developers who need accurate OPEX and CAPEX forecasts to secure financing and justify investments. The gap between demonstrated initial performance and twenty-year operational reality represents one of the industry's most significant barriers to adopting world-leading technology. Customers need more than efficiency specifications or nameplate lifetime projections—they need confidence that their systems will perform predictably, degrade gracefully, and deliver the total cost of ownership that business cases depend upon.
EIS as the foundation for stack diagnostics
Electrochemical impedance spectroscopy (EIS) has emerged as a versatile and powerful diagnostic tool for electrolyser stack performance, capable of probing ohmic, charge-transfer, and mass-transport processes in real time. By tracking changes in equivalent circuit parameters and distribution of relaxation times (DRT), we can detect specific failure modes within the stack before they impact hydrogen output or efficiency. EIS provides a robust foundation for understanding stack health under the dynamic operating conditions that characterise renewable-powered hydrogen production—where load fluctuations and intermittency create unique degradation stresses.
Transforming data into predictive intelligence
The implementation of artificial intelligence in EIS analysis shows immense potential to fundamentally transform how impedance data becomes actionable diagnostics for stack operators. Rather than relying solely on manual interpretation of Nyquist or Bode plots—a time-intensive process requiring specialist expertise—AI enables a systematic approach that ensures data quality, accelerates analysis, and links electrochemical signatures directly to physical degradation mechanisms within the stack. This creates a predictive framework for equipment degradation that translates directly into the cost certainty customers require.
Machine learning models trained on accelerated testing data and early field deployments can identify degradation patterns that would take years to recognise through conventional analysis. The integration of AI strengthens diagnostic capability by automating data processing, detecting measurement artifacts, and enabling predictive analytics such as remaining useful life (RUL) estimation. This addresses the predictability challenge at its core by allowing us to:
Compress learning cycles and extract maximum insight from limited operational data
Translate early field experience into reliable long-term performance forecasts
Provide customers with quantified confidence intervals around lifetime costs
Ultimately, AI-enhanced diagnostics lower total cost of ownership, reduce unplanned downtime, and position advanced electrolyser systems as bankable, scalable solutions for green hydrogen production. By bridging the gap between cutting-edge technology and operational certainty, we can accelerate the deployment of the highest-performing systems without sacrificing the financial predictability that projects demand.
