Demand quality data
Renewable energy is one of the fastest-moving engineering sectors today, and right now its attention is fixed firmly on software
Global investment in energy transition technologies exceeded US$2trn in 2024 for the first time, with a growing share directed at predictive maintenance platforms, cloud-based monitoring systems, and digital twin modelling of wind farms, solar storage arrays and hydrogen facilities. It is an exciting development. It is also, potentially, a dangerous distraction.
Unanswered questions
The question that remains unanswered is this: how good is the physical data feeding all of those intelligent systems? Advocates of software-led intelligence argue that machine learning can compensate for imperfect sensor data and that drift can be corrected algorithmically. Engineers who have learned otherwise – often expensively – know that no analytics platform can recover a fundamentally corrupted data stream. Studies suggest poor data quality costs industrial operators between 15-25% of their revenue in operational inefficiencies. In capital-intensive renewable energy, that is not an acceptable margin for error.
Nowhere is this more consequential than in hydrogen. PEM electrolysers and fuel cells operate within thermal windows as narrow as ten degrees Celsius before efficiency degrades or safety margins are breached. Temperature measurement points inside a fuel cell stack are not data points for a performance dashboard – they are real-time inputs to safety-critical decisions. A sensor that drifts or fails silently is not a software problem to be corrected upstream. It is a direct operational and safety risk.
Wind energy
Wind energy tells a similar story with a sharper commercial edge. Offshore turbines are engineered to run for 20 to 25 years in locations where access is difficult and costly. A single unplanned maintenance visit in the North Sea can exceed £30,000 in vessel and logistics costs alone. Many of those call-outs are driven not by genuine mechanical failure, but by drifting sensors issuing false signals to predictive maintenance systems that cannot distinguish between a real fault and a measurement error. The software is doing its job. The instrumentation is not.
Battery energy storage, deployed at scale alongside solar PV to meet grid balancing obligations, raises the stakes further. Global installed capacity is forecast to exceed 1,500 gigawatt-hours by 2030. In large-format lithium-ion arrays, temperature uniformity is safety-critical, and as energy densities increase, acceptable measurement tolerance shrinks. A sensor reading just two or three degrees outside its calibrated range is not a rounding error – in a system storing megawatt-hours of energy, it is a potential pathway to thermal runaway.
Hydrogen, wind and solar storage share one fundamental vulnerability: they all depend on physical measurement as the foundation of safe, efficient operation. Sensors and transmitters embedded in these systems are not legacy components waiting to be superseded by smarter software. They are the point where the physical world meets the digital one. If that point is unreliable, everything built above it is compromised.
For more information visit: www.jumo.group/renewable-energies
Contributed by Maria Silenti from Jumo
