Aging hydropower plants across three continents are getting a digital overhaul — and the sensors doing it can detect algae, cracks, and pathogens in real time
Image generated with artificial intelligenceHydropower plants are among the oldest workhorses in the global energy mix — and in many cases, they’re running largely blind. Europe’s fleet averages 42 to 46 years old; in the United States, that figure climbs to around 64. At facilities like the Ilarionas plant in northern Greece, turbines spin and water flows through penstocks and drainage systems that have never had a sensor watching for early signs of stress.
That gap matters more now than it once did. As demand for reliable, carbon-free electricity grows, aging infrastructure must perform without the margin for undetected failures — whether a crack forming in a pump casing or a toxic algae bloom quietly spreading across a reservoir.
A renewable workhorse running on borrowed time
Hydropower supplies roughly a sixth of the world’s electricity, making it the largest source of renewable power on the planet. But the infrastructure delivering that power is showing its age. Europe’s fleet averages 42 to 46 years old; the United States sits closer to 64. China’s plants are younger — around 20 years on average — yet the global picture remains one of an industry built for a different era, now asked to carry modern grid loads.
The team deployed a portable multiparametric platform — described as a “suitcase” lab — that can be taken to any sampling point across a reservoir.
Aging doesn’t just mean wear. It means monitoring gaps. Many plants were built before digital instrumentation existed, and retrofitting sensors into concrete-and-steel structures designed decades ago is neither simple nor cheap. As grids lean harder on hydropower to balance intermittent solar and wind, the tolerance for undetected failures shrinks considerably. That’s the problem the EU-funded Di-Hydro project — supported under the Horizon Europe Research and Innovation Programme — was designed to solve.
Listening for cracks: structural health sensors go to work
The first tool Di-Hydro brings to aging infrastructure is a structural health monitoring (SHM) sensor node developed by CERTH. Low-cost and low-power, it’s designed to be retrofitted without invasive installation — a critical feature when working with decades-old facilities that can’t easily be taken offline.
The node pairs two systems. An acoustic emission (AE) sensor — built on the Qawrums RAEM-2 architecture — detects the elastic waves that materials emit when they crack, deform, or change under load; faulty bearings and defective gears produce similar signals. Alongside it, a multisensor unit combines a triaxial accelerometer, gyroscope, magnetometer, barometer, and temperature and humidity sensors, all connected to a Raspberry Pi microcomputer via a Sense HAT board. Data from the AE system upload automatically to a cloud server, tracking amplitude, RMS, power, and ASL values over time — building a continuous history that makes anomalies visible before they become failures.
The first real-world deployment happened at the Ilarionas HPP in northern Greece. Sensors were placed on drainage pumps and a penstock, two areas that had never been instrumented before. That’s not unusual. Across aging fleets, critical components often go unwatched simply because no one built a practical way to watch them.
Green soup and silent nitrogen: why water quality monitoring matters
Structural integrity is only half the challenge. Behind every dam sits a reservoir, and reservoirs are ecosystems that can turn hostile with little warning.
The core problem is stagnation. When water stops moving, it stratifies — forming layers with different temperatures and oxygen levels. That layering limits oxygen mixing, creating conditions for microbiological and chemical imbalances. In low-oxygen zones, nitrification can intensify, altering nutrient cycles and allowing harmful nitrogen compounds to accumulate. For reservoirs supplying drinking water to surrounding communities, that’s a direct public health concern.
Then there are algal blooms — what researchers sometimes call “green soup.” When excess nutrients like nitrogen and phosphorus combine with warm, still water, algae can multiply rapidly. Some species produce toxins; as blooms decompose, they strip oxygen from the water and harm aquatic life. They can also physically clog HPP piping systems, reducing or halting power generation entirely. Many reservoirs serve simultaneously as drinking water sources, irrigation supplies, and recreational areas, so degraded water quality raises treatment costs and creates health risks for communities that may have no alternative supply.
A sensor suite that sees what human inspectors cannot
Di-Hydro’s water quality response combines several technologies into a coordinated monitoring platform. INOSENS developed a new electrochemical sensor for ammonia detection, fluorescence-based algae sensors, and an E. coli biosensor — all integrated with instruments measuring temperature, turbidity, pH, conductivity, and dissolved oxygen, giving operators a multi-parameter view of reservoir conditions in real time.
AIMEN’s contribution goes further. The team deployed a portable multiparametric platform — described as a “suitcase” lab — that can be taken to any sampling point across a reservoir. Inside, a tryptophan-like fluorescence sensor estimates pathogenic contamination within seconds, using E. coli’s natural fluorescence as a proxy indicator. The suitcase also carries a portable Digital Holographic Microscope (DHM) that uses laser interferometry combined with microscopy objectives and a digital camera to produce three-dimensional images of microorganisms in a water sample. Image processing then identifies and counts key species — cyanobacteria, green algae, and others — enabling real-time biodiversity assessment that would otherwise require days of laboratory work.
Together, these tools support early-warning decisions: controlled water releases, targeted aeration, or operational adjustments before a bloom or contamination event escalates.
Digital twins and AI: connecting the dots across the whole plant
Individual sensors generate data. What Di-Hydro’s decision-making framework does is turn that data into coordinated action across an entire facility.
Two interconnected AI and machine learning models sit at the center of the system. The first assesses and predicts biological activity in the reservoir, drawing on sensor readings and automated analysis of DHM images. The second integrates those outputs with HPP operational history from the plant’s SCADA system, linking environmental changes directly to equipment performance. One of the technical challenges was reconciling datasets that operate on vastly different timescales — historical water quality reports might offer only a few measurements per year, while sensor streams can produce readings every minute. The unified model bridges that gap.
Digital twins layer on top of this, facilitating data exchange across the plant and supporting optimal coordination of power generation. Inputs include societal demand, weather forecasts, water flow data, and biodiversity indicators — a breadth of information that no human operator could synthesize manually at speed.
What comes next for the aging fleet
Di-Hydro is a research project, and its deployments so far are pilots. The questions it’s answering, though — how do you instrument a 60-year-old penstock? how do you detect a cyanobacteria bloom before it clogs your intake? — are questions that thousands of plant operators around the world are facing right now.
The combination of low-invasiveness sensors, portable field labs, and AI-driven decision tools points toward a path that doesn’t require tearing down and rebuilding aging infrastructure. It requires layering intelligence onto what already exists. As the project moves toward broader validation and eventual commercialization, the technologies developed at Ilarionas and other pilot sites could become the template for how an aging global fleet earns another generation of reliable service.
Carlos is an engineer with strong expertise in technical and industrial topics. He previously worked at international companies such as Siemens and is multilingual.
