Autonomous Hydropower Dam Inspection

Summary

The inspection site

Storfinnforsen Hydroelectric Power Station is a significant hydropower installation in Norrland in Sweden. Commissioned in 1954, it has an installed capacity of 112 MW and an annual generation of approximately 536 GWh. The plant plays a critical role in supplying renewable electricity to the region, contributing to Sweden’s ambitious sustainability goals. It is operated by Uniper, a leading international energy company focused on reliable and sustainable energy solutions.

Background: Maintaining structural integrity

With a 10-year inspection cycle, the dam is subject to detailed assessments of its massive concrete structures. Seasonal temperature shifts and fluctuating pressure from the water and bedrock cause the dam sections to tilt slightly back and forth over time. While this movement is to be expected, it creates stress patterns that can lead to surface cracks, posing a long-term risk to dam integrity.

To counter some of the environmental variables, the dam sections have been encased in protective cladding. This weatherproofing improves control over internal humidity and temperature, creating a more stable inspection environment.

Previously, such inspections required extensive scaffolding to map cracks manually across multiple sections of the dam, a time-consuming and high-risk method that offered limited comparability between inspection cycles. The introduction of drone-based inspection changes all this.

No battery swaps, no pauses: Endurance by design

The Scout 137 is designed for endurance at every level. ScoutDI’s tether system provides both continuous power and stable wired data transfer, but the value goes beyond this. From flight modes to payload integration and operator support, the system is built for continuous operations lasting many hours each day. 

Endurance by design ties directly into automated flight patterns, which only deliver full value when they can run without interruption. By combining unlimited flight time with automation, the Scout 137 enables truly large-scale inspections, covering entire structures in a consistent and repeatable way without unnecessary pauses for battery swaps or maintenance.

Introducing Autonomy

Halfway through the inspection, the workflow shifted from full manual flying to a beta release of ScoutDI’s new autonomy features, including Auto Surface Scan, Auto Image Capture and Distance Lock. The pilot simply defined the scan area by marking the top and bottom points of the wall and selecting the sidestep width, and the drone then carried out the systematic coverage on its own.

By maintaining consistent speed, consistent motion patterns, and consistent framing through Distance Lock, the Scout 137 generated a highly regular dataset. This not only improved data quality but also reduced the strain of continuous manual flying, allowing the pilot to focus on coverage and monitoring.

The pilot controls the Scout 137 Drone System via a controller and this tablet. The 3D point cloud, flight path and video feed are simultaneously visible on the screen in real-time .

Handling data at scale

With automation in place, the inspection workflow became far more structured and consistent. KIWA completed 52 flight hours inside the dam, capturing over 300,000 high-resolution images using the Scout 137 and its new autonomous flight functions. 

But raw imagery is only potential value. The second half of the workflow turns it into insight: the full dataset was handed to Spotscale, a Swedish company that specializes specifically in 3D modelling and analysis for infrastructure, to be reconstructed and analysed. The results are shown below.

From images to a navigable 3D model

Spotscale reconstructed the 300,000 images into a complete photogrammetry model of the dam. Because capture was consistent, the direct result of Auto Surface Scan and Distance Lock, the model is dense and geometrically reliable end to end.

This is where “photogrammetry-ready” becomes measurable: regular input produces a clean model, and a clean model produces trustworthy measurements. Each compartment is reconstructed individually; shown in this video is compartment 4142 (“fack 4142”), navigable in full 3D in the Spotscale portal.

Every crack, detected and measurable

Spotscale’s analysis detects surface cracks and draws them directly onto the model, in place on the surface they belong to. That turns the model into a working tool rather than a picture: every crack is measurable in length and position, cracks can be tagged and filtered for triage, and the full set exports to CSV straight into the asset owner’s reporting.

The result is a single, authoritative record of surface condition; one an engineer can interrogate from their desk, without scaffolding or a return trip to site.

Navigating the model

The model can be rotated, zoomed and inspected from any angle. What once required three sets of scaffolding and months of work at height is now a few seconds on screen; and nothing is missed or forgotten between visits.

Built for comparison over time

Storfinnforsen runs on a 10-year inspection cycle, and its concrete sections shift slightly with temperature and pressure, the very movement that produces stress cracks. A single inspection shows the condition today; the value is in tracking change.

Because the workflow yields the same structured, location-tagged dataset every time, each future inspection aligns against this baseline: which cracks are new, which have grown, which are stable. Autonomy makes this practical; the same coverage, flown the same way, makes models directly comparable rather than approximately similar. That is the difference between inspecting a dam and monitoring it.

Why comparability makes the inspection worth doing

This is not only a gain in convenience. With manual, scaffolding-based crack mapping, one inspection can be compared with the next only very roughly, which makes genuine trend analysis hard to justify.

According to asset owner Uniper, the manual approach is so difficult to compare over time that a repeat mapping might not have been worth undertaking at all. A structured, repeatable dataset changes that calculation: follow-up mappings line up directly against earlier rounds, which is what makes ongoing crack monitoring worth doing in the first place.

Results and impact

The automated flight and image-capture functions proved highly effective. The drone performed smooth, accurate zig-zag flights and maintained a steady distance to the surface, producing a structured and consistent dataset of 300,000 location-tagged, high-resolution images.

Based on the initial inspection of the two largest compartments, KIWA recorded 52 flight hours in total. After introducing the new autonomy functions, the same working time resulted in approximately 50% greater coverage compared to fully manual operation. This corresponds to just over three working days saved on the two first compartments. For the remaining 78 smaller compartments, the same efficiency gain represents an estimated saving of around 40 working days in total.

According to asset owner Uniper, previous scaffolding-based inspections required risky work at height over several months, including both the scaffolding and the inspection itself. Drone-based inspection has replaced this process, fundamentally improving safety, cost efficiency and data quality. The new autonomy functions further enhance all these aspects.

Conclusion: A complete inspection ecosystem

This is an ecosystem, not a single product.

ScoutDI’s photogrammetry-ready Scout 137 Gen 3 captures structured, location-tagged imagery at a scale and consistency manual flying cannot match; Spotscale turns it into a navigable model with automatically detected, measurable and exportable cracks. Asset owner Uniper gets a safer, faster inspection plus a durable digital record that improves every time the dam is re-flown.

As dams, bridges and tunnels face rising demands for safety and lifetime extension, it is this pairing of reliable autonomous capture with advanced analysis that makes data-driven maintenance of critical public infrastructure realistic at scale.