Power Grids: Software Synchronizes Frequency Fluctuations

Every minute, a subtle fluctuation recurs in the frequency of many electricity grids. The signal can be hard to see in ordinary measurements, yet remains identifiable over years. Why do such different infrastructures exhibit an exact 60-second recurrence?

A study published in Nature Energy on October 30, 2026, provides an explanation. Researchers from Forschungszentrum Jülich, the University of Cologne and Karlsruhe Institute of Technology showed that synchronized software operations can create periodic power-demand peaks. They identified this mechanism in battery-backed uninterruptible power supplies, or UPS systems.

The finding connects two disciplines often examined separately: the timing of embedded software and the physical stability of electricity networks. It does not, however, prove that UPS systems alone explain every observed oscillation worldwide.

A 60-second pattern across five continents

Grid frequency reflects the balance between electricity generation and demand. It tends to fall when demand exceeds supply and rise in the opposite situation. Control systems maintain it around 50 or 60 hertz, depending on the region.

The team examined 12 grids across five continents, including Continental Europe, Great Britain, Scandinavia, Australia, the United States, Malaysia, South Korea, South Africa and several island systems. They range from Iceland's grid, serving approximately 400,000 customers, to the Continental European system, serving more than 400 million.

In every case, Fourier analysis revealed a narrow spectral peak at a period of exactly 60 seconds. This method separates frequency measurements by recurring components, allowing a regular signal to emerge even when obscured by larger variations. A clearly visible frequency drop at every minute boundary is not required.

The timing precision suggests clock-controlled actions rather than a purely mechanical phenomenon. It also distinguishes the signal from electricity-trading effects, which recur at 15- or 60-minute intervals, and much faster electromechanical oscillations.

The same period does not imply that separate continents oscillate in perfect synchrony: their peaks need not occur at the same UTC second, and every grid has its own phase and amplitude. The researchers also identified a weaker component near 100 seconds whose cause remains uncertain.

At Jülich, the evidence leads to batteries

To find a concrete mechanism, the team led by Dirk Witthaut, including Carsten Hartmann, investigated the Living Lab Energy Campus in Jülich. The research center reports measurements at 80 locations. Four buildings had a pronounced periodic electricity-demand signal; all contained battery-backed UPS systems.

These systems maintain power during supply interruptions, particularly for data centers, healthcare facilities and other critical installations. Their management software periodically checks battery conditions.

At one instrumented installation, the researchers detected battery current spikes repeating every minute. The manufacturer confirmed they were periodic battery-management self-tests. Their electrical signature could be traced through transformers to the public 110-kilovolt grid connection.

The peaks were clearly visible on the direct-current side but much less apparent on the alternating-current side, where they overlapped with other loads. Fourier analysis recovered their periodic component in power measurements up to the high-voltage interface.

This demonstrates that a local software routine can produce an electrical variation propagating across voltage levels. It does not quantify how much UPS systems, or other devices, contribute to the recurring pattern in each of the twelve grids.

Computer clocks and NTP

Why do the self-tests coincide? Digital equipment commonly relies on a shared time reference. Network Time Protocol (NTP) synchronizes computer clocks, but does not stagger the electrical tasks scheduled by applications.

The researchers compared UPS behavior with NTP enabled and disabled. Without synchronization, the pulses gradually drifted relative to universal time. With NTP enabled, their phase was more stable throughout the day, followed by a discrete correction during a nightly adjustment at midnight UTC. In the investigated installation, the corrected timing drift could reach approximately six seconds per day.

This directly links software scheduling to the electrical signature. NTP is not defective: a problem arises when many programs use the same clock to initiate electricity-consuming actions simultaneously. Other buildings showed different fixed phases or drifting patterns, depending on their equipment.

Three millihertz, about 80 megawatts

Against a nominal grid frequency of 50 hertz, the observed variations are small. They are measured in millihertz, or thousandths of a hertz. For Continental Europe, the study uses an estimated overall power-frequency characteristic of 26.68 gigawatts per hertz.

On that basis, an oscillation with a 3-millihertz amplitude corresponds to approximately 80 megawatts of coherent power variation. This is an estimate of the grid-wide aggregate imbalance, not the demand of one UPS or power directly measured at Jülich.

The effect arises from many small demands adding together. Signal amplitude generally follows daily activity, with weaker patterns at night and stronger ones during the day.

Ireland shows how the amplitude can grow

In Ireland, the frequency drop associated with the pattern increased from approximately 3 to 12 millihertz between 2016 and 2021, according to the study. More recent measurements show individual events exceeding 20 millihertz. Combined with other deviations, these can temporarily bring the grid closer to the limits of its normal operating range.

Data centers accounted for about 18% of national electricity demand in 2022, according to figures cited by the authors. Their growth is a plausible contributing factor, but not proof of exclusive causation: grid characteristics, frequency-control arrangements and inertia have also changed.

Large synchronous machines store kinetic energy that helps moderate frequency changes. A reduction in effective inertia may make a grid more sensitive to the same power imbalance, depending on installed electronics and controls.

The datasets are not identical. Some European frequency series, sampled every second, span 2015–2025; other regions have shorter observations; campus power measurements at Jülich have three-second resolution. These methodological differences limit direct amplitude comparisons between grids.

A burden on regulation, not an impending crisis

The authors state that the disturbances have not reached a safety-critical level in the grids studied. The findings do not demonstrate an imminent threat of blackouts.

Repeated fluctuations can nevertheless tax frequency-control equipment, including hydropower turbines and governors. The paper discusses potential wear, but does not quantify a general increase in maintenance costs or reduced equipment lifetime caused solely by the 60-second signal.

In Scandinavia, amplitudes above 10 millihertz occurred regularly before certain frequency-response requirements were strengthened in 2023. These included tests for oscillations with periods of 10 to 300 seconds. The authors found no comparable exceedances in their 2025 data. In Australia, changes in frequency-response requirements were also associated with reduced signal amplitude.

Such measures change the grid's response; they do not show that the underlying synchronized pulses have disappeared. Growing deployment of batteries, data centers and software-controlled equipment could increase these interactions, without implying that storage technology itself is faulty or that a cyberattack occurred.

Staggering self-tests instead of removing them

The most direct remedy is to spread electricity-consuming checks over time: device-specific offsets, randomized schedules or event-driven rather than strictly clock-driven routines. Similar techniques are used in computing to prevent scheduled tasks from creating simultaneous demand peaks.

The researchers have already obtained a limited practical result. Following discussions with a manufacturer, a software update removed the strict periodicity of current peaks in one battery block. However, some self-tests still occurred at discrete moments, so the mitigation was only partial.

Wider adoption would require testing across equipment brands and generations while preserving diagnostics essential for reliability. Manufacturers, grid operators and users of critical infrastructure would need to coordinate. Frequency-control systems provide a complementary defense by limiting the grid's response rather than eliminating synchronization at its source.

What the sources actually establish

The original Nature Energy article supplies the measurements, UPS experiments and limits of the worldwide interpretation. Germany's Forschungszentrum Jülich highlights potential applications in its own announcement. Its role in the research is relevant institutional context, not evidence by itself of a conflict of interest.

In a separate scientific commentary, ETH Zurich researcher Florian Dörfler examines unintended synchronization in digital energy infrastructure but does not report independent experimental replication. Tech Xplore republishes the Jülich account and therefore offers no independent confirmation.

Searches conducted on the announcement date did not establish three to five independent treatments from different world regions. The study's global scope rests on data collected across five continents, not on an invented diversity of media reporting.

Some research data are available through Zenodo, although certain Gridradar and EirGrid datasets have access restrictions. The Swiss commentary declares no competing interests, and no specific commercial conflict undermining the primary study was established.

When software scheduling becomes a grid-stability issue

UPS systems and their self-tests serve a real need for reliable power. The aim should not be to eliminate these checks, but to prevent independent devices from carrying them out simultaneously.

The next step is to test the long-term effectiveness of software changes across more devices and grids. The study exposes a design constraint of modern infrastructure: stability also depends on when software acts. A grid can be disturbed not because its components malfunction, but because they correctly execute the same task at precisely the same moment.

FAQ

Why does a pattern recur every 60 seconds?

Battery-management self-tests can create synchronized power demand. The mechanism was demonstrated at Jülich, but its precise contribution in each worldwide grid remains unknown.

Do the oscillations threaten electricity supply?

No general safety-critical threat has been demonstrated. However, some frequency drops in Ireland exceed 20 millihertz, and repeated deviations can burden frequency-control equipment.

Can a software update eliminate the problem?

Staggered or randomized self-tests may reduce synchronization. An initial update removed strict periodicity in one battery block but did not eliminate all checks at discrete times.