The 60-second grid frequency dip: clocked battery self-tests, not market trading

Grid operators expect frequency noise, and they have tools for it. A signal that repeats once a minute, with no drift in its timing, is a different kind of problem. TenneT Germany noticed such a signal in the European grid over several years. According to Forschungszentrum Jülich, a team led by Dirk Witthaut, with colleagues at the University of Cologne and the Karlsruhe Institute of Technology, has now traced it to battery systems that check their state of charge every minute. The work appeared on 30 October 2026 in Nature Energy. It describes an engineering failure that is easy to miss: many independent devices run the same software routine and are anchored to the same clock.

01A clock-shaped signature in the frequency

Frequency is a direct readout of power balance. The rate of change of frequency follows the instantaneous power imbalance in the grid. In the paper's reading of the data, a drop in frequency during the first half of each minute means demand is exceeding generation in that interval.

The pattern is not specific to Europe. The authors examined frequency records from 12 electricity systems on five continents. They found a narrow spectral peak at a period of exactly one minute, and its timing stays coherent over long periods.

Size matters here, because the effect is small. The paper estimates that a 3 mHz periodic frequency variation corresponds to a coherent power variation of about 80 MW. In Continental Europe the nominal frequency is 50 Hz, so this is a tiny deviation. TenneT's observation was that it is measurable and unusually regular, even though the deviations were not critical to grid security.

Witthaut's reasoning was that the exactness of the period ruled out most physical explanations. "It quickly became clear to us that the cause must be something controlled by the clock." The amplitude also ruled out any single device. The effect needs many devices drawing power at the same moment.

02Finding the source on a campus

The decisive evidence came from the Living Lab Energy Campus at Jülich. The team put measuring points on end devices at 80 locations. The same periodic consumption pattern appeared exactly where large battery systems provide uninterrupted power supply.

Secondary summaries of the paper add detail on the chain of evidence:

• At one installation the team recorded recurring current peaks at the battery terminals. The manufacturer confirmed that periodic battery energy-management self-tests caused them.
• Spectral analysis followed the resulting power variation through transformers and a busbar to the interface with the public 110-kV network.

The paper's own wording is broader than the headline "battery storage". It names clock-synchronized control software in digitally controlled energy devices, particularly uninterruptible power supplies (UPS) and battery energy storage systems, as the source of periodic power-demand peaks that propagate across voltage levels. The Jülich press material is looser and speaks of large battery storage systems. The mechanism is the same in both. A routine check draws a short burst of power, and it is scheduled at the top of the minute.

03Why the self-tests line up

No one coordinated these devices. Witthaut's explanation is mundane. There are no binding standards for such power checks, and he suspects most manufacturers assumed a 60-second interval was sensible. The start of each minute was simply chosen as the test time. Every vendor made a locally reasonable choice. Taken together, the choices produce a grid-scale signal.

The paper tests the role of time synchronisation directly. In one UPS, the self-test timing was compared with and without Network Time Protocol (NTP) synchronisation:

• With NTP disabled, the self-test signatures drifted slowly relative to UTC.
• With NTP enabled, the daily progression of the signatures became much more pronounced, at about 6 s per day.

A summary of the paper says the phase stays stable during the day and shifts at midnight. I could not check that against the full text. In the paper's framing, external clock synchronisation is what keeps independent devices in step. The authors describe a broader class of cyber–physical synchronization effects in digitally controlled devices that run on externally synchronized clocks or periodic software routines.

In a commentary in the same journal, Florian Dörfler describes a disturbance recurring precisely every minute in grids worldwide, caused by clock-synchronized UPS self-tests that inadvertently act in unison.

The paper does not claim that UPS and battery systems are the whole story. Witthaut said it is entirely possible that battery storage systems are only part of the problem. He suggested that heat pumps, air conditioners and other appliances with digital control could show similar behaviour. That remains a hypothesis, not a measured result.

04What it means for frequency control and scheduling

The amplitude of the effect depends on both the source and the grid it enters. The authors note that lower inertia can increase the frequency response to a given imbalance, while hydropower turbine-governor dynamics can amplify an externally driven signal. Two findings from the paper and its coverage show this:

• Ireland. Data centres with large battery systems account for around a fifth of the country's electricity demand, and the amplitude of the one-minute signal quadrupled between 2016 and 2021. Secondary coverage of the paper puts the dip at roughly 3 mHz growing to 12 mHz.
• Hydropower. An explainable machine-learning analysis found systematic correlations between hydropower commitment and signal amplitude in the Continental European grid. The mechanical cost is that brief demand spikes make turbines slow down momentarily and then speed back up, which adds wear.

Hydro-heavy systems already knew a related problem. Nordic work from the 2010s describes a "60 second oscillation" that induces wear and tear in hydropower units. A 2015 Nordic frequency-quality report said the causes of that oscillation were not yet fully clear, though it noted that the frequency-containment response amplifies it. The new paper does not claim to explain every Nordic oscillation. It does give a plausible external driver for a component that control tuning alone could not account for. A summary of the paper reports Scandinavian amplitudes above 10 mHz, and no amplitude above 10 mHz in 2025 after frequency-control rules were revised in 2023. I could not verify that against the paper's full text, and the paper may not attribute the improvement to the rule change.

For control design, the distinction is between stochastic and deterministic disturbances. Reserves are dimensioned for random imbalance, such as forecast errors or plant trips. Frequency also reflects market behaviour, including the abrupt switching of power plants, which the Jülich team lists among the many influences on the pattern. A periodic one-minute component sits on top of all that. It is predictable in phase and period. The paper says system-level measures can reduce the frequency impact of deterministic disturbances. It also says the root cause can be addressed directly through control design.

In practice that means two levers, and I am describing them in my own terms, not quoting the paper. Operators can account for the known signal in how reserve units and hydro governors respond. Device makers can stop creating the signal.

05The fix, and what remains unproven

The fix on the device side is cheap. Witthaut: "You would simply have to distribute the checks randomly, rather than always starting them at the same time." A random offset in the scheduler spreads the load over the minute, so the sum of many devices approaches a constant. Nothing about the battery hardware has to change. Because no standard governs the checks, the likely route is a requirement or guideline on test scheduling, applied to new devices and delivered to existing ones by firmware update. Neither the paper nor the press material says such a change is under way.

Several points remain open:

• Attribution. The causal chain is shown in detail for specific installations at Jülich. Its share of the worldwide signal has not been quantified. The authors themselves leave room for other devices.
• Growth. The paper warns that rapid expansion of battery storage, data centres and the Internet of Things may further amplify such oscillations. The Irish record is the main evidence so far for amplitude growth.
• Severity. The authors report that the disturbances have not reached safety-critical levels in the grids studied. The costs are mainly wear on hydro equipment and extra load on frequency regulation.

The wider lesson is about software. As grids fill with power electronics and networked controllers, a default such as "run at :00 every minute" becomes part of the grid's dynamics. A scheduling choice made for convenience in firmware can show up in a continental frequency record years later.