The Few Seconds a Generator's Spinning Mass Gave Away for Free — What Gets Harder as Grid Inertia Falls

The Blackout Lasted Ten Hours, and the Rise in Deaths in Spain Came the Two Days After

On April 28, 2025, electricity went out for about ten hours across mainland Spain and Portugal. An area where 47 million people live went dark all at once (DOI: 10.2139/ssrn.6551460, preprint), and 8 deaths were reported (DOI: 10.1038/s41467-026-75581-w). The numbers that went out in the news that day stopped there.

An observational study that counted deaths separately after the blackout came out afterward. Excess deaths are deaths that exceed the number expected under normal circumstances. This blackout was associated with an increase in deaths in Spain, but not in Portugal. On the day of the blackout itself, there was little evidence of excess deaths. The increase in deaths in Spain occurred over the next two days, and the increased number was estimated at 167. The 95% credible interval—the range likely to contain the true value based on the data—is +28 to +300 people. Relatively speaking, this was 2.4% more than usual, and it was particularly prominent among women aged 85 and older (DOI: 10.1038/s41467-026-75581-w, observational study).

Differences were also visible depending on the region. The authors write that it is uncertain whether what made that difference was the severity of the blackout, original vulnerability, or the conditions of local medical and infrastructure systems. The authors believe these results show that the health burden of blackouts has been underestimated (DOI: 10.1038/s41467-026-75581-w, observational study). Whether a grid holds up or collapses ultimately leads to numbers like these.

Starting the day after the event, a sentence followed: "Wasn't it caused by having too much renewable energy?" This piece examines that sentence. However, this piece does not judge the cause of that day. The conclusion of the official investigation lies in the final report introduced in the primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together", and the academic papers read here ask what that event revealed. Along with each citation, we note whether it is measured data, an estimate, or a simulation. Most studies covering this event are still preprints, meaning manuscripts released before undergoing peer review.

What Happened That Day — Where the Papers Point

The proceedings of that day followed the primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together". An electrical grid, called the system in industry terms, is a network that connects power plants and places that use electricity with wires so they run as one. The system's frequency is the count of how many times a second the direction of the alternating current swings, and it moves when the amount being made and the amount being used do not match. That day, the Iberian system began to break down from overvoltage, where the voltage gets too high, and as power generation facilities dropped out one after another, the frequency was dragged down. The picture of this beginning is the same in two analyses (DOI: 10.2139/ssrn.6895119, preprint) (DOI: 10.2139/ssrn.6551460, preprint). As for the process that amplified the collapse and its roots, the papers point to different layers.

An analysis that reconstructed the event second by second using public waveform data estimates that the frequency, whose normal value is 50 Hz, dropped to about 48.5 Hz, and concludes that a chain starting on the voltage side closed off the time to act before frequency reserves could even react (DOI: 10.2139/ssrn.6895119, preprint based on public data reconstruction and estimation). This 48.5 Hz is an estimate reconstructed from public data. The official final report introduced by the primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together" notes that the frequency dropped below 48 hertz.

A journal paper combining high-resolution measurement analysis and event reconstruction looks at the amplification process separately. Under conditions where synchronous generation was reduced, rapid oscillations appeared between regions, accelerating the frequency drop. On top of that, grid-following inverters, protection settings tailored to automatically disconnect facilities if the frequency changes quickly, and limited interconnection capacity interlocked to contribute to the cascading trip of generation and load. The authors note that the reduced effective inertia vastly cut down the time available for corrective measures (DOI: 10.24084/reepqj26-376). These three points will be unpacked one by one in the later sections of this article.

A preprint summarizing policy implications based on the ENTSO-E investigation finds the root in the institutions. Even though there were renewable energy facilities certified with the ability to regulate voltage, that ability could not be used due to regulatory barriers. It follows by pointing out that market design conflicted with real-time stability requirements, and crisis response management systems were fragmented (DOI: 10.2139/ssrn.6551460, preprint). An editorial in an academic journal views that despite having robust facilities and dispatchable generation capacity, steep solar output variations, low demand, and limited voltage regulation overlapped to cause the collapse (DOI: 10.32397/tesea.vol6.n1.911, editorial).

Among the papers read for this article, none cite the share of renewable energy itself as the cause. Even in the closest editorial, solar output variations are placed side by side with low demand and limited voltage regulation. The public waveform reconstruction analysis notes that this event exposed the vulnerabilities of a grid with a high share of renewables and converter dominance—namely, its susceptibility to rapid frequency drops and voltage instability (DOI: 10.2139/ssrn.6895119, preprint). What this analysis notes is the vulnerability of such a grid, rather than the share as a cause. The final report introduced by the primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together" also does not single out one cause, and includes the operation of renewable facilities with fixed settings that do not respond to voltage changes as one of several factors.

The opposite conclusion—that it has "nothing to do with renewable energy"—is likewise not supported by these papers. When generation that connects to the grid through power electronics instead of rotating generators increases, the system's inertia decreases, and when inertia decreases, the frequency collapses faster even during an accident of the same size (DOI: 10.3390/en19133063, review) (DOI: 10.24084/reepqj26-376). Therefore, what we must ask, rather than the name tag of the generation source, is what properties that facility gives to the grid. The first of those properties is inertia.

The Things That Hold On Without Anyone Telling Them To

The primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together" compared the grid to a giant merry-go-round pushed by many people together. The people pushing are synchronous generators, meaning generators whose rotation speed matches step with the grid's frequency, and the speed at which the board spins is the frequency. Coal, gas, nuclear, and hydro power plants all end up with rotors weighing tens of tons spinning to the rhythm of the grid. If more electricity is used, that burden immediately grabs the rotor, slowing its rotation and dropping the frequency.

As that primer defined, grid inertia is the rotational energy possessed by the chunks of metal spinning while directly connected to the grid, and this slows down the rate at which the frequency changes without any instructions. Even if a single generator suddenly drops out, the remaining chunks of metal give up a little of their rotational energy at a time to slow the descent. No one ordered it, and neither communications nor controllers intervene. Equipment that increases generator output, reserves, and an operator's judgment all take time to move. Inertia is the first cushion that operates ahead of all those responses, waiting for no one.

The magnitude of inertia is usually written in units of seconds. It is a value obtained by dividing the rotational energy stored in the grid by the size of the power generation facilities attached to the grid, so it is read as how many seconds the power generation facilities can produce electricity equal to their own size using only that rotational energy. The preceding reconstruction analysis estimated the inertia of the Iberian grid that day to be 2.21 to 2.71 seconds. This is an estimated range obtained by a method of calculating backward from the speed at which the frequency moved during the accident and a method of summing the values of the facilities attached to the grid (DOI: 10.2139/ssrn.6895119, preprint, estimate).

What changes when inertia decreases is read through the second tick seen in the primer. The rate of change of frequency (RoCoF) is a value measuring how many hertz per second the frequency is collapsing, in other words, the speed of frequency change. The smaller the inertia, the greater the rate of change of frequency for the same accident, and this is the starting point of the low-inertia grid problem. Following that, the dangers of cascading failures and the operation of protection devices trail along (DOI: 10.3390/en19133063, review).

If the rate of change of frequency is large, as the primer showed, the time remaining until measures such as cutting off loads or reducing generation becomes short (DOI: 10.3390/en13040816, simulation). In addition to that, the rate of change of frequency itself causes equipment to drop out. Among protection devices, some are set to disconnect their equipment from the grid if the frequency changes too quickly. When inertia becomes thin, it becomes easier to reach that setting value even in the same accident. A journal paper covering the Iberian incident also counts these protection settings as one of the factors that contributed to the cascading trips (DOI: 10.24084/reepqj26-376). This means that in a grid with thin inertia, the fact that the frequency changes quickly can itself become the trigger for the next dropout.

The point where the frequency hits its lowest as it goes down is viewed separately. What stops the fall is the primary frequency response that comes in afterward, in other words, the immediate reaction that generators and electricity consumers show without anyone telling them to when the frequency deviates from the standard. Inertia slows down the falling speed to buy time for that response to rise, and the primary frequency response stops the fall. The lowest point is determined by the two together. This distinction is used again later when separating the two responses of inverters.

Inertia Is Decreasing — But Not Uniformly

Disturbance records for the California grid from 2013 to 2024 were introduced in the primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together". A decreasing trend in inertia was visible, and the local rate of change of frequency was up to 6 times higher than the value for the interconnected grid as a whole. The authors attribute the recent rise in inertia back to the deployment of battery storage systems equipped with inertia-mimicking functions (arXiv:2511.21387v1, preprint based on empirical data estimation). The number 6 means that even if the system-wide average looks fine, a specific region can collapse much more steeply.

When inertia is geographically uneven, frequency stability becomes a local problem in addition to a system-wide one. Dispatch planning, which determines in advance which generators to run and when, usually approximates the entire grid as a single average frequency, and plans made that way protect weak regions less (DOI: 10.2139/ssrn.6957145, preprint simulation). With inertia, where it is located matters just as much as how much of it is in the grid overall.

There is no single number for what percentage of renewable energy makes things dangerous. The conclusion of a study that simulated the continental European grid for every hour of the year across numerous weather scenarios was that a single threshold percentage cannot be defined and must be recalculated for every dispatch interval (DOI: 10.1049/iet-rpg.2015.0141). The percentage figures appearing in this article should also all be read only as values for specific grids under specific conditions.

There are also gaps in measuring inertia. Inertia is often calculated backward from the movement of frequency after a disturbance. In a simulation comparing various estimation methods, any method was accurate when looking only at rotating generators directly tied to the grid. When frequency control functions were added to wind power disconnected from the grid, large discrepancies arose between the methods. The authors argue that a new estimation approach that even includes simulated inertia is needed (arXiv:2005.02091v1, preprint simulation). This means that once inverters start mimicking inertia, even the task of measuring how much inertia remains in the grid becomes difficult.

Inverters Used to Only Follow

As defined in the primer, an inverter is a device that changes direct current into alternating current and feeds it into the power grid. Solar panels, batteries, and most modern wind turbines connect to the grid through this device, and by themselves, they do not add weight to the merry-go-round. Rotating generators spin while tied to the grid to create frequency, but an inverter's waveform is drawn by software. Therefore, how an inverter behaves during an accident is determined more by its control method than by physics.

A grid-following inverter is an inverter that looks at the voltage already in the grid, matches its own beat to it, and then pushes in a set amount of power. The primer called this the "method of following someone else's beat," which, in terms of the merry-go-round, is a machine that first checks the speed of the board and then pushes from the side to match it. Because it needs a beat to follow in order to follow, this method relies on a solid external grid to keep the beat (DOI: 10.63090/ijtrs/3139.1788.0014, small-scale journal).

A grid-forming inverter is an inverter that, instead of following someone else's beat, establishes its own voltage and frequency on its own and makes other equipment use it as a reference. Because it can spin without the reference of a pre-existing grid, it is widely cited as a technology that will make grids with a very high proportion of renewable energy possible (DOI: 10.63090/ijtrs/3139.1788.0014, small-scale journal). While rotating generators were plentiful, there was always a beat to follow. At the time of a 2024 review paper, grid-following was a long-operated method, and the relatively new grid-forming was still sparsely deployed. The same review paper also analyzes real-grid demonstration cases (DOI: 10.1049/rpg2.12991, review paper).

The weakness of the following method is revealed when the grid is weak. Grid strength means how firmly the surrounding grid holds the voltage of a specific point when something happens there, and it is usually measured by the magnitude of current that can flow into that point when a fault occurs. Inside a grid-following inverter, there is a circuit that looks at the grid voltage and matches the beat, and if the grid is weak, this circuit has a high probability of missing the beat. Therefore, the authors suggest that during the fluctuations immediately after an accident, nearby grid-forming equipment must support this (DOI: 10.3390/pr13123816, simulation). This means that a following inverter performs its role only when a party that produces the beat is nearby.

The comparison between the two methods does not always lean to one side. A hardware test in which two commercial inverters were run alternately in both modes while changing grid conditions is an example. The property of a line blocking current includes resistance lost as heat and the property of blocking changes in alternating current. In the test, the ratio of these two affected voltage stability more than grid strength did. Changing toward a side where the share taken up by resistance becomes relatively larger was disadvantageous to grid-forming and advantageous to grid-following. The authors interpret that grid-forming may be more suited to the transmission side that carries large amounts of electricity far away, while grid-following may be more suited to the distribution side that divides it into neighborhoods (DOI: 10.20944/preprints202504.1145.v1, preprint, hardware test with 2 commercial devices). This means which method is better can differ depending on the line where the inverter is attached.

Grid-forming also has limitations. A low-inertia grid case study incorporating synchronous machines and multiple grid-forming converters together confirmed that grid-forming helps the frequency stability of synchronous machines. The same study notes that because there is a limit to the current the converters can flow, how that limit is applied is decisive. The interactions that fast grid-forming converters and slow synchronous machines exchange can even contribute to instability (arXiv:2003.04715v1, preprint, case study).

Synthetic Inertia and Fast Frequency Response — Two Ways of Measuring and Then Pushing

There are multiple ways to make an inverter mimic inertia, and two different ones are frequently lumped together under the same term.

Synthetic inertia is when an inverter measures the frequency, calculates its rate of change, and momentarily outputs power proportional to that rate to mimic rotational inertia. This is the method seen in the primer, where a battery outputs electricity in proportion to the speed at which the frequency drops. The difference from real inertia lies in the measurement stage. A chunk of metal does not know what the frequency is, does not calculate it, and does not wait for a signal. The inertia of a chunk of metal resists without measuring, whereas synthetic inertia measures and then pushes.

Therefore, what is measured and how it is measured determines the performance of synthetic inertia. Synthetic inertia based on frequency differentiation uses dedicated circuitry to extract the grid frequency. When the signal-to-noise ratio—which indicates how much larger the intended signal is than the background noise—is low, this approach faces the usual challenge of filtering out the input signal (DOI: 10.3390/en18143776, experimental and field measurement). To filter out noise, a filter must be used, and a signal that has passed through a filter comes out correspondingly delayed. The point that filters create delays is a general property of signal processing, and it is not a result measured by that paper.

Another approach compared by the same study amplifies the inertial response that naturally emerges from actual synchronous machines. Thus, it can avoid the input-side filter, but a filter becomes necessary on the output side instead. The two were compared using a microgrid experiment and field measurements of a 55 MVA hydroelectric generator connected to the Nordic system. In the specific cases compared, the authors observed that the amplification approach performed better in smaller systems with large frequency fluctuations, while synthetic inertia performed better in larger systems (DOI: 10.3390/en18143776, experimental and field measurement). This means that whether synthetic inertia is useful can depend on the size of the system in which it is placed.

Fast frequency response is looking at how far the frequency has deviated from the baseline and pushing a set amount of power within a set time. In terms of the merry-go-round, this is the side that looks at how much slower the merry-go-round has become and adds force by a pre-established amount.

A study comparing the two from the perspective of a system operator puts forward the very act of clearly distinguishing the meanings of the two concepts as its contribution. The subject of review is the Nordic system in a future scenario with large-scale wind power integrated (DOI: 10.1049/iet-rpg.2017.0370). Because the abstract of this paper does not state the details of the distinction, the four paths below were gathered by this article from other studies.

First, what they look at is different. One looks at the rate of change, while the other looks at the degree of deviation from the baseline. This contrast is already contained in the names each of the two studies attached, with one calling synthetic inertia frequency derivative-based (DOI: 10.3390/en18143776), and the other calling fast frequency response frequency deviation-based (DOI: 10.1049/iet-rpg.2017.0370).

Second, delay dictates effectiveness. A modeling study analyzing fast frequency response examined the effects of response delay and power ramping speed for each resource on the effectiveness of frequency control, and outlined desirable specifications. The same study also showed that deploying this response can significantly reduce the reserves that need to be held in advance (DOI: 10.1049/joe.2018.8599, model-based analysis).

Third, the constraints they handle are different. In a simulation solving the problem of deploying virtual inertia and fast frequency response together, two constraints of frequency stability determined the required amounts of each different resource. The two milestones separated earlier—namely, the rate of drop and the lowest point hit—are those two constraints. For pulling up the lowest point, fast frequency response was far more cost-efficient than virtual inertia. What this paper calls virtual inertia is emulated inertia. At a 70% renewable energy share, the virtual inertia demanded by this model was 2.89 seconds, and as the share rose, the required amounts of both resources continued to increase (DOI: 10.3390/en19081848, simulation).

Fourth, they can only deliver if they have spare capacity. This applies to both approaches. Rotational inertia comes from a chunk of metal that is already spinning, but for an inverter to momentarily output more, it must reserve a corresponding amount somewhere. The inertia that a wind turbine can deliver depends on its operating state at that moment, separate control is required to retain as much inertia as possible below a set output, and exiting the inertia control at the right time becomes a separate issue (DOI: 10.3390/en13051067, simulation). A study running large-scale future scenarios of the Australian system compared three options: equipment that only spins without generating power, synthetic inertia from wind, and wind response with output intentionally reduced to create spare capacity. The most effective option was the last one, the side that left room for what it would produce in the first place (arXiv:1708.00739v1, preprint and scenario analysis).

There is also a wall in the time it takes to measure. In the simulation introduced by the primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together", when an imbalance of up to 40% occurred in a system where inverter-connected generation exceeds 80%, fast reserves had to fully emerge within 100 ms (in both simulated system sizes). The authors conclude that current frequency measurement techniques and injection times do not guarantee stability under such large imbalances (DOI: 10.3390/en13040816, simulation). The method of measuring and then pushing is delayed by the time it takes to measure, and in this simulation, the very large imbalances in an inverter-dominated system could not be compensated for by the technology of 2020.

It Was Not Just a Matter of Inertia

When a fault occurs, the first thing the grid does is to cut off the faulted location. If a tree falls on a power line, that section must be quickly disconnected so the rest survives. The device that makes this judgment is a protective relay, and relays usually figure out a fault by seeing a large current rushing into that spot at the moment of the fault. Rotating generators can supply such a large current at the moment of a fault. Inverters have a limit to the current their semiconductor components can withstand, so they cannot match the level of fault current that rotating generators used to provide. Expressed in terms of grid strength defined earlier, the smaller the current that will flow during a fault, the weaker the grid at that location.

According to a study analyzing a scenario where the Puerto Rico grid is converted to 100% inverter-based generation, generation connected via inverters cannot match the level of fault current that rotating synchronous generation used to supply. Therefore, the authors ask whether existing protection schemes remain reliable during that transition. The same study also quantifies methods to boost the fault current using facilities that only spin without generating power or voltage compensation equipment (DOI: 10.3390/electronics13071225).

When the fault current is low, the problems appear concretely. Low fault currents fail to trigger protective relays, fail to prop up the voltage during a fault, and also fail to properly help voltage recovery after the fault is cleared (DOI: 10.1049/rpg2.12149, simulation).

There are even more subtle problems. The characteristics of grid-forming equipment change abruptly the moment current limiting is engaged, and then the quantities that relays observed and judged behave differently than before the fault. The reliability of the relay monitoring elements that rely on those quantities is shaken to that extent. The authors propose a method to adjust control during faults so that it does not conflict with grid protection, and verify it through hardware-in-the-loop testing (arXiv:2504.21592v1, preprint, hardware-in-the-loop test).

The scale used to measure grid strength is also shaking. Existing grid strength evaluation methods were created on the premise that there are synchronous generators supporting the voltage, so they do not apply to a 100% inverter grid where synchronous generators are entirely absent. The same paper reports that expanding an index previously developed by the authors makes it possible to evaluate stability against small disturbances even in such grids, and shows its effectiveness by comparing the results of two different analysis methods in a simulated grid (DOI: 10.36227/techrxiv.21120517.v2, preprint, simulation validation). Just like inertia estimation, a new scale to measure what is disappearing must be created, and that work has begun at the preprint stage.

The final layer is ride-through. Fault ride-through is the ability of equipment not to trip and to remain connected when it encounters voltage and frequency anomalies in the grid. A study validating a procedure for testing this ability with commercial solar inverters notes why this became a standard. As the generation share of distributed energy resources rises, the risk grows that a single grid fault will simultaneously trip many or all distributed resources, leading to a large supply-demand imbalance and cascading blackouts. The authors demonstrated an automated procedure to verify the ride-through requirements for voltage, frequency, and rate of change of frequency, respectively, using two commercial inverters (DOI: 10.3390/en14216936).

What actually happened in Iberia also meets in that both reconstruction analyses show it was a cascading trip of facilities. One analysis puts weight on the fact that a chain starting from the voltage side closed the window of time to take action, while the other emphasizes the subsequent cascading factors: inverters, protection settings, interconnection capacity, and reduced inertia (DOI: 10.2139/ssrn.6895119; DOI: 10.24084/reepqj26-376, both post-event reconstructions). Inertia was one of several layers on that day, entangled together with voltage regulation, protection settings, and regulations. However, the simulation in the official final report introduced in the primer "Alternating Current and Frequency — A Giant Merry-Go-Round the Whole Country Pushes Together" suggests that even if inertia had been much greater, the sequence in which the events unfolded on that day could not have prevented the Iberian grid from losing step with the continental European grid. That reduced inertia shortened the time to take action and that it could have been prevented if inertia had been ample are two different things.

Common Misconception Four

Misconception ① "If there is a lot of renewable energy, blackouts happen." The papers this article read do not list the share of renewable energy itself as the cause. What the papers list is a list of properties: inertia, the ability to support voltage, fault current, and ride-through settings. What we must ask is whether there are enough facilities with those properties, and the name tag of the power source is just a crude stand-in for that question. As shown by renewable energy facilities that were certified for voltage regulation capabilities on a given day yet blocked from use by regulatory barriers (DOI: 10.2139/ssrn.6551460, preprint), the same facility can be part of the problem under one system and part of the solution under another.

Misconception ② "So it has nothing to do with renewable energy." This is also wrong. The relationship that as generation passing through power electronics increases, grid inertia decreases comes from both measured data and continental-scale simulations (arXiv:2511.21387v1, measurement-based estimation; DOI: 10.1049/iet-rpg.2015.0141, simulation). That there is a relationship and that it is the cause on a given day are different things, and what the papers read by this article support is up to the preceding sentence.

Misconception ③ "If you deploy lots of batteries, inertia is created." Batteries do not create inertia; they mimic it. To mimic it, frequency must first be measured, and when the signal-to-noise ratio is low, filtering out that measurement is tricky. However, in the compared cases, synthetic inertia performed better in larger grids (DOI: 10.3390/en18143776, experiment and field measurement). Mimicked inertia and fast frequency response handle different constraints (DOI: 10.3390/en19081848, simulation). This does not mean batteries lack power. The recent recovery of inertia in California was attributed by the authors to the widespread deployment of battery storage systems equipped with inertia-mimicking functions (arXiv:2511.21387v1, preprint).

Misconception ④ "This is all still just theory." As of a 2020 report, real-time frequency stability tools were in operation in the control room of the Southwestern Australia grid, and in various cases they accurately predicted frequency trajectories following incidents (arXiv:2010.14016v1, preprint). However, techniques that mimic inertia have varying levels of technological maturity (DOI: 10.3390/en19133063, review).

Remaining Debates — Buying It Versus Mandating It

To buy inertia in a market, you must first be able to write it down as a price, but there is no agreement even on its definition. Ancillary services are services purchased to keep the grid stable, separately from electricity itself. According to research addressing the issue of procuring inertia as an ancillary service, despite much effort from academia and the field, there is no agreement on the definition and treatment of inertial services in market operations. It is especially unsettled what to set the inertia variable as, and how to divide the inertia of synchronous generators and the virtual inertia of inverters (arXiv:2601.04504v1, preprint, simulation).

The same study proposes a definition based on inertial response, even including negative inertial response while the frequency is dropping within the definition. This reflects in the pricing system that inertial response does not always come out in a helpful direction. The authors report that this bidirectional definition creates a price signal reflecting the economic value of providing inertia (arXiv:2601.04504v1, preprint, simulation).

Without a definition, there is no price, and without a price, procurement goes awry. A study modeling the procurement problem of the British grid reports that if the inertia level is not considered together when purchasing frequency response, the cost of procuring ancillary services can increase by 165%. The authors also point out that Britain's ancillary service market opens one month in advance (DOI: 10.52843/cassyni.jxg0nc, non-peer-reviewed published work, model-based research).

The other path is to mandate it instead of buying it. An analysis reconstructing the Iberian incident recommends that inverter-dominant regions need coordinated grid-forming capability requirements, location-reflective reactive power and fast frequency response procurement, and a mandatory lower limit on inertia (DOI: 10.2139/ssrn.6895119, preprint). A journal article examining the amplification process of the same incident also concludes that criteria for judging whether inertia and fast frequency response are sufficient must be specified, grid-forming capabilities deployed, and protection settings adjusted to an environment where frequency changes rapidly (DOI: 10.24084/reepqj26-376). Enforcing it as a requirement can sidestep the definition debate. Instead, the question of who bears that cost remains unchanged.

There are also attempts to provide answers in numbers. In a simulation of conditions where the grid is 100% filled with inverters, when 30% of them were grid-forming, the maximum rate of change of frequency was suppressed to 0.58 Hz/s, and when all of them were grid-following, it reached up to 2.9 Hz/s (DOI: 10.63090/ijtrs/3139.1788.0014, simulation, small-scale journal). The direction is clear. However, this 30% is a value from a single paper derived under a specific baseline grid, specific control methods, and specific fault conditions, so it cannot be directly transposed into a policy number.

What Has Vanished

What has vanished is a few seconds for which no one paid.

For decades, the grid did not budget for those few seconds. It did not need to. When you bought a generator, a chunk of metal came with it, and that chunk of metal delivered rotational energy without any instructions. Inertia was more of a byproduct than a product.

As buying a generator stopped coming with a chunk of metal, what used to be a byproduct became something you had to buy separately or order separately. And the moment you try to buy it, it becomes clear that there is still no agreement on even how to define and sell it in the market (arXiv:2601.04504v1, preprint and simulation).

There are five questions that can be answered. First, how many seconds are left in the grid right now? Second, in which region are those seconds left? Third, do the facilities claiming they will take that place react after measuring or without measuring? Fourth, if they react after measuring, how fast are they? Fifth, which is a question that remains even after defending the frequency, when a fault occurs, can that spot be cleanly cut off, and can the rest endure while it is being cut off?

"Is Having Lots of Renewable Energy Dangerous?" is a question that cannot be answered as it is. What can be answered are those five lines, and each line contains the properties that the facility gives to the grid instead of the name of the power source.

Source Papers

Articles published on arXiv, SSRN, and Research Square are marked as "preprint" and whether they were subsequently published in an academic journal was not separately verified. Most studies covering this event are preprints, and other tiers are indicated in the body text each time.

• "Excess mortality attributable to the 2025 Iberian blackout" (2026) — DOI: 10.1038/s41467-026-75581-w — Spain +167 people (95% confidence interval +28 to +300) two days after, Portugal not observed.
• "The Iberian Blackout of 28 April 2025" (2026, preprint) — DOI: 10.2139/ssrn.6895119 — Reconstruction of the public waveform. Lowest frequency estimated at roughly 48.5 Hz (different from official report value), estimated inertia 2.21 to 2.71 seconds, recommendation for an inertia lower bound.
• "The 2025 Iberian Peninsula Blackout: Policy Implications" (2026, preprint) — DOI: 10.2139/ssrn.6551460 — Attributes root cause to institutional failure. Certified voltage control capabilities were blocked by regulations.
• "Grid Resilience under high shares of Non-Synchronous Generation" (2026) — DOI: 10.24084/reepqj26-376 — Following resources, RoCoF-sensitive protection, and intertie constraints contributed to cascading disconnections.
• "Challenges of variable energy resource integration" (2025, editorial) — DOI: 10.32397/tesea.vol6.n1.911 — Points out solar surges and drops, low demand, and limited voltage control side by side.
• "Review of Virtual Inertia Techniques" (2026, review) — DOI: 10.3390/en19133063 — Inertia reduction → RoCoF increase → cascading failures and protection device operation. Classifies techniques by maturity.
• "Required Power Response of Inverters for FFR Support" (2020) — DOI: 10.3390/en13040816 — If inverters exceed 80% and imbalance is 40%, full activation is needed in under 100 ms, which cannot be guaranteed by current techniques.
• "Regional Inertia Dynamics in CAISO" (2025, preprint) — arXiv:2511.21387v1 — Actual measurements from 2013 to 2024. Regional RoCoF up to 6 times. Recovery attributed to synthetic inertia batteries.
• "Inertia-Inhomogeneity-Aware Unit Commitment" (2026, preprint) — DOI: 10.2139/ssrn.6957145 — Areas where average frequency approximation is weak are underprotected.
• "Impact of high penetration of variable renewables" (2016) — DOI: 10.1049/iet-rpg.2015.0141 — A single critical penetration threshold cannot be defined. The share of load self-regulation also grows.
• "Analysis of power system inertia estimation" (2020, preprint) — arXiv:2005.02091v1 — When wind frequency control is included, estimation techniques misalign. Alternatives are needed.
• "Grid Strength Assessment for 100% Inverter-Based Systems" (2022, preprint) — DOI: 10.36227/techrxiv.21120517.v2 — Existing evaluation methods presuppose the voltage support of synchronous generators and thus cannot be applied. However, it shows that that grid can also be evaluated with an expanded generalized short-circuit ratio.
• "Grid-forming control for IBRs: A review" (2024, review) — DOI: 10.1049/rpg2.12991 — Forming is still rarely deployed. Also analyzes real-grid demonstration cases together.
• "Coordinated FRT Control for GFL and GFM in Weak Grid" (2025) — DOI: 10.3390/pr13123816 — In a weak grid, following's synchronization circuit has a high probability of missing a beat.
• "Impact of SCR and X/R on Stability of GFM and GFL Inverters" (2025, preprint) — DOI: 10.20944/preprints202504.1145.v1 — Tests on two commercial units. Lowering X/R is disadvantageous for forming and advantageous for following.
• "Frequency Stability of Synchronous Machines and GFM Converters" (2020, preprint) — arXiv:2003.04715v1 — Alongside benefits, the interaction between fast converters and slow synchronous machines can contribute to instability.
• "Amplified Inertia Response vs Synthetic Inertia" (2025) — DOI: 10.3390/en18143776 — At low SNR, synthetic inertia suffers from input filtering challenges. However, in large grids, synthetic inertia prevails.
• "Synthetic inertia versus fast frequency response: a definition" (2017) — DOI: 10.1049/iet-rpg.2017.0370 — Presents the distinction between the meanings of the two concepts itself as a contribution (distinction details outside abstract).
• "Fast frequency response for effective frequency control" (2018, model-based analysis) — DOI: 10.1049/joe.2018.8599 — Analyzes the impacts of response delay and rate of rise to summarize desirable specifications.
• "Optimal Configuration of Virtual Inertia and FFR" (2026) — DOI: 10.3390/en19081848 — The two constraints determine the requirements of different resources. FFR has the advantage for nadir improvement, virtual inertia 2.89 seconds at 70%.
• "A Framework for Frequency Stability Assessment" (2017, preprint) — arXiv:1708.00739v1 — Among synchronous condensers, wind synthetic inertia, and de-loaded wind, the last is the most effective.
• "Optimal Inertia Reserve and Control for Wind Farms" (2020) — DOI: 10.3390/en13051067 — The inertia wind power can provide varies depending on the operating point, and the disconnection timing is a separate issue.
• "System Strength Reduction in an Island Grid" (2024) — DOI: 10.3390/electronics13071225 — Inverters cannot match the short-circuit level of rotating generation. Quantifies supplementary facilities.
• "Optimal short-circuit current control of the GFM converter" (2021) — DOI: 10.1049/rpg2.12149 — Low short-circuit current fails to trip protective relays and voltage support is also weak.
• "A Protection-Interoperable Fault Ride-Through Control" (2025, preprint) — arXiv:2504.21592v1 — Overcurrent limitation shakes the reliability of relay monitoring elements.
• "Commercial PV Inverter IEEE 1547.1 Ride-Through" (2021) — DOI: 10.3390/en14216936 — Risk of multiple or all units dropping simultaneously from a single fault.
• "Online Security Assessment of Low-Inertia Power Systems" (2020, preprint) — arXiv:2010.14016v1 — Operated in the control room of the southwestern grid of Australia as of 2020.
• "Definition and Formulation of Inertia Service" (2026, preprint) — arXiv:2601.04504v1 — There is no consensus on definition and treatment in market operations. Proposes a bidirectional definition.
• "Linking Inertia and Frequency Response Procurement" (2023, non-peer-reviewed posting/model-based) — DOI: 10.52843/cassyni.jxg0nc — Ancillary service costs can increase by 165% if inertia is not considered.
• "Grid-Forming Inverter Control" (2026, small journal/simulation) — DOI: 10.63090/ijtrs/3139.1788.0014 — With 100% inverters, 30% forming yields a maximum RoCoF of 0.58 Hz/s, whereas if all are following it is 2.9 Hz/s.