Warnings have emerged of the potential for further blackouts in Spain after Red Eléctrica de España (REE) reported steep voltage swings in recent weeks during periods of low demand, high solar output and slow response from generation. REE has asked the regulator (CNMC) for exceptional, time-limited powers to manage voltage while longer-term reforms are designed.
The request relates to four operative procedures: PO 3.1 on programming and scheduling, PO 3.2 on technical constraints, PO 7.2 on secondary regulation and PO 7.4 on voltage control. The CNMC notes that the measures could materially affect balancing, voltage control services and commercial opportunities across all markets. Tighter service conditions could shrink supply, reduce market competitiveness and raise costs to demand. For that reason the CNMC wants any approval to be clearly exceptional and temporary, pending a short consultation and deeper analysis of causes and remedies.
What this means in practice is that REE has forgotten how the grid works. Synchronous generators, as I explained in my last blog Ghosts on the grid: why the phantom concept of vars risks our energy security, dampen voltage disturbances as well as frequency disturbances. Essentially “inertia” applies to voltage AND frequency, but unlike frequency, voltage is local, so the location of synchronous generators matters. REE has allowed the Spanish power grid to become dangerously unbalanced and Scotland is heading in the same direction.
Recap of the April blackout and the new powers sought by REE
In April the Iberian system suffered a full collapse in which eleven people died. A faulty solar inverter triggered voltage and frequency oscillations. REE instructed both renewable and conventional units to deliver reactive power under PO 7.4, but many did not. Initial stabilisation was achieved, but when solar output was curtailed in response to negative prices frequency began to fall and the voltage disturbance recurred. As the system hunted for equilibrium, Spanish frequency dipped into the 49.6 to 49.5 Hz band, comfortably inside the ride-through envelope required by PO 7.3 and the EU RfG.
A large number of inverter-based generation (mostly solar PV and some wind) tripped as frequency crossed 49.5 Hz even though the code requires immunity down to 47.5 Hz for twenty seconds. Roughly 8 GW vanished within about five seconds, sending frequency through 48 Hz – automatic load shedding followed and the peninsula split. Conventional machines did not begin to trip until below 47.5 Hz, which complied with the code as frequency had already fallen below the level they were required to withstand. The real cause of the blackout was the widespread failure of renewable generators to remain connected during a modest frequency deviation. Without those premature trips, the grid would probably have survived the voltage disturbance.
A government report by the 28A-Committee has different perspectives on events that day, placing more blame on REE for not scheduling enough reactive power and in particular not reacting to a maintenance outage announced by one of the plants scheduled to provide it. Both that report and REE’s own analysis identify generating units that disconnected before voltages reached regulated limits. REE explicitly attributes several losses to incorrect triggers and notes instances of forced oscillations likely due to internal plant anomalies. It also records failures under PO 7.4, including a unit injecting reactive power when it had been instructed to absorb it, and emphasises that its calculations were based on an assumption that all rules would be complied with, an attempt to shift responsibility for the failure to the service providers.
The new powers seek to close the gaps exposed in April. The direction is for stronger operational levers for dynamic voltage and reactive control under stressed conditions, real-time transparency through daily reports to the CNMC, and acceleration of medium-term rule changes so that compliance by inverter-based and other non-synchronous sources is enforceable in practice rather than only in theory. Specifically:
- Stricter operational procedures for voltage / reactive control: giving REE more leeway to enforce dynamic voltage control obligations under extreme conditions, allowing REE to more aggressively demand that generators operate in compliance, eg instruct setpoints, impose stricter reactive power margins, etc
- Improved monitoring, reporting, oversight: REE must provide daily reports to CNMC on how the exceptional measures are being used and monitor their effects, to increase transparency and accountability in real time
- Reinforcement of normative / regulatory changes in the medium term: progress with plans to update normative frameworks so that grid code compliance, particularly for inverter-based and non-synchronous sources, becomes more robustly enforceable.
Yet the proposals are explicitly temporary. They emphasise operational flexibility more than deterrence. Unless enforcement, penalties and disconnection powers are sharpened, compliance incentives will remain weak. There is also a tension with market signals, as units that are compelled to provide voltage support without appropriate remuneration will be reluctant to comply. Distributed IBR fleets complicate monitoring, modelling and remote control. Any new powers must also meet Spanish legal tests of proportionality and withstand challenge.
Inertia and voltage control in weak grids, and why location matters
REE and the 28A Committee both say the April collapse was not an inertia problem, and REE says current disturbances are not about inertia either. That is not correct. Frequency reflects active-power balance, while voltage reflects reactive balance – in weak grids they couple. Synchronous machines damp both because the excitation fields and torque of synchronous machines both respond to voltage, while inverters rely on phase-locked loops (“PLLs”) to track grid voltage and phase and voltage oscillations disturb that tracking, altering their active-power injection.
In a stable grid reactive power and the response of equipment (eg excitation systems, inverter voltage control) create negative feedback – if voltage rises, reactive absorption increases and voltage is pulled back down. But in an unstable grid that feedback turns positive as the controllers or network reactances push voltage further in the same direction.
Lower inertia speeds up the system’s electromagnetic state changes. Faster frequency movement means larger instantaneous swings in terminal voltage and PLL phase. Weakly damped PLLs convert fast frequency deviations into phase noise of voltage, exciting electromechanical modes that appear as slow voltage ripples. Without heavy synchronous torque the voltage–current phase relationship can “wander” as voltage magnitude changes are no longer anchored by mechanical stiffness.
Inertia therefore provides phase damping as well as frequency damping – remove it and the voltage-control problem becomes highly sensitive. REE is right that the trigger in April was reactive, but low effective inertia worsened damping and allowed the swings to grow.
On paper reactive reserves may look adequate, but in practice dynamics matter. Rapid solar ramping, price-driven curtailment and wind variability create fast active-power changes. Volt-var and volt-watt controls across many IBRs can fight each other, especially when capacitor banks and long cables inject energy under light load (Ferranti effect), resulting in an aggregated flicker and low-frequency voltage oscillation.
REE has observed exactly these behaviours and is now seeking stronger tools to manage them. Despite downplaying inertia, it has quietly instructed the operation of more gas units since April to raise inertia and damping, an implicit admission that the grid had become too “light”.
Location becomes far more important in weak grids. Frequency is broadly uniform across the grid, so the location of synchronous generators does not matter. Voltage, however, is a local property that must be supported by local equipment. However, Spain’s conventional generation is concentrated in the north and east, while the south is dominated by renewables, making the southern network weak and increasingly difficult to control.
Spain’s transmission corridors are long, lightly loaded, and reactive-dominant (impedance is much greater than resistance). PV output in Castilla–La Mancha and Extremadura can vary by hundreds of MW in seconds under cumulus conditions. When a large PV cluster suddenly cuts output (due to market signals or a protection trip), voltage at the export node can rise, because the line charging of grid capacitance is no longer counterbalanced by inductive current from the inverters.
The solution is not purely procedural. Yes, REE must test and enforce grid code compliance far more rigorously with regular audits of generator responses to control instructions, periodic testing of fault-ride-through capability, and meaningful penalties for non-performance. But it must also address the structural imbalance in the grid. New conventional generation in the south is unlikely, but synchronous condensers (large spinning machines that stabilise voltage without producing electricity) and modern electronic devices such as STATCOMs can provide the same service.
Grid forming power electronics are simply not mature today and too little is known about their behaviour
Inevitably, my social media has been full of people claiming that batteries and grid forming power electronics (“GFPE”) are the answer. As I described in this recent post, the use of GFPE are limited because neither the economic frameworks or technical capabilities yet exist. Where GFPE are installed they are not actually forming the grid.
A 2024 report by Transgrid which operates the high voltage transmission systems in New South Wales and the Australian Capital Territory, notes that, in relation to short-circuit level controls that “…practical examples of such designs are known where the inverter will effectively switchover from the GFM [grid forming] to GFL [grid following] during fault conditions, where the GFM capability might be most necessary. Even if an indirect current limiter is applied, in many cases GFM stability is degraded when operating at or above the rated current”.
All IBRs need to maintain their total current within safe levels to protect semiconducting switching devices. Depending on the design, grid forming inverters may or may not have a direct current limiter – for those that do, the inverter can behave like a grid following device when the current limit is reached for example during faults. While it is proven that grid forming inverters can provide several grid support functions, a well performing grid forming inverter in terms of stability will always operate below the rated current which isn’t ideal for meeting minimum fault level requirements.
“Noting the risks discussed and as the impact and mitigation measures are not currently known deterministically, it is recommended to exclude GFM BESS and STATCOM from meeting the minimum [system strength] level,”
– Advice on the maturity of grid forming inverter solutions for system strength, Transgrid
Surprisingly, grid forming performance is also degraded under high system strength conditions. This is counterintuitive given the limitations of grid following inverters under low system strength conditions, but can be explained by differentiating between current sources and voltage sources. Grid following inverters cannot form a voltage source and so require connection to a relatively strong voltage source. The best performance can be achieved when connecting to an ideal voltage source which naturally comes with excellent current control capability that does not require connection to a current source.
In general, the opposite applies to grid forming devices since they can form their own voltage source, but don’t have the same tight current control exercised in the grid following inverters, or, if they do, this capability can adversely impact other aspects of system stability. This means that they are most susceptible to unstable behaviours under strong grid voltage with a low impedance. Such instabilities are not experienced in synchronous machines under high system strength conditions as the concept of loose/tight converter current control does not apply.
System instability with grid forming inverters can still be experienced under low system strength conditions. However, most of these instabilities are attributed to exceeding the maximum power transfer capability from a steady-state standpoint rather than inverter control susceptibilities.
Since grid forming inverters are primarily intended for low system strength conditions, the impact of this limitation may not be immediately apparent, but if more grid forming inverters begin to connect in currently strong parts of the network, there is a possibility that they will create high system strength conditions which could then adversely impact the stability of these grid forming inverters. That’s quite the catch-22 to manage: install lots of grid forming inverters to stabilise the grid and you will destabilise the grid as a result, due to their internal voltage regulation conflicting with the stiff grid voltage.
Another known limitation of inverters is the possibility of very little or no injection under very low or very high (close to 90% residual voltages). Fault current contribution can vary depending on steady-state operating conditions, making it more difficult to correctly account for a grid forming inverter’s fault current contribution for all conceivable operating conditions in fault current calculation engines.
The report sets out five risks associated with the deployment of grid forming power electronics:
- Correct operation of protection systems
- Magnitude and duration of fault current
- Voltage change due to reactive plant switching
- Considerations for distributed IBR stability
- Lack of accurate grid forming models in commercial simulation tools
Of these Transgrid says that the correct operation of protection systems “remains a significant unknown” ie whether protection systems designed for use in grids dominated by synchronous machines will function correctly in an IBR-dominated grid. Protection relays expect large, high fault currents with characteristic waveforms and phase shifts that depend on the type and location of the fault. Inverter-based resources, by contrast, limit their current (to protect semiconductors) and inject it with controlled phase angles. The result is very low, highly variable, and unpredictable fault currents. Whether relays and protection systems will still detect and clear faults quickly and selectively when fault currents are small, shaped differently, or delayed is currently unknown.
This is not just about the magnitude of fault current (enough amps to trip a relay), the bigger issue is that some modern protection systems make decisions based on impedance or sequence component analysis, and that depends on system physics that change dramatically when IBRs dominate.
When a fault occurs such as a short-circuit between phases or between a phase and ground, the three phase voltages and currents (A, B and C) become unbalanced. In normal operation, the system is perfectly balanced: each phase is 120° apart and the magnitudes are equal, so if you draw them as vectors they form a perfect equilateral triangle.
When a fault happens eg one phase touches the ground, the symmetry breaks: the three phasors are no longer equally spaced or equal in size. This makes direct analysis awkward, because protection relays need to know what kind of fault it is, where it is, and how severe it is. So engineers decompose the messy, unbalanced system into three simpler, balanced “sequence” systems:
The positive-sequence (also called “forward” sequence): three phasors of equal magnitude, spaced 120° apart, in the normal A–B–C rotation. This represents the healthy system — how the grid behaves under normal conditions. Synchronous machines naturally produce positive-sequence current
The negative-sequence: also three phasors of equal magnitude, 120° apart, but rotating the opposite way (A–C–B). This represents imbalance with faults between phases or unbalanced loads. In a healthy system, it’s nearly zero, but during faults, synchronous machines produce a predictable amount of negative-sequence current, which relays can detect.
The zero-sequence: three phasors in phase with each other (no angular separation). This appears during earth faults (when one or more phases are connected to ground) and is used to detect ground faults through the return path via neutral or earth.
Together: Total current = Positive + Negative + Zero-sequence components
Protection relays measure phase currents and voltages, transform them into sequence components, and then act based on characteristic patterns. A positive-sequence indicates normal power flow, a negative-sequence indicates phase imbalance or fault between phases, and a zero-sequence indicates earth fault. For example, a line-to-line fault gives a strong negative-sequence signal, a single line-to-ground fault gives strong zero-sequence, and a three-phase fault gives mostly positive-sequence. By monitoring these ratios, relays can classify the fault type and decide whether to trip.
Synchronous machines respond inherently to sequence components since their electromagnetic fields generate predictable currents in all three sequences. This is why protection schemes based on sequence analysis work well in conventional grids. However, inverters don’t respond inherently – their control software decides how much positive, negative, and zero-sequence current to inject. Many limit or even suppress negative-sequence current to protect their semiconductors. This means the relays may see nothing unusual during an unbalanced fault and fail to trip.
Transgrid’s report warns that “in an IBR-dominated system, the amount of positive- and negative-sequence current injected depends on control-system response and current availability… this cannot be assessed with sufficient accuracy using commercial fault current calculations or protection coordination tools”.
And because inverter behaviour depends on grid impedance, fault current levels can vary massively between, say, a light-load summer day (weak system) and a heavy winter evening (strong system). In a weak grid, an inverter’s contribution to fault current may collapse almost instantly; in a strong grid, it may contribute a bit more but still below what protection expects. So protection settings that work one day might fail the next.
This isn’t just a problem for grid stability, it’s also a safety issue: an inability to detect a fault may mean that faulted elements will remain energised when they should not.
Inverters, including grid forming devices, create issues with harmonics
Unlike synchronous machines, inverters can inject harmonics, flicker and voltage unbalance into the power system. Where synchronous generators act like a sink for some harmonics, including significant ones such as the second harmonic, grid-forming inverters typically emit rather than dampen harmonics. Although they can be programmed to cancel out certain harmonics, the impact on the total current available, and the trade-off with other important capabilities, each requiring part of the same limited current capacity, must be carefully considered.
For the uninitiated, a harmonic is a frequency component that is an integer multiple of the grid’s fundamental 50 Hz (or 60 Hz) frequency. The first harmonic is the fundamental, 50 Hz, the second harmonic is 100 Hz, the third is 150 Hz, and so on. These harmonics appear whenever the current or voltage waveform deviates from a perfect sine wave, usually because of non-linear equipment such as power electronics, rectifiers or inverters.
The three main concerns about harmonics on the electrical system are resonance, heating and electromagnetic compatibility with equipment.
Resonance can lead to catastrophic failure of equipment or systems. It is like putting a microphone in front of a speaker and creating a feedback loop where the signal continues to amplify until something breaks. On the electrical system, this happens when the capacitance and inductance align with a harmonic frequency to create a similar feedback loop. This is why installation of capacitor banks, filters and reactive compensation systems must always be evaluated for harmonic interactions.
Heating of equipment due to harmonics is a well-documented effect. Harmonics increase both copper and iron losses. In conductors and transformer windings, higher-order harmonics increase heating, as each frequency component adds its own resistive losses. In magnetic cores, the oscillating fields from harmonics induce eddy currents, producing additional core losses and localised hot spots that accelerate insulation ageing. Eddy-current losses grow with the square of frequency, so even modest distortion at higher harmonics can produce disproportionate temperature rises.
The phenomenon of skin effect compounds this problem. As frequency rises, current flows increasingly near the conductor surface, effectively shrinking the usable cross-sectional area and raising apparent resistance. This makes higher-frequency harmonics more dissipative and more likely to cause overheating.
The increased impedance becomes particularly critical when dealing with third-harmonic return flow in the neutral conductor. Third harmonics do not cancel out but instead sum together in the neutral wire. Combined with the higher resistance from skin effect, this can cause an increase in third-harmonic stray voltage along the neutral path, leading to excessive heating, nuisance trips and even melted neutrals in poorly designed systems. Over time, these effects translate into reduced transformer efficiency, cable degradation and shortened asset life.
The least understood effect of harmonics is faulty operation of equipment caused by electromagnetic compatibility issues. Numerous cases have been documented where harmonics have caused process-control malfunctions in industrial systems, flickering LED lights, false readings in metering and SCADA systems, and audible interference in communications equipment. These problems are often difficult to diagnose and require extensive data collection, harmonic analysis and investigation of system interactions.
The second harmonic, 100 Hz, is especially problematic because it distorts the voltage waveform in a way that can interfere with transformer magnetisation, metering accuracy and relay operation. It can bias converter controls, cause oscillations in dc links, and often appears during unbalanced faults or in systems where converters rectify and invert ac multiple times. In grids dominated by rotating machines, second-harmonic currents are usually very small because the machines’ physical and electromagnetic characteristics absorb or suppress them.
Second harmonics are particularly associated with solar PV. When grid voltage drops, the output current of PV inverters contains a high proportion of second harmonics that cannot be ignored. Studies have shown that PV inverters inject harmonic currents into the grid even under normal conditions, and these harmonic currents worsen under disturbance. Rather than producing a clean sinusoidal current, the inverter output includes significant second, third, fifth and higher harmonics, distorting grid voltage and confusing protection systems.
During a low-voltage fault, the inverter must stay connected and support the grid instead of tripping off. To do this, it switches control modes from power-maximising to voltage-support mode. Inverters have multiple control loops, inner current control and outer voltage or power control. During a fault, the ride-through mechanism activates, and the inner control loop suddenly operates in two regimes, a linear region where voltage and current behave predictably, and a non-linear region where voltage dips and the control system saturates because the inverter’s current limit is reached. When the inverter transitions between these regions, its transient behaviour becomes non-linear and produces current spikes as large as a full ac cycle’s peak current on the output side.
Instead of a smooth, sinusoidal fault current, there is a burst of distorted current dominated by harmonics and transients. At the start of a fault, the dc component of the inverter’s fault current can reach around 34% of the fundamental amplitude, while the second harmonic can reach roughly 39%. These components decay exponentially over time, but during the first few cycles they dominate the waveform. The result is a current that is not a clean sine wave but a mix of the 50 Hz fundamental, the 100 Hz second harmonic, the dc offset and smaller higher-order harmonics. (The dc offset is the amount by which the current waveform deviates from being centred around zero, creating a non-zero average over a full cycle.)
This behaviour is completely different from that of a synchronous machine, where fault current is large and mostly fundamental, with a predictable dc offset that decays with rotor time constant. Harmonics are minimal, and the waveform is well understood and easy for relays to interpret.
For inverters, the fault current is smaller, highly distorted and rich in transient harmonics, especially second-harmonic content. This confuses conventional over-current and distance protection relays, which assume that the waveform represents a simple 50 Hz sinusoid plus a dc offset. When they encounter heavily distorted currents, they may miscalculate impedance or direction and either trip unnecessarily or fail to trip when they should.
The shift from synchronous to inverter-dominated grids introduces a new class of harmonic and thermal risks that were largely absent in traditional systems. Synchronous machines not only supplied inertia and dampened voltage disturbances, but also acted as natural filters and harmonic sinks, quietly absorbing distortion and maintaining waveform integrity.
Inverters, by contrast, can both generate and amplify harmonics, especially during transient events and low-voltage ride-through conditions. The resulting distortion increases thermal stress on cables, transformers and switchgear, shortens insulation life, and undermines the reliable operation of protection and control systems.
As inverter penetration rises, managing harmonics will require system-level coordination rather than just local filtering. Operators will need better harmonic monitoring, adaptive filtering strategies and realistic testing of inverter behaviour under faulted conditions. Without that, the grid of the future may be clean in carbon terms, but far dirtier in electrical terms: a noisier, hotter and less predictable system than the one it replaced.
[Aside on harmonic filtering:
Readers more familiar with optical systems will note that optical filters absorb or reflect unwanted wavelengths of light. A passive electrical harmonic filter, made of inductors, capacitors and sometimes resistors, is tuned so that at a particular harmonic frequency its impedance is very low. This means the harmonic current is diverted down that path instead of flowing through the rest of the network.
The filter is connected in parallel with the main circuit. At the fundamental frequency (50 Hz), its impedance is high, so it barely conducts. But at, say, the 5th harmonic (250 Hz), its impedance drops sharply, and the harmonic current is diverted into the filter. Once inside, part of the energy may be dissipated as heat in resistors, but most simply circulates within the L–C branch, isolated from the grid.
Active filters, however, behave more like destructive interference in optics. They measure the harmonic current and inject an equal and opposite waveform, cancelling the distortion at the grid connection point. The cancelling energy has to be created by the filter’s power electronics.
In contrast, optical filters passively absorb certain wavelengths because of their inherent physical properties, while active electrical filters determine the nature of the harmonic and synthesise an equal and opposite waveform to cancel it out. Both exploit fundamental properties of electromagnetic waves, but in optics this behaviour arises passively from the material, whereas in electrical systems it is actively generated through control and computation.]
The economic frameworks for grid forming inverters also do not exist
The report also sets out some of the economic challenges of using grid forming inverters to provide system strength. In the same way that using IBRs for inertia requires haircutting the real power output of the generating source or battery in order to provide headroom for frequency response, there is a similar challenge with the use of GFPE for grid stability: they can provide many different capabilities, however, each requires the use of some of the limited current available. Not all grid forming capabilities can be provided at the same time because most of these capabilities will rely on a portion of the total current, active or reactive.
Whilst grid forming inverters provided by some OEMs can provide fault current contribution of up to 2 pu and even higher, this comes at a cost, and the most widely used devices come with a capability which is at or just slightly above the inverter rated current. A prudent planning principle is therefore to assume that grid forming inverters can only contribute up to their rated current. In fact, operation below the rated current will assist in avoiding adverse effects on other aspects of GFPE performance, even if the inverter has a higher technical capability.
Switching between the capabilities is possible, but this means more complex design, modelling and compliance studies. A conflict in priorities due to the limited current is most likely to arise is the provision of fast frequency response /inertia when there is a combined voltage and frequency disturbance. Since batteries are the most widely used grid forming technology currently in use, this gives rise to two further limitations: limited time that sufficient charge can be maintained while retaining an ability to be ready to respond to the next event, and a life cycle of around 1000 deep charges and discharges (this may vary from c500 to c3000). Providing a large fault current will mean larger changes in both active and reactive current even under conditions such changes are not desired from a lifetime perspective, for example a rapid change in active current from zero to full discharge to provide a large current.
Premature reliance on GFPE will carry significant risks
Transgrid concludes that both grid forming batteries and STATCOMs should not form part of minimum system strength capabilities until these issues can be resolved. It says that if TSOs start relying on grid forming batteries to provide the baseline level of inertia or fault current needed to keep the grid stable then conventional synchronous generators may not be dispatched as often. When a thermal unit isn’t dispatched, it shuts down or stays offline, and it can then take several hours to bring it back from a cold start. This means it is unable to respond to a later fault as both its energy output and stability services (inertia, reactive power, short-circuit strength) will be delayed until it warms up.
This can’t be resolved by putting a constraint on the system (for example, limiting inverter output or mandating certain dispatch patterns) because dynamic instabilities happen too quickly for market-based constraints or dispatch instructions to be effective. In other words, the system operator can’t rely on dispatch tools or constraint equations to stabilise the grid after an instability starts.
In other words, before relying on grid forming batteries for core stability, TSOs need to have a very high confidence level that they will behave correctly during disturbances because it won’t be possible to bring synchronous support online fast enough if they don’t.
Transgrid also points out that it took two years for the full deployment of the four synchronous condensers in South Australia as AEMO and ElectraNet worked to gain the necessary confidence in their behaviour and interactions with other grid equipment, and this was with a much simpler synchronous condenser technology. The deployment of grid forming inverters can be expected to take much longer.
Scotland may well be next
Britain should pay close attention. Scotland has only two conventional generators left, one of which, Torness, will close in March 2030. Voltage disturbances and frequency oscillations are already being observed, and NESO is struggling to dampen them. Various corrective projects are under way (the Pathfinders), but given the experiences in Australia, they may not be fully operational before Torness closes, leaving the Scottish grid weak and exposed.
A recent report by the GB system operator, NESO, claimed that what happened in Spain could not occur here. It can. In its report, NESO says that sub-synchronous oscillations are normal, and well managed. Several such oscillations were observed in Scotland in 2023 the range 5-9 Hz, mainly at about 8 Hz (approximately 1/6th of 50 Hz nominal frequency). These events caused disturbances on the power system which included the tripping of generation, tripping of an interconnector and HVDC link, and in one case a transmission circuit trip. There were other documented sub-synchronous oscillations including 8 Hz in North Scotland in 2021, and recurring events centred in Scotland in Summer 2023, with further occurrences in January and May 2024.
However, sub-synchronous oscillations are not the only types power grid instability observed in Scotland. Over the past few years Scotland has seen multiple examples of other types of grid instability including persistent harmonic distortion hotspots (eg 5th-order at Crossaig) now being treated with harmonic filters, and harmonic mitigation around Mark Hill, as well as a system-strength deficit (low inertia / low short-circuit level) that led NESO to procure stability services focused on Scotland. These sit alongside high-voltage management challenges under light-load /high-wind conditions. NESO and the other transmission operators have responded with filters, STATCOMs, stability procurement, wider phasor measurement unit rollout and development of electromagnetic transient models to better characterise inverter-grid interactions.
Harmonics issues have been identified at Crossaig in Argyll, Mark Hill in Ayrshire and the Caithness–Moray HVDC zone.
At Crossaig, Scottish & Southern Electricity Networks has identified “unusually high levels of harmonic distortion at the 5th harmonic order” which breach the planning limit and fluctuate within the band between the planning and compatibility levels. The cause was identified as an amplification of the harmonic within the Scottish Power licence area, caused by the Kintyre – Hunterston 220 kV subsea cables (but in that area it is not above planning levels – the physical characteristic of the cables results in amplification across and into the SSEN licence area).
At Mark Hill, a lot of large windfarms are being connected to what Scottish Power describes as “relatively weak 132kV networks” which are also characterised by the increased use of long cable circuits. The combination of a relatively high source impedance with higher cable capacitance leads to resonance at lower frequencies in the network, typically below the 20th harmonic (1 kHz), creating a need to ensure the grid complies with harmonic level standards.
The Caithness-Moray HVDC line was constructed to allow increased flow of electricity from renewables in the north of Scotland, such as the Beatrice and Dorenell wind farms. This project had specific targets for Total Harmonic Distortion levels, which are monitored to ensure the system functions correctly. The system has two operating modes: “normal” and “extremely weak”.
“In GB we recognise the trend toward increasing voltage fluctuations and it is an issue that NESO is taking steps to manage. We continue to develop our modelling capability to enhance our understanding of voltage challenges and continue to push forward,”
– Future-Proofing GB’s Power System: Reflections on the April 2025 Iberian Event, NESO
NESO’s report has a reassuring tone, but if read in detail, an awful lot of aspirations rather than finished work are described.
The GB grid is demonstrating signs of weakness
I had heard from engineers working in the power sector that some unusual frequency behaviours had been observed on late May, and decided to investigate. I fell at the first hurdle, because of the poor quality of data reported by NESO to Elexon. I complained, and Elexon requested re-submission of the data numerous times, but at the time of writing, almost five months later, the junk data are still present.
Here is one example, but there were various hours during the period of interest showing such zero values. I then looked at the 1-second data on the NESO Portal. These files are very difficult to handle since they are larger than the number of rows in Excel (thank goodness for ChatGPT which can take the full csv and break it into daily files). Here there are no zero values in May 2025 (although when I did earlier analysis of times frequency had gone outside the operational and statutory limits I did find zeros that had to be cleaned).
The data indicate that between 24 and 27 May 2025, the GB grid exhibited mild but distinct frequency irregularities, consistent with low-damping oscillations and weak-grid dynamic interactions rather than measurement artefacts or large system imbalances. These anomalies were short-duration – typically 20 to 60 seconds – of elevated frequency variability (rolling 20-s σ > 0.02 Hz) and rapid reversals (|Δf| > 0.03 Hz s⁻¹). These clusters were most frequent during low-demand, high-renewables periods: overnight and early-afternoon hours over the late-May bank-holiday weekend.
The deviations do not resemble a simple generation-loss or load-shedding event. Instead, the frequency oscillates around the nominal 50 Hz baseline with alternating positive and negative excursions of a few tens of millihertz, what can be described as a lightly damped “ringing” motion. This is typical of a poorly damped frequency–voltage interaction rather than a step disturbance, and is characterised by:
- Short-term resonance behaviour with repeated swings every 5 to 15 seconds, decaying gradually rather than instantly, implying that natural damping in the system is weak
- Asynchronous bursts where the pattern doesn’t propagate uniformly, suggesting it’s not a single-source fault but multiple IBRs (inverter-based resources) responding non-coherently to small voltage/frequency perturbations
- Correlation with low inertia periods which coincide with times when there were few synchronous machines meaning system inertia and reactive reserve were likely both low
While only operational logs could confirm the details of why these behaviours were observed, several mechanisms could explain the frequency data:
- Low inertia / weak-grid dynamic coupling: with reduced synchronous mass, even small active-power fluctuations from wind and solar translate into noticeable frequency movement. In low-load, high-IBR periods, inverter controls can interact through the grid impedance to produce electromechanical oscillations typically around 0.2 – 1.0 Hz. These are in-band for system frequency measurements, and therefore visible in the 1-second NESO data
- Grid-following inverter control interactions: many IBRs use phase-locked-loop (“PLLs”) controllers to synchronise to the grid waveform. When the grid becomes electrically “soft,” PLLs can oscillate or chase each other, producing alternating over- and under-frequency responses. These control-loop fights create the quasi-sinusoidal patterns observed here
- Automatic Generation Control (“AGC”) chasing noise: during holidays, AGC algorithms often run against sparse or fluctuating signals. When system inertia is low, even modest AGC corrections overshoot, amplifying the apparent frequency noise
- Local voltage oscillations coupling into frequency: voltage disturbances, possibly triggered by high PV reactive output or tap-changer actions, can indirectly affect frequency through converter control cross-coupling. This is consistent with reports from the control room that “voltage and frequency were behaving oddly” over that weekend.
The May-bank-holiday oscillations point to marginal stability under certain system conditions, rather than an immediate reliability failure. They show that the GB grid can approach low-damping behaviour when synchronous commitment is minimal and renewables dominate, that control-loop interactions among IBRs are beginning to have visible, system-level effects in the 0.1–1 Hz range, and that the standard tools (primary/secondary frequency control) remain effective but may be masking emerging weaknesses in reactive power management and grid-forming capability. If this pattern persists or intensifies, it could foreshadow the kind of frequency-voltage coupling that caused the Iberian collapse in April, albeit on a smaller scale.
To the trained eye, this pattern says that the grid didn’t fail, but it did twitch. These micro-oscillations are the electrical equivalent of small tremors before a major earthquake. They tell us the damping provided by synchronous machines is now insufficient in certain operating states, and that inverters are beginning to interact through the network rather than with it. They were not classical sub-synchronous interactions but lightly damped electromechanical or inverter-control oscillations consistent with low-inertia conditions.
These effects are not yet large enough to be dangerous, but it’s a warning shot. Without tighter grid-code enforcement, better tuning of inverter controls, and more distributed inertia or synchronous support (such as synchronous condensers), these oscillations could one day align and amplify, creating the same positive feedback that brought down the Iberian grid in April.
In the operability analysis supporting its Clean Power 2030 advice, NESO confidently asserts that “installation of new grid-forming technology can reduce additional stability requirements and, alongside obligations in the Grid Code for users, Transmission Owner (TO) built High Voltage Direct Current (HVDC) links with grid-forming characteristics can also play an important role in supporting system stability. To address these network service requirements, we require new assets, new stability markets and policy changes by 2030. These additional network services requirements can be secured and delivered for 2030 by following the existing NESO procurement processes and working with Great Britain’s TOs”. Transgrid may disagree.
Grid Code compliance is also an issue in GB
NESO also says in relation to the Iberian blackout that “key factors cited to date were the lack of delivery of obligated reactive compensation from thermal generation and the lack of dynamic voltage regulation from renewable generation (which was the predominant source of power). This is not the case in GB. We require reactive capability and voltage control from all generation plants including non-synchronous renewables. Further, we also have compliance testing processes to test this capability ahead of connection to the system and then periodically once connected, NESO has the processes and ability to limit operation of non-compliant generators.”
Again, this sounds better on paper than in reality. Grid Code compliance IS tested when new generators connect to the power grid, and when they return after an extended outage. But the ongoing compliance has historically been rare, and compliance failures typically only detected after the equipment fails during a fault. This was the case during the 2019 blackout when both Hornsea One and Little Barford failed to meet fault ride through requirements and were subsequently fined.
NESO says that since this incident it has implemented 5-yearly compliance testing – indeed, improved compliance monitoring was one of Ofgem’s requirements after the blackout. However industry sources tell me that grid connected batteries have poor code compliance which is not being addressed. A Scottish Power report from January this year mentions ongoing Grid Code compliance challenges with batteries.
I can find no evidence that the new requirement for 5-yearly Grid Code compliance testing is actually happening. With the help of ChatGPT and Google, I searched for any evidence of compliance audit reports by NESO or Ofgem detailing the number of sites tested and the number found to be compliant or non-compliant, any evidence of the types of compliance problems identified, or any fines issued for non-compliance. I find it unlikely that every single site has been fully compliant, so the lack of information is surprising.
Typically, fines are public, because the authorities want to underscore the importance of following the rules – publicising fines, even if the parties are not identified, allows the market to understand the issue is being taken seriously. So I have sent FOI requests to both NESO and Ofgem to see if I can find out if this testing is actually happening, and if not, when it is expected to start. I will report back….
Complacency, racing to meet net zero targets and erosion of engineering capabilities amplify risks
The Iberian crisis is a warning of what happens when the energy transition progresses faster than the engineering needed to support it. As synchronous generation closes, grids grow weaker, and system operators must learn to manage not just frequency but voltage. At the same time, expertise within these organisations has been hollowed out. Where once they were run by engineers who owned, maintained and operated the grid as a single system, they are now managed by economists and programme directors. Although REE still owns its transmission network, NESO does not and has lost the intimate connection it once had to the physical equipment. Teams operating in silos exacerbates the erosion of knowledge.
The Iberian blackout has exposed gaps in REE’s understanding of the fundamental behaviour of the grid it is operating. Australia, which has been grappling with the issues created by inverter-based resources for longer, has demonstrated that much more work is needed to fully understand the way in which inverters interact with essential grid equipment such as protection relays, and that electronic solutions such as grid forming power electronics have hard limits to their capabilities linked to the physical limits of the semiconductors, and may under some fault conditions, revert to grid following. The need to reserve limited current for multiple grid support functions is one area for further study, and the economics of these functions, which are delivered for free by synchronous generators, have yet to be determined.
Fundamentally, system operators lose sight of the physics of the grid, as REE appears to have done, no amount of regulation or software will save them. NESO’s “nothing to see here” report does little to inspire confidence it has better insight. Rather than superficial reassurances, system operators need to follow the example of Transgrid and set out the risks associated with trying to stuff direct current devices onto alternating current grids without sufficient understanding of their limitations and how they interact with existing grid equipment. Policymakers need honest assessments of these risks, and not be fed a false sense of security around deliverability of solutions.
Leave A Comment