Thomas's projects

Apagon de 28 de April - 28th April Spain Blackout

The issue of the blackout was a combination of an Iberian grid oscillation and poor control of the phase between the current and the voltage (reactive power) on the grid. Several events accumulated. I propose to describe in this post the chain of events and the physics behind it.

All information comes from the following report: Version no confidencial del informe del comite para el analisis de las circunstancias que concurrieron en la crisis de electricidad des 28 abril de 2025

Technical notions

Reactive power and voltage in a grid

The phase between the current and the voltage describes the grid's capacity to transmit power.

In the alternating current domain, electrical power (called apparent power) is divided into two elements: active power (also called true power) and reactive power. If the current and voltage are in phase (0° of dephasing), the power is composed solely of active power and is transmitted correctly. Conversely, when the current is completely out of phase (90° of dephasing or -90) with the voltage, the reactive power is at its maximum and the active power is zero. Due to the properties of the grid, power cannot flow in this scenario. The ratio between apparent power and active power is called the power factor. It ranges from 0 to 1. A value close to 1 indicates that the power is composed mainly of active power.

Example of dephasage of 0°

Example of a dephasage of 90°

Reactive power is generated by the reactive component on a network. There are three main components:

From electronics-tutorals

From electronics-tutorals

From electronics-tutorals

Since grids are combinations of these elements, they generate reactive power. The wire itself acts as the resistance, the line forms a small capacitance with the ground, and the line also acts as a small inductance. Therefore, the grid, depending on its geometry and the amount of current flowing through it, generates reactive power. If a large amount of current flows through a line, the inductive part of the line will cause a delay for the current (the voltage will be early). If a high voltage is applied to the line, the votlage will be delayed.

A representation of a power line as a succession of resistors, capacitance and inductance. From Wikipedia

Reactive power is therefore useful in a grid to maintain the voltage at specific levels. If too much reactive power accumulates on the grid, the voltage will increase. If too little reactive power is present, the voltage will decrease. The reactive elements (capacitors and inductors) will not be adequately charged and will attempt to compensate for the missing reactive power, resulting in lower voltage. This, in turn, impacts the transmission of active power.

When transmission system operators (TSOs) design and operate a grid, they must carefully balance these effects. The grid will never be perfectly balanced. However, TSOs can rely on producers and consumers to help balance the grid. Indeed, many types of equipment can provide or absorb reactive power. The only requirement for this is to introduce a phase shift between the current and voltage when consuming or providing power to the grid.

Frequency oscillation

The European grid is not a monolithic block. All elements must provide and receive a 50 Hz power wave. However, due to various effects, the frequency is never exactly 50 Hz. One of the causes is grid oscillation.

This phenomenon appears as cyclic changes in frequency over time and is caused by the 'natural modes' of the system. When a string on a guitar is excited, the string (and the instrument) will always vibrate in the same way. This vibration is called a standing wave and can be decomposed into a sum of modes. The fundamental mode of a guitar string can be described as follows: the extremities remain fixed, and the middle of the string vibrates significantly. The length of the string defines the frequency of the sound. By changing its length, the frequencies of the modes are altered. A musical note is defined by the combination of these modes at specific frequencies. If you change the geometry of the string and the instrument (for example, by replacing the guitar with a violin), the modes will change, and the note will sound different even if the frequency remains identical.

Different vibration mode of a rope. FromWikipedia

Several vibration mode of a gitare. From 10.4236/oja.2020.103003

Several mode from a violin. From 10.1007/s00107-020-01538-5

The same principle applies to the European grid. The entire grid vibrates at 50 Hz, which is its fundamental mode. However, other modes can also emerge. Several well-known and extensively studied modes include:

Local vibration exist too. We can name:

Apagon de 28 de April 2025

Now that we have covered these technical aspects, we can describe the issues with the Spanish grid.

Phase 0 - Initial state

The Spanish grid controls voltage (and thus reactive power) through several methods:

Additional methods include:

These three elements exist in the Spanish grid but were not involved in the 28th April blackout.

You can find more information on the wikipedia page Voltage Control and reactive power management.

On April 28th, the weather in Spain was favorable, being sunny and neither too cold nor too warm. The load on the Spanish grid on that day was not high. Solar production was expected to be significant, and market prices were anticipated to be low.

Thermal power plants were available in sufficient numbers according to the RED (Spain's TSO) estimation. These plants were instructed to follow the voltage evolution and adjust the reactive power accordingly. They were dispatched throughout the country.

One power plant in south west was not available during the 28th reducing the total number to 9.

Until 10:00, no significant issues were noticed on the grid, which was behaving correctly. After 06:00, some variations on the interconnection with France occurred. The export changed from 2590 MW to 1600 MW due to higher-than-expected demand. This change caused some frequency variations, but nothing dramatic, as the grid had to adjust to the changing conditions. It is important to note that not all operators react immediately or at the same time. The voltage and frequency variations remained within the grid's operational limits.

Phase 1

Several little European oscillation (0,2Hz)

Several frequency oscillations were noticed, with the grid's frequency changing periodically over time.

To dampen these oscillations, the operator chose to connect more lines to the network. By doing so, the vibration mode of the grid changed. A vibration mode depends on the geometry of the grid and the equipment connected. By connecting new lines, the mode was altered, and its energy was dissipated. This strategy is common and worked effectively in this instance.

The 0.2 Hz oscillation is a known phenomenon and corresponds to certain vibration modes of the European grid.

First large oscillation (0,6Hz and 0,2 Hz) 12:03

At 12:03, a new oscillation appeared. This time, it was a 0.6 Hz oscillation with an amplitude of 70 mHz.

This new oscillation was observed from Portugal to Germany. It caused voltage variations in the Spanish grid of up to 32.7 kV over 400 kV. This oscillation was also monitored on the France-Spain interconnection. At this moment, the 0.2 Hz oscillation reappeared, but the grid's damping effect was less effective.

The fact that the 0.6 Hz oscillation propagated as far as Germany is a strong indication that it is a mode of the grid. However, the oscillation was most pronounced in the Iberian Peninsula, specifically between Porto and Malaga. All the photovoltaic power plants had a steady production output, except for one. Its production oscillated at the same frequency with an amplitude of 70% of its initial production. This is an abnormal behavior, as photovoltaic technology in Spain is designed to produce power with a fixed power factor. It cannot be ruled out that this plant is the source of the oscillation. By entering into resonance with the grid mode, the oscillation was not absorbed and could propagate, leading to a more significant impact.

At this moment, since the behavior of the photovoltaic (PV) plant was not known to RED, the Transmission System Operator (TSO) applied a protocol previously defined with France. The interconnection was set to export 1500 MW to France. Additionally, the interconnection mode was changed from an AC mode (called Pmode 3) to a DC mode (called Pmode 1). This change was made possible through the use of inverters. This interconnection is inherently a DC interconnection. However, thanks to this mode, it can simulate an AC line (bidirectional for power and oscillations) or a DC line. Power is collected on one side and injected on the other. Oscillations cannot travel through it. Furthermore, the power collected is defined by a command and is not automatic, as it would be for an AC line.

There are several interconnections between France and Spain. The interconnection affected by this change is the one in the east, highlighted in pink. The other interconnections are AC interconnections.

From ENTSOE Grid Map

The RED TSO also decided to connect additional lines to change the grid topology and attenuate the vibrations.

The oscillation is successfully attenuated at 12:07.

A specific electrical line (400 kV Cedillo-Falagueira) is known to create unwanted oscillations. To reduce the power transiting through this line, the RED TSO requested that the REN (Portugal's TSO) decrease exports to Portugal from 2500 MW to 2000 MW. The REN accepted and planned the reduction for 13:00.

Second large oscillation (0,6Hz and 0,2 Hz) 12:19

At 12:16, the 0.6 Hz oscillation reappeared, and the same PV plant exhibited the same reactive power oscillation.

At 12:19, a new 0.2 Hz grid oscillation appeared. This time, the magnitude of the oscillation reached up to 200 mHz locally and lasted until 12:21:30. This caused voltage oscillations of up to 23 kV (over 400 kV) on the grid. In response, RED decided to further increase the meshing of the grid by connecting additional lines. They also chose to execute the reduction of the export to Portugal earlier, at 12:30.

At 12:05 and 12:20, frequency variations induced voltage fluctuations on the RED 400 kV grid. However, these fluctuations remained within operational limits.

At the end of this phase, 10 lines were added to the RED grid.

Exports toward France and Portugal were reduced. Spain exports 1500 MW toward France with a DC line and 2500 MW toward Portugal. It is planned to reduce slowly the export toward France to 1000 MW and more quickly the export toward Portugal to 2000 MW.

The addition of the lines in a context of low demand generates more reactive power on the grid. Indeed, the capacitors and inductances are generating opposite reactive power. An equilibrium can be found by balancing these components. However, the inductance generates reactive power if current passes through. For capacitors, only the voltage is necessary. Hence, by adding more lines, the current on each line is reduced, and the reactive power was less balanced, making it more sensitive to reactive power variations.

Phase 2

When the export change are applied surtension are observed all over Spain.

First deconnection at 12:32

Export toward France started to decrease at 12:00. At 12:27, the interconnection line with Portugal started to reduce its exchange too. It happened 3 minutes earlier due to short notice and limited time to synchronize all the productions. The production was reduced to follow this change. The voltage increased at the same time (less reactive power consumed).

At 12:30, with the general overvoltage event, the export increased again before falling quickly at 12:32 on all interconnections (except Portugal).

Interconnections are sensitive to active power. They show the disconnection of generators and loss of power. The analysis links this loss to 525 MW of small producers. Immediately after, at 12:32:57.140, a generator that was consuming 165 MVAr of reactive power was disconnected. This plant was connected to the secondary grid (220 kV). It was due to overvoltage on this part of the grid. The reason is not yet known, but it could be due to a transformer's tap changer reacting too slowly. By staying on the inappropriate tap longer than expected, overvoltage was delivered in the secondary grid. The frequency dropped but was restored 3 seconds later. The export to France fell to 0 MW. The synchronous production (thermal plants) was not affected by the event.

Since less reactive power was consumed and evacuated toward the France interconnection, it started to accumulate in the grid, especially close to Grenada (South), and voltage rose.

In this graph 39, you can see the phase shift inside the grid increasing. Ideally it should be zero. However due the reduction of the flow between the north and the south the stations are shifting. Note: The hours are not correct. It shows the phase shift at 12:32.

Second deconnection

At 12:33:16.460, a second disconnection occurred close to Badajoz (southwest). 730 MW were lost, and the frequency dropped by 55 mHz. The export with France became an import of 895 MW. Several small disconnections were triggered by this loss in the following seconds and by the rise in voltage.

Third disconnection

At 12:33:17.780, a third large disconnection (550 MW) occurred close to Sevilla (southwest). Again, in the following seconds, several disconnections occurred due to the overvoltage. The frequency (–75 mHz) was again affected, and the import from France rose to 2405 MW.

In this area, the voltage of the secondary grid increased from 220 kV to 240 kV. The frequency was no longer close to 50 Hz.General view of the behavior the France-Spain interconnections. Notice that the DC interconnection stay close to 1000MW during the whole Phase 3.

In a few seconds, the wave will propagate throughout all of Spain, and the frequency will continue to drop. This table shows how quickly the disconnections impacted the grid.

The disconnections due to underfrequency occurred at the very end, 6 seconds before the blackout.

Phase 3

The third phase describes the blackout itself. Two events occurred at the same time: an overvoltage wave appeared on the primary grid with very high voltage levels, and a drop in frequency. This phase lasted 5 seconds.

The overvoltage on the secondary grid (220 kV) continued to propagate. 247.6 kV was measured locally, triggering new disconnections. These disconnections decreased the consumption of reactive power, which was then sent back toward the primary grid (400 kV). The wave began to spread through the network, with local voltage measurements reaching up to 443.8 kV.

These values are above the security threshold, and disconnections are expected. The local stations, which follow the voltage by absorbing or generating reactive power, were disconnected to protect themselves. Once again, the reactive power was not compensated, and the wave continued to propagate. The entire transmission grid was affected (400 kV saw 440 kV, 220 kV saw 240 kV, 130 kV saw 146 kV).

At the same time (12:33:19.320), the frequency continued to decrease.

At 12:33:19.620, the interconnection with France reached its maximum capacity. Synchronization with the rest of Europe was lost. This loss of synchronization caused large fluctuations in the transmitted power, exacerbating the event.

General view of the behavior the France-Spain interconnections. Notice that the DC interconnection stay close to 1000MW during the whole Phase 3.

In a few seconds, the wave will propagate throughout all of Spain, and the frequency will continue to drop. This table shows how quickly the disconnections impacted the grid.

The disconnections due to underfrequency occurred at the very end, 6 seconds before the blackout.

Conclusion

The cause is multifactorial. It started with the 0.2 Hz European grid oscillation combined with a 0.6 Hz internal oscillation from a PV plant. The increase in grid meshing damped the oscillations but increased the reactive power on the grid. The surge in reactive power—and consequently the voltage—was the main factor leading to the blackout. Equipment disconnected when they detected that operational limits had been reached.

The dynamic voltage control from power plants was insufficient. If enough of these plants had been online, the reactive power could have been dissipated. Renewable power plants are capable of providing such control if they have the correct hardware installed, but currently, regulations prevent them from doing so.

The DC interconnection between France and Spain was set to export 1000 MW. If the AC mode had been activated, it could have provided an additional 2000 MW. However, this would not have been enough to prevent the blackout.

In summary, the event chain looks similar to this:

Graph showing the chain of event causing the 28th April blackout.

To dampen the oscillation, exports were reduced and the RED added more lines to the transport grid. This caused an increase in reactive energy. A surge in voltage occurred locally on the 220kV grid and disconnected local generators, causing the accumulation of more reactive energy and then increasing the voltage on the entire network (400kV, 220kV, 130kV). This new surge caused the disconnection of other equipment and the cascade effect started. At some point, not enough power was available, causing a frequency drop and leading to the blackout.