Living with electricity grid instability

As most readers will know, the Iberian Peninsula transmission grid covering Spain and Portugal suffered a severe failure shortly after 12.30 CEST on Monday April 28^th. Nearly 24 hours later power supplies have been restored to most customers in the peninsula, but full recovery to the status quo ante is likely to take several days. In the meantime, both businesses and households experienced substantial costs and inconvenience as a direct or indirect consequence of the grid failure.

The full details of the outage may not be known for many months, if at all. Governments and grid operators are usually very reluctant to admit to what has gone wrong in such cases. Even when official inquiries are established, experience tells us that concerted efforts are made to obfuscate or place the blame on minor factors, especially when there is a high level of political polarisation as in Spain.

Even so the broad outline of what happened seems clear. Shortly after12.30 CEST the system frequency dropped suddenly from 50 to 49.85 Hz, still well within conventional limits but indicating an abrupt loss of power. This loss of power is believed to have been solar generation in the south of Spain, perhaps because of synchronized inverter oscillations. The drop in frequency was either immediately preceded or accompanied by a widespread disconnection of wind and solar generation that could not operate safely outside a very narrow frequency band.

The electricity system had too little inertia to stabilise system frequency with the consequence that generator disconnections cascaded through the system to the point at which almost all grid connected generation was lost. At the time the amount of gas generation, the main source of spinning reserve which provides system inertia, was very low.[1] Grid connected supply in Spain fell from about 32 GW to close to zero within 30 minutes and there was a similar collapse in Portugal from about 6 GW.

The system might have been stabilised by drawing upon supplies from France. Earlier on Monday morning Spain had been exporting power to France, but these flows had tailed off by the middle of the day. However, emergency supplies from France would have had to travel across the key link down the east coast from Catalonia. This zone was at the heart of the events which prompted the disconnections, so external support could not help. A further complication was that some of the hydro plants required to “black start” – i.e. synchronize generators which were restarting - the system were offline for scheduled maintenance, thus delaying the process to restoring grid power supplies.

Other sources refer to rare meteorological conditions. In truth, such explanations don’t matter very much as electricity systems are supposed to plan for an abrupt loss of power at different points in their networks. The point of contingency planning is not to foresee every possible eventuality but to build a system that is resilient enough to withstand the abrupt loss of some defined proportion of its power supply – often up to 10% of total demand. What we saw was that the Iberian grid was not able to cope with whatever caused the original frequency drop. From that point on the only question was how severe the grid outage would be.

What happened is discussed in more detail in another Substack article by Michael Shellenberger. In essence, he points out that the outage was not a bug but a feature of any electricity system with a high level of reliance on solar and wind generation. That may be overstating the case, but there have been repeated warnings in Europe of the increased risks of grid instability when system inertia, usually provided by thermal power plants, is too low.

Major grid failures are a regular occurrence around the world, even though they are infrequent for individual systems. We tend to remember the most serious ones that affect us or are widely publicised such as the US North-East and Italian blackouts in 2003 or the Southern Brazil blackout of 1999, all of which affected more than 50 million people. On a more local scale there was a shorter blackout in the UK which affected mainly Eastern England and London in August 2019.

There are important lessons to learn. First, a frequency drop to 49.85 Hz is remarkably small to cause a major system failure. Using data for the GB system from 2014 to 2020 the minimum frequency in each 30-min settlement period was less than 49.85 Hz in more than 10% of settlement periods and less than 49.8 Hz in more than 1% of settlement periods. What happened in Spain was not normal, but neither should it have had such severe consequences. Second, the share of generation from solar and wind in the hour from 11.30 to 12.30 (CEST) was 69%, high but not close to 100%. The PR claims for very high reliance on renewable generation rely upon the large contribution from hydro plants.

If there is a connection between the two aspects of the grid failure, the inference is that systems that rely predominantly on solar and wind generation may be more, perhaps greatly more, prone to grid outages due to random failures or external events than systems with diversified supplies that are less reliant on types of generation that provide no associated system inertia. Alternatively, some companies offering flywheels – described as synchronous compensators – claim that they can offset the risks associated with relying on solar and wind generation.

Economic analyses of flywheels are highly speculative, largely because they are usually presented as alternatives to battery storage systems. As pure grid stabilisation devices, they are expensive with a projected capital costs of $1,000 to $1,500 per kW or $150 to $200 per kW per year at 2024 prices – see Mustizer Rahman et al (2022). That would amount to £1.6 billion a year for a reasonable capacity of 10 GW for the GB market. Some might judge that as affordable as a way of reducing reliance on gas plants for system inertia. Even so, widespread adoption seems unlikely while outages are infrequent and can be blamed on random factors.

Such details matter because the shift to greater reliance upon low inertia forms of generation will leave electricity grids much more vulnerable to loss of power due to failures at both generating plants and network nodes. Whether it is a fire at a substation or a lightning strike that takes a grid line out of operation, the margin of resilience will be less and the probability of a cascading sequence of failures is greater.

On the demand side, the pressure to electrify transport and heating systems will increase the probability of network failures because such changes will increase network loads. In the UK there are plans to strengthen both transmission and distribution networks to handle increased usage, but the upgrades required are both complex and expensive. Recent experience is that network upgrades have lagged far behind increases in solar and wind generation. What happened in Spain is a clear example of the consequences of such delays if things turn out badly.

So, how much is it worth spending on reducing the probability of network outages and their impact? From a network perspective the calculations are harsh and probably unpopular. Major network outages are infrequent events – certainly less than 1 in 20 years for most large systems. In the case of Spain and Portugal the total lost load was of the order of 300,000 MWh. At a value of lost load (discussed in my article on the power outage at Heathrow) of £15,000 per MWh, that implies a cost of about £4.5 billion – a very large sum.

However, with a probability that such an event might occur in any year of less than 5%, it would only be worth spending £225 million per year on measures to mitigate such a risk. That is far less than the annual cost of providing a large capacity of inertia in the form of flywheel installations. Low-cost measures can be justified. Even so, it is unlikely that that the incremental benefits from spending more than, say, £100 million per year would be high enough to justify the expenditures incurred.

The implication is that any system that relies heavily on solar and wind generation is likely to experience more frequent power outages, though usually not on the scale experienced in Spain and Portugal. That is just an inevitable outcome of the loss of system inertia and the change in the trade-off between the costs of ensuring grid stability and the costs of grid outages of various extents and duration.

From the perspective of electricity users, the increase in the risks of grid outages should change their calculations about how much to invest in backup. The lesson for large and medium organisations is that either generators or other forms of backup are likely to be essential. Widespread grid outages rarely last for more than 12 hours, at worst 24 hours, so backup generators will only be required to run for limited periods. The amount of fuel that must be stored is not large but how much of the organisation’s operations should be protected is an important issue.

Battery systems are rarely satisfactory as they are usually designed to operate for 2 or 4 hours, which is certainly not sufficient. The capital cost of 8-hour or 12-hour storage, whether batteries or compressed air or other options, tends to be high. System and other incentives favour short duration storage, so it is nearly impossible to construct a plausible economic case for medium duration storage.

The options for households and small businesses are even more limited. Neither small generators nor household-scale battery systems are suitable for high load uses – heating, cooling. cooking or vehicle charging – so these need to need to be disconnected or switched off. Safety regulations complicate the installation of generators as alternatives to regular electricity supplies and most are not designed to run for more than 6 to 8 hours. Larger household battery systems of up to 13 kWh can provide backup for lighting, electronic equipment, etc for 12 to18 hours but they are expensive as backup for infrequent use. When connected to domestic solar systems, they may make economic sense but any insurance against power outages is incidental in such cases.

There is a larger issue that has barely been addressed in the rush to promote solar and wind generation. As societies we have invested heavily on the assumption that power is always on. We rely increasingly on the centralised provision of financial, information and control services – from banking to the management of transport systems. This is both efficient and convenient, but only so long as distributed communications are working.

Data centres will rapidly go out of business if they do not have adequate backup. But what about your local mobile phone mast or your local fibre optic node that connects you to the central network node? In the past, telephone exchanges had battery backup which enabled phone lines to work despite local power outages. Customers are advised to have UPS backup but most such units don’t run for more than 1 hour and many won’t work at all because they are not set up properly. In any case, such backup arrangements are useless if the wider network is not functioning. It is trite but crucial to remember that networks are only as strong as their weakest link.

In summary, the extreme disruption caused by the network failure in Spain and Portugal highlights the inconsistencies of current economic and social trends. The rapid shift towards reliance on solar and wind generation means that electricity networks will experience higher risks of both minor and major power outages. As we rely more on always-on methods of communication, the costs of such power outages will grow and will fall heavily on households and small businesses that have limited capacity and face very high costs to insure themselves against such outages.

On the other side, network operators cannot justify the large costs of effectively reducing the risks of widespread outages, while they have limited interest in more minor outages. For them the costs of accommodating increasing levels of solar and wind generation look likely to stretch their financial resources beyond breaking point. The concerns of electricity consumers rank well behind the pressure from governments to respond to the demands of producers.

The eventual outcome looks likely to be a significant deterioration in the quality of network services for small customers and a significant increase in costs for larger customers. Not a new story, but yet another reason why the energy transition is easily seen as a slow-motion commitment to economic decline.

[1] The best way of thinking about system inertia is by using the analogy of a massive flywheel. If the input which powers the flywheel is cut off it will continue to spin though its speed will slow down due to friction and any use of the magnetic fluxes created by the flywheel. The inertia is the angular momentum of the flywheel – i.e. mass times rotational speed. Gas plants – and, in the past, coal or oil plants – provide the equivalent of such flywheels and the amount of power input can be rapidly adjusted to keep the flywheel spinning. Nuclear plants are less useful because they are designed run at constant speed whenever possible with changes being gradual rather than very quick.

There are (semi-serious) proposals to provide more system inertia for systems that rely on solar and wind power by building huge flywheels that would be powered by electricity. They are semi-serious because it is unclear where the injection of power required to offset the slowing down effect of drawing on the flywheel’s rotational inertia would come from when electricity supplies are abruptly cut-off. Batteries powering flywheels as a way of supplying system inertia has all the flavour of Heath-Robinson inventions.