Thinking about power outages

This article has been prompted by the major power outage at Heathrow Airport that occurred on March 20^th-21^st and its aftermath. I cannot claim any special understanding of the reasons for the outage or why the failure of one electricity substation had such a drastic impact on the operations of the airport. For such information I urge you to read the admirably clear and dispassionate description provided by Kathryn Porter in her Watt-Logic blog. Rather I want to use the episode to examine how we should think about system reliability and the costs of power outages.

Countries have very different expectations about electricity network reliability. As an example, I spent an extended period working with power companies in Hong Kong. Their system reliability target was 99.99% or no more than 50 minutes of outages in any year. In practice, their performance was closer to 99.999% reliability. In contrast, when I lived in Washington DC, power outages – albeit short ones – were a regular event during the summer. This was because the system was heavily loaded during the hot months of the year. Peak demand was no higher than in Hong Kong on a comparable basis, but the PEPCO network was vulnerable to the frequent lightning strikes that occur during hot weather.

During the same period the Scottish Power network serving our house in a rural part of the Scottish Borders was horribly unreliable with frequent, although usually brief, outages. The local joke, perhaps true, was that power outages were caused by cows rubbing against electricity poles. Its reliability has been better in the last decade, as Scottish Power invested (relatively) a lot to upgrade the network in an area of low population density. Even so, a large expansion in forestry and the ageing of older trees has meant that network reliability is deteriorating again because regular windstorms mean that falling branches and trees bring down electricity cables.

I give these examples to highlight three related questions about electricity network reliability:

A. Not all outages are the same, so how do we decide what matters most? Localised but brief interruptions may be infuriating if you lose material in open computer files and must redo work, but appropriate software can minimise the losses. But, when critical equipment must be reset and checked after even short outages, the costs of short interruptions can escalate rapidly. Longer outages are often not under the control of power companies, because falling trees often block access to damaged power lines. In rural areas, network resilience is often severely affected by the failure of other infrastructure providers to invest adequate resources in maintenance and improvement of their assets. This is a critical problem in the UK where local authorities have responded to budgetary and legislative pressures to shift spending from the provision of infrastructure services to the provision of social services, broadly construed.

B. What should be the balance between customer and system provision of reliability services? Households and businesses may invest in UPS protection for key equipment, but what is the trade-off between spending on private measures and on greater network resilience. In Scottish Borders and the North of England, there have been two episodes of extended power outages in the last 5 years caused by windstorms. Many households living in rural areas will consider whether installing some combination of a generator, solar panels, and battery systems is a sensible investment. These may not help those using electric vehicles, heat pumps or other forms of electric, while solar panels are of little help for the 4+ months of Scottish winter. Thus, improving network resilience is an essential – though largely neglected – component of any program to promote alternatives to fossil fuels in rural areas.

C. How much are we (either individually or collectively) willing to pay to reduce the probability of power outages? There may be a significant difference between the collective willingness to pay for insurance and the sum of individual valuations. In part, this is because the costs of supply interruptions vary greatly across customers. Some customers are willing to pay for private protection while other customers may believe that they will – or should – be able to rely upon what is implicitly some kind of social safety net.

Studies of how electricity customers value the benefits of reducing the frequency and length of power outages – what economists refer to as the value of lost load - are difficult to carry out and are thus rare. In the UK the primary source of evidence on the issue is a study carried out by London Economics for Ofgem that was published in 2013. [I have converted the results of the report from 2012 to 2024 prices by applying a price multiplier of 1.37.]

London Economics recommended that the overall value of lost load should be £23,200 per MWh averaged over different types of user and different methods of measurement. That average is useful when considering random power outages caused by, for example, widespread weather events or network failures. However, the average masks enormous differences between types of users, and type of power outage.

One distinction, which is common to the type of study carried out by London Economics, is between what is called the willingness to accept (WTA) power cuts – i.e. deliberate load shedding – and the willingness to pay (WTP) to reduce the risk of power cuts. In the case of household customers, the average WTA was far higher than the average WTP. In some cases, the average WTA was more than 5 or 10 times the average WTP. Perhaps surprisingly, the average WTP was not significantly different from zero for both peak and off-peak periods on weekdays.

London Economics argue, with good reason, that WTP estimates are likely to be seriously biased. It is known that respondents are often confused by requests to value changes in the probability of infrequent events. Being asked to accept a power cut lasting 1 hour at within a known window of time is a clear proposition. In contrast, it is harder to attach a value to reducing the risk of a 1-hour power cut from 2 times per year (roughly 1 per 5,000 hours) to 1 time per year (roughly 1 per 10,000 hours). Further, respondents often reject the basic premise, i.e. they believe that there should be no unplanned power cuts on the grounds that they are paying for a reliable network service.

The average WTA for household customers was £13,500 per MWh. Businesses, especially SMEs with limited bargaining power, have much higher values for lost load with an average WTA of £51,600 per MWh. This is not surprising as the impact of power cuts on output is large, while SMEs may have limited options for buying insurance by investing in generator or battery backup.

If the business value of lost load is used the cost of the power outage at Heathrow – about 1.1 GWh over 24 hours – was about £55 million. This is comparable in magnitude with reports that the wider cost to the airline industry might be £60-£70 million. Yet, since up to 500,000 passengers were affected either directly or indirectly (due to disruption of flight schedules) by the power outage, this cost seems relatively low. It is perhaps £100-£150 per passenger affected. I suspect that most would claim that extra expenditures on meals, accommodation, etc due to delays were at least that amount without any allowance for inconvenience and stress.

Whether the total cost of the Heathrow incident was £55 million or £220 million is only the beginning of any assessment of how much should be spent on avoiding a repetition and if so by whom. As is often the case with major system breakdowns, the closure of the airport was due to a conjunction of events – (a) a serious fire in transmission equipment at a key grid substation, (b) difficulties in switching to electricity supplies from other substations, and (c) either an absence of or a shortfall in local backup supplies.

While grid substations have low failure rates, fires, lightning strikes and breakdowns which cause substations to shut down partially or completely occur intermittently across the transmission system. It may be argued that trying to reduce costs the regulator and transmission operators have extended the life of transformers and other transmission equipment to a point where age-related failures will become increasingly frequent. Certainly, the average age of GB grid assets is considerably higher than for networks such as Hong Kong.

The primary lesson for managing risks of power outages is that it does not make sense to apply uniform replacement criteria to all network assets. It is both efficient and reasonable to replace assets which serve Heathrow and dense urban areas after 40 or 50 years, while replacing assets in less densely populated areas after 60 or 70 years. The key requirements are (i) to be clear and honest about such a strategy, and (ii) to ensure that network charges are differentiated to reflect the costs of meeting higher reliability standards in the areas which require or want them.

However, while the precipitating factor was a grid failure, the primary lesson from the Heathrow episode is that large – and medium - electricity consumers cannot be excused when they fail to implement adequate backup procedures. From the usual confusion it seems clear that Heathrow had access to power from two other substations with the capacity to meet the airport’s demand. The problem, though it is barely credible, was that airport was unable either to switch its main supplies and/or that equipment could not be restarted safely without a prolonged interruption.

The power outage at Heathrow from far from the worst-case scenario. It occurred late at night when the number of incoming and departing flights was at its lowest. There was a period of 4 to 6 hours during which power supplies could have been switched or backup supplies mobilised. Certainly, staff had to be contacted and to come into the airport during the night, but that is what emergency rotas are for. The electricity network cannot be blamed for what has the appearance of incompetence and lack of preparation on the part of the airport’s management.

This brings us back to the division of responsibilities between the network operator and electricity customers. One can infer that National Grid believed that it had fulfilled its responsibility by providing multiple network connections, of which only one failed. Heathrow seems to have acted on the basis that each of its separate supplies was completely reliable and prepared for what I would describe as internal or on-site network failures.

What happened could easily have been avoided had there been clarity about who was responsible for what. As Kathryn Porter points out, other airports rely on a combination of battery storage plus either aero-derivative gas turbines or reciprocating engines. Such equipment could be used to backup supplies from each of the substations serving Heathrow with minimal disruption and internal changes. Alternatively, Heathrow could have opted for internal management and failover of the separate supplies. That would require the installation of an internal power network with appropriate automatic switching. In practice, there was no such clarity, so the critical choices were ducked.

Over a life of 20 years the annual cost of installing and operating a combination of batteries and gas turbines providing up to 60 MW of backup capacity would be about £14 million per year at 2024 prices using Heathrow’s regulated cost of capital. That is equivalent to a cost of about 14p per passenger for the 83.9 million passengers who used Heathrow in 2024. That sum is equivalent to 0.4% of Heathrow’s total revenue of £3.35 billion in 2024 and 1.5% of its pre-tax profit of £917 million. None of these figures seem disproportionate as a sum to pay for insurance against a very infrequent but extremely disruptive event.

Heathrow’s airline customers are notoriously reluctant to accept increases in the passenger charges which they pay. This is because they believe that Heathrow’s operating costs are excessive and that they do not benefit sufficiently from the commercial revenues generated by very high levels of passenger traffic. There is a clear case for splitting the cost of insurance between regulated passenger and other charges, which account for70% of total revenues, and commercial activities. The cap on passenger charges for 2025 is £23.73. Increasing that cap by 10p would certainly not be a significant imposition, provided it is accompanied by immediate action to ensure that there is no repetition of the extended outage that closed the airport.

In considering the trade-off between network and customer responsibility for supply reliability, cases like Heathrow are relatively straightforward. Large businesses have the financial and human resources required to invest in and manage the risks of power outages. Network operators should not be expected to spend the money necessary to achieve very high levels of reliability for such customers, unless they are permitted to recover the costs from those customers.

For households and SMEs, the costs of individual measures to enhance the reliability of electricity supplies are often prohibitive. Small generators are useful only to protect individual circuits. Larger generators to supply a household or business must be isolated from network supplies and assume that users are able to store sufficient fuel to run for more than a few hours. The costs of buying and installing battery storage systems are even higher and they will barely cover 1-day’s electricity consumption for the type of households who can afford them.

Ofgem has adopted rules which specify the amount of compensation that distribution network operators (DNOs) should pay to their customers if they suffer a power cut longer than 12 hours. The goal is to encourage DNOs to give more weight to minimising the impact of longer power outages. Unfortunately, like all too many regulatory incentives, it is half-hearted and poorly designed.

If we use an average load of 1 kW during the winter months when power cuts are most likely, the average value of lost load cited above is equivalent to £557 for a 24-hour outage. In contrast, the Ofgem compensation for normal weather is £95 for a 12-hour outage plus £40 for each subsequent 12-hour period. The total compensation for a 24-hour outage is less than one-quarter of the value of lost load.

For power cuts caused by severe weather, the compensation is reduced to £85 for a 24-hour (less severe storms) or 48-hour (more severe storms) outage depending on the storm severity plus £40 for each subsequent 6-hour period. The determination of storm severity is made by Ofgem, not by the Met Office or any external referee.

Ofgem’s arrangement is absurd as an incentive scheme. It is patently designed to limit the liabilities of DNOs arising from their failure to storm-proof distribution networks. Average compensation per hour of normal weather outages falls with the length of the outage. This is completely counter-intuitive, as almost everyone would agree that the inconvenience of an outage per hour increases with its length. However, the logic is obvious: outages which cannot be fixed remotely – and, thus, quickly – require repair teams to be mobilised. DNOs do not want to spend lots of money on out-of-hours (including weekend) repair work.

Even if it is accepted that storm-related outages should be partially covered by force majeure provisions, any compensation scheme that reflects the interests of customers rather than suppliers ought to incorporate a steeply increasing schedule of compensation for outages greater than either 24 or 48 hours. Many people have little choice other than to move out of their houses – and businesses to shut down – during extended outages. This is particularly the case for storm damage that occurs during winter in northern regions when the amount of daylight is very limited.

If DNOs are allowed to limit their liabilities as at present, the condition should be that in rural areas that are most prone to extended outages DNOs should be required to facilitate and, indeed, actively promote the adoption of backup supplies. In effect, the goal should be to split the costs of greater service resilience between network operators and their customers.

It is easy for commentators to call for greater clarity about matters which operators and regulators prefer to fudge. However, fudging is not a good strategy if we need businesses or households to invest or act in other ways to reduce the impact of power outages. Clarity about what level of reliability network operators can and should offer ought to be a primary requirement.

Even so, lack of clarity is hardly an excuse for any large business for which power outages are likely to cause substantial inconvenience and financial losses. In the case of Heathrow there is, I believe, a reasonable case on current evidence for imposing severe penalties commensurate with the consequences of their incompetence or carelessness.

It is more difficult to determine the appropriate balance between network and customer responsibility for managing the risk of power outages for household and SME customers. A uniform policy makes little sense, but that has been the approach followed by Ofgem and DNOs, perhaps more in rhetoric than reality. A much greater level of honesty is required.

No reasonable person believes that a network operator can maintain the same level of network reliability in the more remote valleys of the Scottish Borders as in centre of Edinburgh. Such differences in service quality might be compensated is by different network charges or by support to enable customers in areas with lower levels of network reliability to invest in individual options to lessen the impact of power outages.

Unfortunately, there is a catch – at least in the UK. Both Ofgem and the government have strongly resisted the idea of unbundling electricity charges. Instead, either because of ideas about how competition should work or to hide the role of government levies in increasing the prices paid by domestic and SME customers, they insist that energy suppliers should offer a single all-in electricity price for customers in a specific DNO area.

That policy is, of course, fundamentally incompatible with a recognition of differences in network reliability in prices. The whole structure of electricity tariffs for household and SME customers rests on a deliberate pretence that all customers can benefit from a similar level of service and be charged the same prices wherever they are. Any action which might undermine that pretence is strongly resisted. In dealing with practical issues concerning network reliability, we finish up with yet more pretence on top of confusion and a system of compensation that is completely inconsistent with the basis on which system reliability is addressed.

Such are the consequences of a regulator whose independence has been entirely compromised by political interference and lobby groups.