Affordable and reliable energy: where to from here?

Here is the transcript of a speech I gave to the New South Wales Energy Forum on 19 July. Unfortunately it was by Zoom and not in person…

***************************

Good afternoon, and thank you for inviting me to speak today about the practical realities of the energy transition and what lessons we can draw from the experiences in Australia, Britain and elsewhere.

Electricity systems don’t respond to political ambition or financial models. They respond to the laws of physics. These are not laws that politicians can repeal!

Let me be clear at the outset: I do not believe the energy transition, as currently conceived, is sensible. I’m not against cleaner energy, technological progress or reducing genuine environmental harm. Nor do I believe that the electricity system should remain frozen in the form it took fifty years ago.

But replacing reliable, energy-dense generation with large volumes of intermittent wind and solar is a poor use of both capital and natural resources. It requires us to build far more generating capacity than we need, duplicate it with dispatchable capacity for periods when the weather does not co-operate, expand the network to connect increasingly remote and dispersed assets, and then procure additional equipment and services to replace the essential electrical characteristics lost when synchronous power stations close.

This isn’t an elegant transition from an old system to a better one. It’s the construction of several overlapping systems, at enormous cost, in an attempt to reproduce the reliability we already had.

Britain is presented as a success because coal, which supplied around 40% of our electricity little more than a decade ago, has disappeared from the power system, while offshore wind has grown rapidly. In fact the choice to move away from coal had nothing to do with environmental concerns – coal started to run out just as North Sea gas was discovered and a switch between them was economically rational.

Had we discovered coal when gas was running out we would have made the opposite choice and we’d be crying about the emissions now.

But this fuel switching decision and the subsequent choice to replace gas with wind and solar does not tell us whether the policy was rational or the resulting system secure. Installed capacity is not dependable capacity, annual production is not production when electricity is needed, and a megawatt hour is not the only thing a power station provides. Policymakers have counted visible inputs while overlooking total system performance and cost.

The central problems are intermittency, energy density and trying to force direct current technologies onto alternating current grids.

Wind farms generate when the wind blows and solar panels generate when the sun shines, regardless of the level of demand.

In Britain, we tend to talk about cold, dark and still winter weather. Demand is high, wind output can remain low for days, the solar contribution is meagre and it disappears altogether before the evening peak. That’s a serious problem, but it isn’t the only weather pattern that matters.

Australia’s more obvious challenge is heat, particularly heat combined with still air. When temperatures soar, air-conditioning demand surges. At exactly the same time, a stationary high-pressure system can suppress wind generation across a very large area.

Solar may produce strongly in the middle of the day, but it ramps down far faster than people expect, and extreme heat itself reduces the efficiency of solar panels.  Demand doesn’t disappear when the sun begins to set.

Europe has faced the same combination this summer: intense heat, weak wind output and high cooling demand. On 23 June, Britain came uncomfortably close to power shortages as demand exceeded the system operator’s forecast and available reserves became dangerously low.

The control room curtailed exports and there are now whistleblower allegations that the grid was insecure and the system operator’s management has been covering it up. Investigations are underway after questions were asked in Parliament.

Elsewhere in Europe, high river temperatures forced cuts to nuclear output over reduced access to cooling water, and low river levels have restricted coal deliveries to thermal power stations.

Heat also increases line losses and losses from transformers. Thermal power stations are also less efficient: the entire grid has to run faster just to stand still. Transformers are also harder to cool, increasing the risk of equipment failure.

So whether the critical event is a cold, dark and windless winter evening or a hot, still summer evening, the underlying requirement is the same. The system has to meet demand continuously, not merely on average and not merely when the weather is helpful.

This is why claims that wind and solar are the cheapest forms of electricity are so misleading. They may have a low marginal cost when they happen to generate but consumers don’t buy isolated megawatt hours at the gate of a wind farm. They buy a reliable electricity service, delivered at the right voltage and frequency every second of the year.

The relevant cost isn’t simply the cost of the wind turbine or the solar panel. It’s the cost of the whole system: generation, backup, storage, transmission, distribution, balancing, constraint management, reserve and stability services.

If wind and solar replace dispatchable generation, what supplies the system during a prolonged period of low renewables output? If dispatchable generation has to remain available, then it hasn’t really been replaced – we’ve simply built another generating fleet alongside it and forced the original fleet to recover its costs over fewer running hours.

That makes the dispatchable capacity less economic, even as it becomes more important to security of supply.

The inefficient use of capital is matched by an inefficient use of physical resources. Wind and solar have low energy density and relatively low capacity factors, so producing a given quantity of electricity requires large numbers of machines spread across very large areas. Each machine needs steel, concrete, copper and other minerals.

Offshore wind also requires extensive foundations and undersea cables, while solar consumes land, glass, aluminium and network equipment.

Batteries add further demand for processed minerals, yet most grid batteries store only a few hours of output. That can be extremely useful for rapid response or for moving some solar generation into the early evening. But a battery that can discharge for two or four hours isn’t a solution to several days of low wind.

All these materials have to be mined, refined, manufactured, transported and eventually disposed of or recycled. Much of the mining and processing takes place in countries with weaker environmental safeguards and poor labour protections. We call the final electricity clean because we choose not to count the damage elsewhere in the supply chain.

There’s an extraordinary asymmetry in this debate. The environmental impacts of fossil fuels and nuclear power are examined in minute detail, as they should be. But the land take, mineral intensity, habitat disruption and waste associated with wind, solar, batteries and thousands of kilometres of new transmission are too often treated as incidental. We’ve confused low carbon at the point of generation with low environmental impact overall.

There’s also a deeper electrical-engineering problem. Generating electricity and operating an electricity system aren’t the same thing. A large synchronous generator does far more than simply produce energy – it contributes inertia, fault current, voltage support and system strength because of its fundamental physical characteristics.

Historically, those services arrived bundled with the electricity, so markets didn’t need to value them separately and policymakers barely noticed them.

Wind turbines, solar panels and batteries connect through inverters and don’t behave in the same way. Once synchronous generators retire, the system operator has to procure services that were previously broadly free. We replace the energy, and then we pay again to replace the engineering characteristics of the plant we displaced.

So a system can have far more installed generating capacity than it could ever realistically need and still be insecure, because capacity isn’t the same as capability. If the output is in the wrong place, arrives at the wrong time or can’t support the network during a disturbance, its nameplate rating is irrelevant.

Britain shows how these costs accumulate. Our transmission network was designed largely around dispatchable power stations built near established industrial centres. It’s now being reconfigured to connect weather-dependent resources dispersed around the country and far offshore, often hundreds of kilometres from the largest centres of demand. That requires enormous investment in new lines, substations and subsea cables.

Yet building more network doesn’t abolish constraints when generation is developed faster than the infrastructure required to transport its output. On windy days Britain can have too much electricity in northern Scotland and not enough grid capacity to move it south.

The system operator then pays wind farms to reduce output while paying gas plant elsewhere to increase output. Consumers therefore pay twice – for the electricity they actually use from the gas power station and the electricity they don’t use from the windfarm.

Britain’s constraint and balancing costs have risen sharply, while the range of ancillary services required to keep the grid secure has expanded. Synchronous condensers are being installed to provide inertia, voltage support and fault current without producing electricity.

Gas power stations are sometimes instructed to run for stability even when their energy isn’t needed. Batteries are paid to provide rapid frequency response. Each intervention may be technically sound, but together they expose the friction that wind and solar introduce to the grid.

Nor does storage solve the fundamental problem. Batteries are excellent at responding quickly and can be valuable for frequency management, short-duration balancing and moving some solar output from the middle of the day into the evening.

But the challenge in a renewable-heavy system isn’t merely a five-minute mismatch. It can be a multi-day or even multi-week period of low wind, sometimes accompanied by unusually high demand.

Providing enough battery capacity to bridge such periods would require a huge overbuild of generation to charge it, an immense quantity of minerals and capital, and storage assets that might be called upon only occasionally. Their economics are difficult precisely because the strategic reserve we most need must sit idle much of the time.

Pumped hydro can provide longer-duration storage where geography allows, but suitable sites are limited. Hydrogen remains inefficient, expensive and largely unproven as a route for storing electricity at power-system scale.

The Iberian blackout in April 2025 revealed the physical risks behind these economic distortions. The initial public reaction divided along familiar tribal lines. Renewable advocates insisted that wind and solar had nothing to do with the collapse, while critics described it as definitive proof that renewables could never be operated securely. The detailed reality is both more complicated and more important.

The disturbance involved voltage instability, inadequate dynamic voltage control, protection behaviour and the failure of inverter-based generation to remain connected as required. More than eight gigawatts of inverter-based resources disconnected within seconds, including installations that should have ridden through the fault conditions.

With relatively little synchronous generation online, the system couldn’t arrest the cascade and the Iberian Peninsula went dark.

The lesson isn’t simply that the system lacked inertia. Inertia has become a convenient catch-all whenever a grid disturbance occurs, but synchronous machines also provide short-circuit current and voltage stiffness. These characteristics interact, particularly during faults, and they can’t safely be reduced to a single metric.

Iberia demonstrated the difference between compliance on paper and performance in the real world. An inverter may be certified against a grid-code model and still respond differently when exposed to a complex disturbance involving both voltage and frequency changes across a weak network.

Protection settings designed for a system dominated by large rotating machines may also behave unpredictably as fault currents become smaller and less familiar. Simulations and certificates are valuable, but neither is a substitute for demonstrated behaviour under credible system conditions.

The blackout also illustrates the danger of normalisation of deviance. As synchronous generation falls and inverter-based penetration rises, system operators may repeatedly operate closer to boundaries that would once have been considered uncomfortable. Most days nothing catastrophic happens, and each successful day is then treated as evidence that the reduced margin was safe.

But the absence of failure isn’t proof of resilience. It may simply mean the system hasn’t yet encountered the disturbance capable of exposing its weakness. It’s a bit like thinking running out in traffic is safe because you did it a few times and lived to tell the tale.

Electricity networks operate in milliseconds, not annual averages. They must survive the wrong fault in the wrong place on the wrong day. Iberia had abundant generation before the blackout but it lacked the ability to keep that generation connected and the system stable when conditions deteriorated.

Australia is especially important because it’s further along this path than almost any other country. The National Electricity Market spans thousands of kilometres and incorporates some of the world’s highest penetrations of rooftop solar. South Australia has already experienced periods when inverter-based resources supplied almost all local demand, while conventional generators have progressively retired or reduced their running hours.

AEMO and the network businesses deserve credit for being more candid than many policymakers about the resulting engineering challenge. Their work on system strength, minimum synchronous generation, fault levels and protection acknowledges that adding megawatts of renewable capacity doesn’t automatically add the capabilities required to operate a secure network.

The installation of synchronous condensers in South Australia is perhaps the clearest physical expression of this reality. These are large rotating machines that consume capital and resources but produce no electricity. They’re installed solely to restore some of the inertia, fault current, voltage support and system strength lost when synchronous power stations are removed.

Technically, that may be the correct response to the system policy has created. Economically, however, we should ask why consumers are being required to pay for wind and solar generation, dispatchable backup, major network expansion and synchronous machines whose purpose is to recreate services previously provided by the generators being displaced. If a supposedly cheaper technology requires an entire supporting architecture before it can be used securely, then it ceases to be cheap.

Grid-forming inverters are frequently presented as the eventual answer. Unlike conventional grid-following devices, they can establish a voltage and frequency reference rather than merely following one, and in controlled settings they can reproduce some behaviours associated with synchronous generation. The technology is promising and deserves serious investigation, but promise isn’t proof of equivalence.

Some designs revert towards grid-following behaviour during faults, precisely when grid-forming capability is required, and maintaining stable control becomes harder as current limits are reached.

Power electronics can’t produce unlimited fault current, and software-controlled responses introduce interactions that aren’t yet fully understood at system scale. It would be reckless to retire proven synchronous capability today on the assumption that future inverter controls will reliably reproduce all of it tomorrow.

This is where engineering judgement has to confront political ideology. Engineers can devise ways to make a renewable-heavy grid more stable, just as they can solve many difficult problems if they’re given enough time, equipment and money. But the fact that a problem can be mitigated doesn’t mean it was rational to create it. Technical feasibility isn’t the same as economic wisdom.

We’re deliberately building generation that breaks the engineering logic of our electricity grids and this is forcing us to spend $billions on correcting the problems introduced by that generation. When we pay $billions in subsidies to build it in the first place, that adds insult to economic injury.

We could build vast networks, enormous quantities of storage, synchronous condensers, grid-forming batteries and fleets of underused gas stations held in reserve. The relevant question is whether doing so produces a better outcome than investing directly in reliable, energy-dense generation such as nuclear, hydro where geography supports it, and modern gas generation. Those technologies can supply electricity when it’s required while contributing many of the essential characteristics the network needs.

Policy has focused on the cost of carbon while neglecting the cost of complexity, duplication and fragility. Governments count installed renewable capacity because it’s visible and politically convenient, but consumers don’t care how many gigawatts have been connected. They care whether the lights come on, whether their bills are affordable and whether industry can invest with confidence in a reliable supply.

A system that occasionally produces an excess of very cheap electricity but requires billions in subsidies, networks, balancing and backup isn’t genuinely cheap. A system that meets an emissions target by closing domestic industry and importing manufactured goods hasn’t eliminated environmental harm. It has exported both the emissions and the jobs. Energy policy should be judged by its effects on human welfare and economic resilience, not by the elegance of its targets.

A more rational approach would begin with the outcome the electricity system must deliver: secure and affordable power at all times, with environmental impacts assessed across the whole supply chain. Technologies should then compete on their contribution to that outcome, including availability, location, useful life, material requirements, network impacts and system services.

Firm power shouldn’t be treated as an embarrassing bridge to a weather-dependent future. It’s the foundation of an industrial economy. Nuclear power, gas and hydro each involve trade-offs, but they’re energy-dense, dispatchable or both. They require far less duplication than a system designed around intermittent output, and they preserve options while technologies evolve.

If I leave you with one thought today, it’s this: don’t think of the energy transition as the straightforward replacement of dirty generators with clean ones. It’s the replacement of a compact, dispatchable and electrically stable system with a larger, more resource-intensive and more complicated system that still depends on much of the infrastructure it was meant to supersede.

The objective shouldn’t be to complete an energy transition regardless of cost. It should be to build an energy system worthy of the people and industries that depend upon it.

Markets decide who gets paid. Governments may decide which technologies they want. But physics and engineering decide whether the system works, and economics ultimately decides how long society can afford to pretend otherwise.

Thank you.