The energy transition is technological change
Model outline
The energy transition is the greatest technological change in history. This change will decide the competitiveness of entire economies if the responses to it are geared only to the short term and to activism. To meet this change as before, without sufficient foresight and without suitable models, is more than negligent.
There are many models of the energy transition and almost all of them have failed. But there are reliable models. The empirical evidence of past successful technology transitions has proved to be a reliable estimator for technology transitions currently under way. That is the basis for reliable models, for managing the economic challenges, for holding one's own in competition.
Empirically based
Computer model
At the beginning of 2018 we created an empirically based computer model, and it is proving to be outstandingly reliable. Beyond that, our model has received confirmation from globally leading research, through the parallel and independent development of the same concept. We regard this as extraordinary confirmation of our modelling. The model likewise meets the highest banking-regulatory requirements for validation1 with flying colours.
But are validation and confirmation by globally leading science enough? Why should the forecasts be able to be correct? What about the many obstacles to be overcome, such as scarce resources or high investment costs? The plausibility check of the model forecasts turns out to be impressive; real obstacles exist only at the political level.
The empirically based modelling of technology transitions primarily uses the characteristics of successful technology transitions and the steady price reduction of those very technologies that accompanies these transitions. This price reduction in turn supports the further growth of that technology. This powerful relationship fundamentally explains2 the success of technology transitions. Comprehensive scientific analyses confirm this and provide the basis for the modelling used.
But can the energy transition be reliably modelled at all for the next 20 – 25 years? Have past transition models and their forecasts, such as those of the IEA, not shown with their spectacularly wrong failure3 that modelling is not reliable? Can new developments, for example the renaissance of nuclear energy, not reduce any modelling to absurdity? The answer, anticipated here, is clearly no.
What is remarkable about many previous models is that they do not even come close to reflecting the course of the transition so far; that was not even an objective in the development of the models. These models would therefore fail to stand up to the regulatory model validation standards4 from the outset.
The problem of unsuitable transition models has meanwhile found prominent recognition. Among others, the Bank for International Settlement (BIS) calls for epistemologically new approaches in forecasting the transition5. It also calls for new approaches in risk management. The reason for the latter is that traditional risk management systems are fundamentally “backward looking” and therefore cannot recognise emerging new risks by design.
Even before the BIS call, we had set ourselves the goal of developing a new and reliable approach. The empirical evidence of past technology transitions is a good estimator for transitions currently under way. At the beginning a successful technology grows exponentially, up to ~ 50% coverage of the market. This phase is followed by linear growth, finally flattening out and converging towards the market. This course is known as the S-curve. Renewable energies have shown exponential growth for a long time. As an example, the figure below shows the modelling of the trend (up to 2023 the historical development) on the basis of the known characteristics.
1 FED SR 11 – 7, April 4, 2011, https://www.federalreserve.gov/supervisionreg/srletters/sr1107.htm
2 This relationship between volume growth and price degression is observed in highly scalable products.
3 PV Magazine: IEA versus the reality of solar PV, November 20, 2018, https://www.pv-magazine.com/2018/11/20/iea-versus-solar-pv-reality/
4 Federal Reserve System (FED), SR 11-7: Guidance on Model Risk Management, 04.April 2011
Figure #: Wright’s Law for photovoltaics, also known as Swanson’s Law. The chart uses a double logarithmic scale.
Also empirically proven is an almost law-like price behaviour (Wright’s Law) that accompanies the expansion of production. For products that can be standardised (PV panels, flat screens, batteries, smartphones, etc.), costs fall with every doubling of the total volume produced by a technology-specific percentage, the learning rate.
In photovoltaics this regularity has been demonstrated impressively for 48 years. The globally installed capacity is increasing by 36% p.a. and costs fall by 23% with every doubling (learning rate). Since 1976 the costs of PV modules have fallen from $ 100 Wp6 to under $ 0.10 Wp. The chart shows a certain volatility around the trend, which is not particularly surprising and also not of great relevance.
This development applies to all successful7 technologies and thus also to renewable energies and batteries; they are increasingly displacing fossil and nuclear power generation.
The empirical evidence of past technology transitions, both their volume growth and their price degression, is the foundation of our model.
6 For the watt peak (Wp), standardised output under defined parameters
7 We regard technologies as successful if they show a sufficient track record of volume growth and price decline, a sufficient total volume, social acceptance and no known resource constraints.
Stability of the model
Objectives of the model development
An important objective of the model development was and is the stability of the model. A reduced parameter set is a decisive factor in this. Extensive parameter sets and over-complexity fundamentally lead to lower reliability, as is known from research into risk models in the financial sector, among others. We rejected complex socio-economic assumptions for modelling behaviour not only because of the large number of parameters and their partial interdependencies as well as their susceptibility to temporary social effects; the poor performance of these models was the most important aspect in doing without such models.
The characteristics of technology development, such as the relationship between the total volume produced in the energy technologies (PV, wind, batteries, etc.) and the price, Wright’s Law, are by contrast empirically proven and the central model component.
The chosen approach avoids both arbitrary assumptions about the transition and the demonstrably weak expert estimates8 of future technological development.
Particularly worth emphasising is the rising energy efficiency of new technologies, such as the replacement of fossil power generation. On a global average it has an efficiency of only 38%; the rest is lost as waste heat. Every kWh of renewable electricity therefore avoids the consumption of 2.63 kWh of fossil energy. Cars with combustion engines need about 4 x more energy than an electric vehicle, and the efficiency gain from using heat pumps is similarly high. The increase in efficiency contributes decisively to the energy transition and is taken into account accordingly in the modelling.
Technologies that are so far at prototype stage or are used on a small technological scale we have excluded as insufficient, given the order of magnitude required in the energy sector. This concerns technologies such as carbon capture and storage (CCS). Realistically, in the short time up to 2045 they cannot achieve a material share of the energy shift, even with exponential growth at realistic rates.
Furthermore, neither the CO2 prices for emission allowances nor the costs of the damage from the physical effects of global warming were taken into account. CO2 emission allowances are highly heterogeneous globally. There are still too many exemptions and free allocations of certificates; as a result, even within individual countries such as Germany no uniform cross-sector prices can be determined, and no global effect on the energy transition can be established so far.
Restrictions on consumption behaviour such as going without were not taken into account. As the past has shown, warnings about general risks remain almost ineffective. This has been shown both by the Club of Rome with its impressive warnings and by the continuing warnings about advancing climate warming; despite very high prominence, the warnings have not translated into consistent changes in behaviour.
The chart below shows the expected transition as it results from the empirical modelling outlined above.
Total energy consumption – by source – to 2050
The Corona break in 2020 is clearly visible. Striking is the decline in total energy consumption up to ~ 2033. What at first seems surprising is largely explained by the increase in efficiency in the energy technologies. The efficiency gains will continue after 2033 as well, but from about 2033 the growth in energy consumption asserts itself again. The additional demand is then covered by the growth in renewable energies.
The model also shows the possibilities and limits of hydrogen production well. Despite exponential growth up to about 2043, hydrogen will primarily only be able to be available for applications that cannot be electrified. These are primarily the petrochemical industry (production of plastics and lubricants) and the fertiliser industry. Fuels for long-haul flights and shipping transport are hardly material. In road traffic there are, on the one hand, alternatives that are considerably more energy-efficient than hydrogen propulsion. On the other, this development would come far too late; electrification will have been completed by then.
About 48% of oil consumption is attributable to road traffic; this share will be essentially replaced by electricity (BEVs – battery electric vehicles) by around 2042. It is also striking that green hydrogen (produced with the help of renewable energies) and synthetic hydrocarbons (dotted area within the solar share) are virtually invisible before 2034. From about 2034 it will probably have reached break-even with fossil fuels, but its share of energy supply will by then amount to considerably less than 1%. In terms of price and volume this is a striking analogy to PV in 2022.
Further technical progress and growth in transmission and storage technology support the energy transition. The collapse in battery prices in 2024 to partly under $ 60/ kWh underlines the trend. For efficiency reasons, coal-fired power generation will be displaced worldwide by around 2033 by renewable energies and, to a small extent, by the more flexible gas peakers (peak-load power plants). Because of the short operating times of the gas peakers, these will also cause only low CO2 emissions. The use of natural gas in the heating sector will increasingly be reduced by electrically powered heat pumps.
We have assumed nuclear energy to be constant on the basis of the trend. In recent decades there has been hardly any appreciable change in the globally installed power plant capacity. On the one hand, the extreme costs of new nuclear power plants stand in the way of development, and on the other, new nuclear power plants cannot be available in time if modern safety standards are complied with.
Current capacities in CCS are, on the one hand, very small and, on the other, the growth rate is likewise insignificant. CCU (carbon capture and usage) may have a future as a carbon source for synthetic fuels, but this cannot be reliably modelled owing to a lack of historical data. CCU will not have a material influence on the CO2 content of the atmosphere by 2050.
The statements above concern the probable, the expected course of the transition. Deviations from the forecast course are possible; clear deviations, particularly over longer periods, are however unlikely, as the empirical evidence impressively shows. Nevertheless, we point to the drastic collapse in prices for PV modules in 2023/2024 as well as in the battery sector in 2024. Even if a market shake-out is to be expected, particularly in PV module production, the current collapse in prices may bring about an acceleration of the transition. For the possible consequences we refer you to the section on risks and to risk management.
8 Jing Meng, Rupert Way, Elena Verdolini, and Laura Diaz Anadon. Comparing expert elicitation and model-based probabilistic technology cost forecasts for the energy transition. Proceedings of the National Academy of Sciences, 118(27), 2021. ISSN 0027- 8424. doi: 10.1073/pnas.1917165118. URL https://www.pnas.org/content/118/27/ e1917165118.
Correct basis, better results
Costs of the transition
The supposedly high costs of the transition have been cited so often that for that reason alone they are almost beyond doubt. But anecdotal evidence is a very poor guide. The argument was always made only on the basis of current costs, while the further strong cost degression (Wright’s Law) of renewable energies was left out. Meanwhile new large-scale renewable plants are the cheapest source of electricity, and the further cost degression will extend the advantage of renewables. Even the direct costs of fossil fuels are consistently higher than the costs of renewables. The empirical evidence points to a further rapid fall in the cost of renewables. Grid expansion could therefore at most cause considerable costs. Against this stand the drastic growth and the equally drastic cost reductions in batteries. Smart integration into the distribution grids fundamentally makes an expansion of the distribution grid unnecessary. The transition can therefore hardly be more expensive than the further unrestricted use of fossil fuels. In addition, with renewables neither the further processing costs of fossil fuels nor the environmental damage caused by exhaust gases have to be taken into account in order to make the economic advantage clear.
According to the modelling, the costs of the transition are about 25% below the direct expenditure on fossil fuels up to 2045. The time horizon of 2045 is still heavily restricted here. In the no-transition case the costs for fossil fuels would continue to arise every year. Those for renewable energies fall significantly, since investments in renewables are fundamentally limited to replacement after the transition. Expenditure on fossil fuels amounted to about $ 5 trillion in 2022. Added to these costs are, among other things, the subsidies quantified by the International Monetary Fund (IMF), 7 trillion for 2022 alone. Of this, more than one trillion is explicit direct subsidy, which would have to be added to the expenditure on fossil fuels. This clearly underlines the advantage of the transition. We have refrained from including the processing costs of fossil fuels.
Finally, we point out that we have not included the expenditure on the infrastructure expansion of the transmission and distribution grid in the modelling. The allegedly high costs of infrastructure expansion are repeated many times without being questioned. These cost estimates were, however, drawn up without the high potential of decentralised energy storage in batteries. Infrastructure expansion can therefore fundamentally be limited to a few transmission corridors. In distribution grids and even in transmission grids, electricity transport will be shifted in time through the use of batteries. In low-load periods (at night and between the peaks) extensive capacity is available; over the course of the day (24h) the grids are only used to a small extent. We have taken the expansion of battery capacity into account in the modelling. Set against the investments in the expansion of the transmission grids during the transition are primarily the subsidies for fossil fuels, the processing costs of fossil fuels that have not been taken into account, and the investments in the expansion of fossil energy infrastructure in the no-transition scenario. We have so far refrained from a precise comparison, as they are of a smaller order of magnitude.
Expected consumption of fossil energy
CO2 emissions – the Paris climate target
From the modelling, the expected direct CO2 emissions from the consumption of fossil fuels up to 2050 can be derived. Empirically they add up to about 335 Gt, which makes meeting the Paris climate target of 1.5° appear possible. This does not, however, take into account the current development, the globally warmest months since records began and the warming of the earth that may thus be accelerating. Contributing to this acceleration are, among other things, CO2 emissions from forest fires or the effects of rising methane emissions from thawing permafrost soils in the Arctic or further effects (increased microbial activity due to global warming).
This is by no means intended to say that we consider the problem of global warming to be solved. But the modelling expresses that anthropogenic CO2 emissions can with high probability be reduced to the pre-industrial level and that this transition is even economically attractive. The results of the modelling are an additional incentive not to slacken in the efforts to achieve the energy transition.
Expected consumption of fossil energy
Validity of the model and plausibility check of the results – model validation
Many models have drastically underestimated the speed of the transition, and almost all forecasts of the cost reduction of renewable energies, particularly of PV, were even spectacularly wrong 9&10 .
But why should our modelling be better? Especially as the results appear too optimistic at first sight. That is exactly what we asked ourselves, and below we have analysed all the counter-arguments known to us. The best-known problems to be solved are energy storage, the availability of the raw materials required (catchword: rare earths), available land for the use of renewable energies, the possible resistance of the fossil energy sector, other energy technologies (e.g. nuclear energy, fossil fuels with CCS), socio-economic aspects, sufficient electricity availability for e-mobility or heat pumps, the expansion of the energy infrastructure, the time to net zero and, not least, the costs of the transition.
The costs have already been dealt with above. On the argument of scarce raw materials, let us simply take the example of lithium and the frequently cited scarcity: the lithium reserves known worldwide are sufficient to electrify the car fleet 8 x over, and the lithium can be recovered from old batteries. In addition, with every new generation of batteries less lithium is needed to achieve the same battery capacity. There can be no question of scarcity here. This applies in principle to the other raw materials as well.
On the counter-arguments to the energy transition let us only briefly add: in principle they can all be refuted. To date we have not found a single valid reason apart from politics that could delay the transition.
9 Marc Jaxa-Rozen and Evelina Trutnevyte. Sources of uncertainty in long-term global scenarios of solar photovoltaic technology. Nature Climate Change, 11(3):266–273, Mar 2021. ISSN 1758-6798. doi: 10.1038/s41558-021-00998-8. URL https://doi.org/10. 1038/s41558-021-00998-8
10 C. Wilson, A. Grubler, N. Bauer, V. Krey, and K. Riahi. Future capacity growth of energy technologies: are scenarios consistent with historical evidence? Climatic Change, 118(2): 381–395, May 2013. ISSN 1573-1480. doi: 10.1007/s10584-012-0618-y. URL https: //doi.org/10.1007/s10584-012-0618-y.
Management of model risk
Model validation – formal aspects
Validation standards would increase model quality and confidence in the modelling. In the banking sector there are exactly these high requirements that models have to meet before they are recognised by the supervisory authority. Our model meets the leading banking-regulatory requirements for model validation with flying colours. Beyond that, leading research has taken up the same approach in parallel and has thereby underlined the model approach.
The use of inappropriate models entails considerable risks; not using models entails greater risks. An initial plausibility check already gives a first indication of the modelling; gross errors in the assumptions, the data or the modelling would already show after 6 years of use. The model risks cannot, however, be adequately captured and managed with a simple plausibility check. For this reason the FED11 has drawn up a comprehensive catalogue of requirements for the management of model risks with SR 11 – 7. We have imposed these leading requirements on ourselves as a benchmark. Our model meets the banking-regulatory requirements of SR 11 – 7 in outstanding fashion.
Among other things, the forecasts of growth rates and cost degression drawn up at the beginning of 2018 have been impressively confirmed by actual developments over the six years of use to date. Even the slight decline during the Corona crisis was subsequently made up again. The expansion of renewable energies and the cost degression are proceeding as forecast. This confirms the development of renewable energies into the cheapest source of energy.
In addition to the formal validation, the model concept has gained scientific recognition through parallel and independent development by a leading university. The scientific publications produced by Oxford University in this context12 have underpinned the significance and reliability of empirical evidence through comprehensive research and thus likewise confirm the model concept.
11 Federal Reserve (FED): https://www.federalreserve.gov/supervisionreg/srletters/sr1107.htm
12 https://www.sciencedirect.com/science/article/pii/S254243512200410X
We develop forward-looking solution concepts
Confirmation by the market
Financial markets are anticipatory; they often anticipate developments. To what extent can the modelling be observed on the stock markets? The chart below impressively shows the weak performance of the oil market compared with the S&P 500 over the past 10 years. It is surprising that even the high energy prices since 2021 and the extraordinary performance of the oil and gas industry since then have not led to a lasting improvement in share performance.
Management of model risk
Reliability compared with other models and expert estimates
The most widespread models in the energy transition are probably the integrated assessment models (IAM). IAMs are intended to take into account policy-relevant and socio-economic developments relating to global warming, taking account of the current economic situation and its development over a time frame extending beyond 2050. The aim is to answer questions of global environmental change and sustainable development by providing a quantitative description of the most important processes in the human and earth systems and their interactions.
The models attempt to model politics, and with it the political steering of the economy, over a period of more than 25 years. How hopeless this undertaking is has been impressively demonstrated by US politics in recent years with the withdrawal from, the re-entry into and the probable renewed withdrawal from the Paris climate agreement.
A further serious weak point of the IAMs is the failure to take technological development into account. These models always start from current costs. The drastic fall in the cost of renewable energies, the rising efficiency that accompanies the expansion of renewables and the lower (primary) energy consumption induced by it are not taken into account. Because of the constantly accelerating technological development, this model error is constantly becoming more severe.
An appraisal of the IAMs is already superfluous because of the long-term objective of these models. Especially as the question of the reliability of the IAMs arises when economic forecasts by prominent institutions (universities, central banks, the World Bank) hardly go beyond a time frame of three years.
Expert estimates
Expert estimates of the transition have likewise proved to be decidedly weak. This is not meant to belittle the experts' knowledge in their specialist field. Comprehensive analyses13 document, however, the poor ability to assess future developments correctly. As a rule, experts can hardly forecast the speed and extent of the transition, nor the cost degression of renewable energies, adequately. Regarding the cost degression of renewables, lower limits below which costs cannot fall are frequently postulated. These have proved to be fundamentally untenable; costs almost always fell below them after a short time through actual developments. That is the primary reason why we did not consider taking expert estimates into account in our model.
Incorrect forecasts
The list of incorrect forecasts of the energy transition is long, and the forecasts were in some cases blatantly wrong. A forecast by the International Energy Agency (IEA) from the 1990s estimated that the share of renewable energies in global electricity consumption would be only 6 % by 2020. In 2022 more than 29 % of global electricity generation came from renewable sources.
The examples could be continued. Almost all forecasts of the growth of renewable energies were overtaken by reality. The error lies consistently in assuming a linear growth trend. Only in the first three years is faster growth assumed; after that it is typically assumed to be linear.
Conclusion
In its book The green swan the BIS has called for new epistemological approaches. Our modelling has shown for six years now that it meets the BIS's demands. The modelling thus also offers a reliable starting point for managing the risks that come with the transition. This includes in particular the possibility of adapting the traditionally backward-looking financial risk models. The BIS's second demand.
13 Jing Meng, Rupert Way, Elena Verdolini, and Laura Diaz Anadon. Comparing expert elicitation and model-based probabilistic technology cost forecasts for the energy transition. Proceedings of the National Academy of Sciences, 118(27), 2021. ISSN 0027- 8424. doi: 10.1073/pnas.1917165118. URL https://www.pnas.org/content/118/27/ e1917165118
Contact
We are delighted to help you identify and seize the opportunities presented by the transition, and to balance these opportunities against the risks.
