REMod pathways study from Fraunhofer ISE, Germany

Please note: 1. The english translations given here, including the name of the study, are unofficial.  2. An official english version of the main report is in progress.  3. None of the published material to date is open licensed — which means that diagrams and tables cannot be extracted and presented on a public forum such as this.


This post covers the published study “Pathways to a climate-neutral energy system — the German energy system transformation in the context of societal behavior” undertaken by Fraunhofer ISE (Sterchele et al 2020a). The study results were made public on 13 February 2020 and I attended the official presentation in Berlin, Germany.

Some background for readers not familiar with research institutes in Germany. The Fraunhofer ISE specializes in solar energy systems. It is located in Frieburg and is one of 72 Fraunhofer research institutes and research units that make up the Fraunhofer Society. Much of the academic research in Germany is structured this way with other notable associations being Helmholtz, Leibniz, and Max Planck.

The reported study investigates development pathways (sometimes known as trajectories) for the German energy system that reduce energy‑related CO2 emissions by variously 95% and 100% by 2050 (with one further 100% scenario targeting 2035). Achieving these goals for the energy sector is entirely feasible from a technical and systems perspective.

Notwithstanding, aggregate societal behavior is now a decisive factor. Detrimental behavior can restrict the technical solution space, exclude otherwise attractive pathways, increase cumulative expenditures, and impact negatively on other high‑level metrics. Whereas better aligned behavior can lead to considerably smaller and cheaper systems, with fewer downside consequences. Indeed, the German energiewende should be seen as much a social transformation as a technical transformation.

Social behavior can be made endogenous in energy system models with varying degrees of success (I have worked on three such attempts). The Fraunhofer ISE study instead introduces such behavior exogenously via scenarios that attempt to capture various aspects of societal behavior and attitudes. This study puts societal responses at the center of the analysis and that is a first to my knowledge.


Four main scenarios (table below) seek to capture various societal and political backstories and are evaluated using the objective of 95% decarbonization by 2050.

Scenario Translation Comment
Beharrung inertia strong resistance to the use of new technologies in the private realm
Inakzeptanz resistant strong resistance to the expansion of large infrastructure
Suffizienz sufficient marked changes in behavior that significantly reduce energy consumption
Referenz reference no additional boundary conditions that promote or impede the overarching objective

Two secondary scenarios (table below) are also reported with their key properties as indicated. In the report, these are further qualified, for example, Referenz_100 becomes the reference plotline for 100% decarbonization and Suffizienz_2035 becomes the plotline supporting very rapid decarbonization.

Target Target year Comment
95% 2050 bulk of report
100% 2050 German government policy as of 18 December 2019
100% 2035 unrealistically high reliance on imported fuels

Unfortunately the only mainstream 100% scenario reported is Referenz_100. Given the new official policy on decarbonization (see next section), hopefully future work will report on the “remaining” 100% scenarios.

A target year of 2035 is advocated by the Fridays for Future climate protest movement inspired by Greta Thunberg.

Official climate protection targets

At the beginning of the study, official German policy was 80–95% decarbonization by 2050, relative to 1990 levels (BMWi and BMU 2010).

Toward the end of the study, the German parliament passed the Federal Climate Protection Act (KSG) which shifted the 2050 target to 100% net reduction (German Parliament 2019). The new target entered into force on 18 December 2019, as follows (my translation):

[This Act] is based … on the commitment of the Federal Republic of Germany at the United Nations Climate Change Summit in New York on 23 September 2019 to pursue greenhouse gas neutrality by 2050 as a long-term goal. (§1.3)

REMod model

The energy system model REMod — Regenerative Energien Modell — developed at Fraunhofer ISE, was used for the simulation and optimization of the scenarios. Previously, the model name had an appended “D” to indicates its scope was Germany: REMod‑D. The model is closed source and programmed in the Pascal language (which later became Delphi when Windows GUI programming support was added). The model is driven by the nonlinear, nonconvex, continuous system, heuristic solver CMA‑ES (covariance matrix adaptation evolution strategy), which makes it somewhat different from model frameworks that employ linear and mixed‑integer programming instead. Erlach et al (2018) describe the software in some detail.

REMod supports an hourly resolution but does not model electricity and gas transmission or AC load flow. The model seeks a pathway that offers the least financial cost amortized over the selected horizon.

REMod supports roll‑out constraints for technologies and mitigation measures more generally and learning curves as those technologies deploy (but neither feature yet for social change :). REMod also supports fleet vintage but does not actively retire or strand capital or compensate owners adversely affected by shifts in public policy.

REMod contains a differenciated building stock database and is able to apply quite well resolved building thermal performance retrofit measures in an optimized context. This level of detail is relatively rare in other modeling frameworks.

REMod supports international trade. Scenarios that rely on imported green liquid fuels and biomass to any degree are often contentious.

The OpenEnergy Platform provides a factsheet on REMod‑D.


For each scenario, the study reports the heuristically‑optimal pathway identified, via the following:

  • annual evolution of the technology mix (structure) and supply mix (operations)
  • annual evolution of energy‑related CO2 emissions, both per sector and total

And the following metrics at points along the entire transformation:

  • cumulative expenditures
  • estimated marginal cost of avoided CO2

In addition, various sector‑specific themes are highlighted, including the form, role, and propagation rate of contributions from:

  • built environment mitigation — including existing building stock retrofit
  • transformation of the traffic sector
  • sector‑coupling, embedded storage, and other flexibility measures
  • more specifically, green hydrogen from electrolysis

The model is run twice: first to determine the structural evolution of the system in question, and second, with that structure specified exogenously, to determine the costs of avoided CO2 over time.


It is hard to discuss exemplary results without being able to reuse the plots and tables contained in publications. The reader is instead referred to the original report (Sterchele et al 2020a) and supporting document (Sterchele et al 2020b).

Nor am I dwelling much on results because the 95% target that forms the bulk of this study is no longer official policy, neither is it especially relevant given full decarbonization was always the goal. Moreover, a 95% optimal system is unlikely to be a waypoint en route to a 100% optimal system.

The only scenario (2035 aside) also reported with 100% decarbonization is the reference scenario. Of note is the relatively small difference in cumulative expenditure between the two, some 30% more for full decarbonization (figure 28) — and certainly not nearly as much as some people opine.

What is rather evident however, is how expensive it is to work around societal inertia and resistance compared with acceptant change. Multipliers are in the order of 320% to 460% — or, in broad terms, an energiewende three or four‑fold more costly than it need be (figure 28). With the proviso that this result needs to be confirmed for 100% decarbonization.

The 2035 scenario resulted in very high reliance of imported biofuels, particularly green hydrogen. Yet it is difficult to imagine that this level of trade would eventuate in practice, particularly in a world where hopefully all countries are seeking to rapidly decarbonize their own economies.


This is a solid study, with clear research questions, understandable scenarios, well presented results, and readily accessible conclusions, both aggregate and specific. I guess a few points stand out for me:

  • societal aspects can be included in analysis without having to characterize and calibrate the underlying individual and social dynamics, whether that be embedding attitudes, replicating decisions, propagating influence, or representing other causal processes
  • with the change in official government policy to 100% decarbonization, the “missing” 100% scenarios need analyzing
  • the difference in system characteristics between 95% and 100% decarbonization (for just the reference scenario) is not especially notable
  • but the disparity between societal inertia and resistance and societal engagement is very significant
  • transitions to zero‑carbon more rapid than 2050 will require that a social tipping point be crossed
  • one model cannot do everything — and REMod is a good blend of temporal resolution, sector scope, and network simplicity for this kind of analytical role — that said, it remains to be seen how models that also embed AC load flow and span the European electricity grid perform in this space in the future
  • the German energiewende is entirely achievable within the next three decades — whereas the consequences of not being carbon neutral within this timeframe are barely worth contemplating

The usual caveats about relying on closed models to inform public policy apply to this study. The practice of using closed policy analysis may well persist in Germany, but the European Commission is increasingly seeking policy transparency and that the trend will doubtless continue until “open by default” becomes the new benchmark for public policy analysis.

As always, please contact me directly if errors need correcting — otherwise feel free to comment below.


Erlach, Berit, Hans-Martin Henning, Christoph Kost, Andreas Palzer, and Cyril Stephanos (April 2018). Optimization model REMod-D: Materialien zur Analyse Sektorkopplung — Untersuchungen und Überlegungen zur Entwicklung eines integrierten Energiesystems [Optimization model REMod-D: materials for the sector coupling analysis: investigations and considerations for the development of an integrated energy system] (in German). Germany: acatech, Leopoldina, Akademienunion.

Federal Ministry of Economics and Technology (BMWi) and Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (BMU) (28 September 2010). Energy concept for an environmentally sound, reliable and affordable energy supply. Berlin, Germany: Federal Ministry of Economics and Technology (BMWi). Superseded policy. Archive copy. Policy database.

German Parliament (December 2019). Bundes-Klimaschutzgesetz (KSG) [German Federal Climate Protection Act (KSG)] (in German). Adopted on 12 December 2019 and entered into force on 18 December 2019. PDF.

Sterchele, Philip, Julian Brandes, Judith Heilig, Danien Wrede, Christoph Kost, Thomas Schlegl, Andreas Bett, and Hans-Martin Henning (February 2020). Wege zu einem Klimaneutralen Energiesysem: Die deutsche Energiewende im Kontext gesellschaftlicher Verhaltensweisen [Pathways to a climate-neutral energy system: the German energy system transformation in the context of societal behavior] (in German). Freiburg, Germany: Fraunhofer ISE.

Sterchele, Philip, Julian Brandes, Judith Heilig, Danien Wrede, Christoph Kost, Thomas Schlegl, Andreas Bett, and Hans-Martin Henning (February 2020b). Wege zu einem Klimaneutralen Energiesysem: Die deutsche Energiewende im Kontext gesellschaftlicher Verhaltensweisen — Anhang zur Studie [Pathways to a climate-neutral energy system: the German energy system transformation in the context of societal behavior — Annex to the study] (in German). Freiburg, Germany: Fraunhofer ISE. Supplementary material.

Sterchele, Philip (2019). Analysis of technology options to balance power generation from variable renewable energy: case study for the German energy system with the sector coupling model REMod. Düren, Germany: Shaker.

Note on open licensing

Occasionally I write for Wikipedia and my mailbox is full of traffic requesting that third‑party copyright holders reissue images — typically screenshots, plots, and diagrams — under Wikipedia‑compatible open licenses so I can add them to articles. In rare cases, I even adapt and redraw key diagrams to escape copyright. These various options are clearly time consuming and I don’t have the inclination to do so in this case. Instead I have asked four of the eight study authors and an institute press officer to consider reissuing their two PDFs under Creative Commons Attribution CC‑BY‑4.0 licenses in order to advance open science and to also allow reuse here. That request is only a few days old and I remain hopeful of a satisfactory outcome.


Some recent tweet streams related to this study:

One tweeter interprets “Beharrung” as “Private sector resists new tech”. That is not correct. I used “private realm” instead to stay close to the original text, but “households” or “public” might have been better choices. The report (§2.3.2) also addresses the “Bevölkerung” or “general public”.

Scenario benchmarking

@tom_brown suggested by direct message that I clarify the following paragraph in my original post:

What is rather evident however, is how expensive it is to work around societal inertia and resistance compared with acceptant change. Multipliers are in the order of 320% to 460% — or, in broad terms, an energiewende three or four‑fold more costly than it need be (figure 28). With the proviso that this result needs to be confirmed for 100% decarbonization.

Indeed, this is what @tom_brown wrote to me:

Maybe your summary could be clearer that the 320–460% multiplier is on the difference to the business‑as‑usual scenario, not on the total costs?

The costs recorded in figure 28 (page 55) from the Fraunhofer ISE report are offset from a business‑as‑usual (BAU) case — which should not, in my view, be described as a scenario for reasons that will become apparent. The BAU case effectively zeros the y‑axis — like setting the sea level of a floating iceberg of unknown depth.

Specifically under the BAU case: no new policy measures are applied, no assumptions on lack of public acceptance or motivated social change are present, and no decarbonization target is sought. In other words, close or identical to the two Referenz scenarios but lacking any ambition on climate protection. Under BAU, plant replacement and technological learning will still occur naturally as lifespans are reached, but dispatch and capital are allocated in the absence of decarbonization policy measures (such as carbon pricing), explicit upper bounded decarbonization trajectories (intermediate goals), an end‑of‑horizon decarbonization goal (the carbon‑zero target year), or some mix thereof. Social learning is not present either — a point I will return to.

Personally I would be more inclined to label this particular contingency as dereliction‑as‑usual (DAU). But terminology aside, the report fails to state the cumulative expenditures that result from BAU, nor what level of decarbonization, if any, eventuates.

The analytical role of BAU is to benchmark the various selected scenarios that do necessarily meet stated decarbonization targets: in this study 95% and 100% in 2050 (along with one further scenario targeting 2035) — while noting that 100% decarbonization across all sectors is now official policy in Germany. The BAU approach allows the results to be reported by difference, but absolute costs cannot be determined here because, as noted, the performance metrics of the BAU case are not provided. Nor should the BAU case be considered a scenario itself because it fails to provide a policy admissible, let alone remotely humane, future (refer leaked JP Morgan report covered below). In contrast, other projects do provide percentage‑wise differences against their BAU cases — thereby enabling astute readers to reverse engineer the absolute values.

The BAU case method allows the marginal abatement cost of decarbonization to be estimated for each true scenario at various points in time — this probably being its most compelling defense. Assuming that is, that one accepts that diffing against some arbitrary policy frozen world does indeed provide useful information.

To expand, it is a moot question whether reporting benchmarked differences against a BAU case that, in all likelihood, fails to confront catastrophic climate change, is indeed useful. At the very least, the financial and emissions characteristics of the BAU benchmark should be reported so readers can better understand the interpreted context of the various scenarios. But I can equally understand why report authors would opt not to provide those values, because the first and only figure the media will pick up is the rather arbitrary, if not entirely spurious, BAU‑plus‑scenario financial cost of the favored transition and publicize this as the price of change.

To return to my earlier comment about implied frozen social learning in the BAU case. Current and future governments can, in theory at least, elect to embargo all new climate protection policies (and one can look to United States at present for an example). But current and future governments cannot mandate that social attitudes and behaviors remain fixed. So although social learning is difficult to describe and extrapolate, social progress nonetheless exists and evolves. Indeed, social progress may be the most cogent factor in future responses to the climate emergency — something that the Fraunhofer ISE study under discussion seeks to highlight and explore.

On a philosophical note, ceteris paribus analysis (CP) — long favored by economists — and complex adaptive systems (CAS) studies do not fundamentally mix well. Particular when the selected CP benchmark is so detached from any feasible reality, as is the case here with BAU. Whereas comparative scenario analysis, made against one or more elected baseline scenarios, is well supported by the CAS paradigm. Indeed, considering any costs beneath the cheapest identified policy‑conforming scenario may well be akin to engaging in the sunk cost fallacy.

My original paragraph was not strictly wrong but neither did it portray the entire story. Nor can that story be fully determined on the information made available. But I am inclined nonetheless to use the least financially expensive, goal‑compliant, identified, feasible scenario (note: scenario not base case) as the baseline (note: baseline not benchmark) to compare options and enable trade‑offs. While acknowledging that my original statement did rely on a model‑determined BAU case to allow the calculations presented to proceed.

To close, benchmarking against some arbitrary and clearly unrealistic counterfactual has many limitations. Just maybe governments and the public should get on with very rapid decarbonization without the need to convince ourselves of the relative merits of doing so using questionable forms of analysis. The various scenarios in the Fraunhofer ISE study are — stand alone — completely legitimate in their own right and perhaps that, and their comparative merits, is the message that analysts should focus in on and promote.

As always, contact me directly for corrections of fact. Otherwise discuss below.

IEA analysis

The 2019 IEA World Energy Outlook adopts a somewhat similar counterfactual approach to that used by Fraunhofer ISE. That report contains a current policies scenario — their terminology — that shows what might happen if the world continues along its present path, without any additional changes in policy. The IEA opines (IEA 2019:1) (emphasis added):

In this scenario, energy demand rises by 1.3% each year to 2040, with increasing demand for energy services unrestrained by further efforts to improve efficiency. While this is well below the remarkable 2.3% growth seen in 2018, it would result in a relentless upward march in energy‑related emissions, as well as growing strains on almost all aspects of energy security.

I don’t know if and how the other scenarios the IEA consider are compared against this scenario because the executive summary doesn’t traverse the question and the full report, at €120, lies well beyond my means.

Leaked JP Morgan report

Regarding the dereliction‑as‑usual case, a leaked internal report from two JP Morgan investment bank economists, David Mackie and Jessica Murray, dated 14 January 2020, was recently reported in The Guardian (Greenfield and Watts 2020). Their report states that under our current unsustainable trajectory “we cannot rule out catastrophic outcomes where human life as we know it is threatened”. JP Morgan subsequently distanced itself from the content of the report.

I don’t suppose even JP Morgan would attempt to cost the failure of humanity. But another decade of inaction and some transgressed earth system tipping point and we (but not especially me, given my age) may well be discovering that price first hand.

Non‑model considerations

Also worth noting is that the assessment of scenarios should be augmented by non‑model information to round out that assessment, including the co‑benefits and co‑disbenefits not represented in the model. Whether these additional factors can and should be monetized or not is dependent on the purpose of the analysis and whether doing so is indeed legitimate.


Greenfield, Patrick and Jonathan Watts (21 February 2020). “JP Morgan economists warn climate crisis is threat to human race”. The Guardian. London, United Kingdom. ISSN 0261-3077.

IEA (November 2019). World energy outlook — Executive summary. Paris, France: IEA Publications. Open access. Landing page.

Text and images licensed under CC BY 4.0Data licensed under CC0 1.0Code licensed under MITSite terms of serviceOpenmod mailing list.