We divide the results in two sections: first, we show the capital cost projections generated by the three IAMs with ITC. Next, we present global results for all scenarios until 2050 derived from TIAM-ECN. We treat the perfect spill-over (OPS) scenarios separately, as variations of the OPT scenarios. We focus on the expansion of the 5 key technology clusters, and we discuss the impact of different R&D and climate policy set-ups on energy system costs.
3.1. Impact of R&D on Capital Costs
In Figure 1 we show capital cost reduction projections relative to 2020 per scenario and technology group as box plots. We present these results for the years 2030 and 2050, for the REF and CB710_OPT scenarios, i.e. respectively the most and least conservative scenarios with regard to R&D policy ambitions. The yellow box plots show results for fossil- and biomass-based CCS technologies (panels a to d), the blue box plots show results for variable renewable electricity (VRE), namely onshore and offshore wind (panels e and f) and solar PV and, in the case of WITCH, CSP (panels g and h). Red box plots show results for the remaining technology groups of advanced fuels (aggregating advanced routes for synthetic fuels and biofuels generation, such as Fischer-Tropsch, as well as hydrogen production from advanced technologies such as biomass gasification and electrolysis, panels i and j) and batteries for passenger and, in the case of WITCH, freight EVs (panels k and l). For the full set of results, including all time periods and scenarios and a detailed list of technologies per technology group, see the SI.
The size of the box plots in Figure 1 is determined by the number of energy conversion technologies representing each technology cluster, and the number of regions where these technologies are implemented in each model, which results in different ranges of capital cost reductions. Means, medians, first and third quartiles are depicted as x’s, horizontal lines inside the boxes, lower limits and higher limits of the boxes, respectively. The whiskers below and above the boxplots indicate the lowest and highest quartiles of each group.
Results in Figure 1 show that capital costs may vary significantly depending on the model and technology group in both reported years despite the harmonization of knowledge-related parameters. While, in general, cost reductions are expected to be more pronounced in scenarios that optimize R&D policy (CB) than in REF, this pattern does not hold for some technology clusters for GEM-E3 and MERGE-ETL. This can be seen, for instance, in the fossil-based CCS results from MERGE-ETL (panels a and b) and the electric vehicles results from GEM-E3 (panels k and l). In the case of MERGE-ETL, this effect is due to R&D investments being allocated per technology component. Different components might benefit to a different extent from R&D-driven cost reductions, depending on the stringency of climate policies in each scenario. For CCS, for instance, R&D investments in the gasification component of coal power plants and in natural gas combined cycle turbines drive down capital costs of coal and natural gas power plants (panels a and b) in the REF scenario. On the other hand, R&D investments are shifted to the CO2 capture component (CO2 scrubbers) in CB scenarios, thus favouring capital cost reductions of biomass-based CCS technologies (panels c and d). In the case of GEM-E3, modest cost reductions in CB scenarios are caused by costs other than equipment, such as increased labour. This effect illustrates that macroeconomic implications of low carbon policies can in some cases offset the expected R&D-induced capital cost reductions related to equipment.
Capital costs of CCS technologies have different cost reduction profiles across models, with an average cost reduction no larger than 20% in the case of fossil-based CCS (see W_CB710_OPT scenario results in panels a and b of Figure 1) and a maximum of 40% average cost reduction of biomass-based CCS (see 2050 result for W_CB710_OPT scenario, panel d). Solar-based technologies display the largest cost reductions: mean values are around 70% in 2050 in both W_REF and W_CB710_OPT scenarios (panel f), and already between 50% and 60% in 2030 (see panel e). Similarly, wind energy technologies show a steeper cost reduction in scenarios derived from WITCH than from the other two models (panels g and h). Largest cost reductions, reflected by the first quartile and median observed at similar levels in W_CB710_OPT in 2050, reach almost 60%, reflecting the steep cost reductions foreseen for offshore wind technologies, and less than 40% in other models’ results, which present more conservative and aggregated cost reductions.
Regarding advanced fuels, the most pronounced cost reductions are observed in CB710_OPT scenarios in 2050: median values in WITCH reach 60% and, in GEM-E3, 30% (panel i). Besides these outcomes, capital cost reductions are modest or absent, especially in 2030, indicating that these technologies might need a longer development time to benefit from R&D investments. Finally, panels k and l in Figure 1 show that electric vehicles are the least affected by R&D and climate policy packages, as they display the smallest cost reductions, with ranges in CB710_OPT scenarios that are similar or more modest than in REF. This indicates that factors other than combined R&D and stringent climate policies drive capital cost reductions for EVs. Moreover, no cost reductions for this technology group are observed for MERGE-ETL because this model does not include R&D for EV batteries.
3.2. Impact of R&D on the Energy System: Technology Diffusion and Costs
With the TIAM-ECN model, we assess the impact of the different capital cost paths on the development of the global energy system up to 2050. We present results for all scenarios in Table 1, but we treat the perfect spill-over scenarios (OPS) as a variation of the OPT scenarios, thus reporting OPS always in comparison to OPT for each technology cluster.
In Figure 2, we show TIAM-ECN projections for global final energy consumption (FEC) per energy carrier. Each of the three panels corresponds to results obtained with TIAM-ECN using capital cost reductions derived from WITCH (a), MERGE-ETL (b) and GEM-E3 (c). Each bar in the chart corresponds to a specific combination of scenario and time period. Figure 2 shows that the overall trends are similar across the three panels. Total FEC grows by about 40% between 2020 and 2050 in REF, while its growth in CB scenarios is less pronounced due to climate policies triggering the deployment of high-efficiency technologies. All CB scenarios have higher consumption of electricity, biomass and hydrogen hand-in-hand with lower fossil fuel consumption compared to the corresponding REF in 2050. TIAM-ECN’s CB scenarios derived from WITCH (panel a) has the largest electrification level, which relates to the steeper capital cost reduction profiles derived from this model for CCS and VRE technologies, as shown in Figure 1. In comparison, TIAM -ECN scenarios with MERGE-ETL’s and GEM-E3’s costs (panels b and c, respectively) present a smaller increase in electricity consumption and a larger consumption of biomass and hydrogen. Moreover, by comparing results from OPT and FIX scenarios in Figure 2, one can note that the different R&D assumptions in these two group of scenarios do not lead to observable differences in the FEC composition until 2050. This indicates that the energy transition is more influenced by the stringency of climate policy than by different R&D frameworks.
Figure 3 shows TIAM-ECN projections for total installed capacity of power plants with CCS from fossil fuels (first row) and from biomass (second row) until 2050. Scenarios derived from MERGE-ETL incorporate cost reductions from CCS in advanced fuel technologies as well, however these are allocated in results for this specific technology group (see SI for an overview of CO2 removal per technology group in each scenario). Each line represents the yearly total installed capacity in a specific scenario. Scenarios in which capital cost reductions are derived from WITCH, MERGE-ETL and GEM-E3 results are presented respectively in shades of blue (panels a and b), orange (c and d) and green (e and f) – this colour convention is consistently applied in all line-plots in this section. REF scenarios are plotted as solid lines with empty squares. Dark shaded lines with full diamonds and light shaded lines with full circles represent, respectively, CB710 and CB1460 scenarios. Dashed lines distinguish FIX from OPT scenarios. These results show that CCS technologies are significantly deployed at similar levels in all low-carbon scenarios, indicating that CO2 mitigation policies play a key role in stimulating CCS deployment. On the other hand, consistent with Figure 2, there is no substantial difference between OPT and FIX scenarios, which suggests that R&D policy only influences CCS technology diffusion to a limited extent. Total capacity of power plants with fossil-based CCS in 2050 is higher in CB scenarios with costs from MERGE-ETL and GEM-E3 (panels c and e) than from WITCH (a), despite the modest capital cost reductions from the former models in comparison to the latter. Higher relative dependence on fossil fuels in the power sector (see SI for a detailed Figure on the evolution of the power sector) resulting from the more conservative capital cost reductions observed for competing technologies (such as solar and wind) in MERGE-ETL and GEM-E3 may justify this difference. Regarding biomass-based CCS, CB scenarios linked to WITCH present the highest capacity level in 2050, which is consistent with the largest capital cost reduction resulting from this model.
Once we add perfect spill-over assumptions to the optimal implementation of R&D policies, we observe a higher influence of R&D on CCS diffusion. Figure 4 shows the absolute difference, in GW, of OPS scenario results relative to OPT scenarios. No difference is observed for scenarios with costs from MERGE-ETL because it considers perfect regional overspill-overs by default, but remarkable differences can be observed in scenarios derived from WITCH and GEM-E3. In both W_CB1460_OPS and G_CB1460_OPS, power plants with fossil-based CCS have higher capacity than their OPT counterparts in 2040 and in 2050. However, installed capacity is lower in the more stringent G_CB710_OPS in all years, as well as in W_CB710_OPS in 2030. A similar trend is observed in results for biomass-based CCS, especially in scenarios derived from WITCH: installed capacity is lower relative to OPT scenarios under both carbon budgets. This downward trend might be explained, by the higher deployment of competing technologies under perfect spill-overs of knowledge, although CCS remains as a key technology for decarbonization due to the persistence of coal and gas in some regions.
Figure 5 depicts TIAM-ECN projections for installed capacity of variable renewable energy (VRE) technologies: solar PV and CSP (first row) and onshore and offshore wind (second row). Long-term impacts of R&D policies are limited for both technologies, since OPT and FIX scenarios are similar. In scenarios with capital costs derived from WITCH, a significantly higher amount of solar PV is deployed (panel a) compared to the corresponding counterparts with costs from MERGE-ETL and GEM-E3 (panels c and e, respectively). These results link directly with the higher electricity share in FEC shown in Figure 2 (panel a). Figure 5 also shows that solar is fairly deployed already in W_REF, indicating that the cost reduction pathway resulting from WITCH render this technology competitive even in absence of stringent climate policies. Low carbon and R&D policy schemes enable additional cost reductions (Figure 1), but do not substantially change the diffusion of solar (Figure 5). TIAM-ECN scenarios using capital cost reductions from MERGE-ETL (panel c) and GEM-E3 (panel e) show a much lower deployment of solar, which kicks-off after 2040 in CB scenarios. In fact, capital cost reductions resulting from these models are more conservative, as discussed in section 3.1, which is a consequence of R&D investments being limited to few components of a technology and of eventual offsets from macroeconomic effects.
Wind energy capacity increases substantially in all three REF scenarios, indicating that these technologies are cost-competitive even without low carbon policies. This is especially true for TIAM-ECN scenarios with capital costs from WITCH: capacity expands worldwide up to almost 8,000 GW (panel b). Results for CB scenarios with costs from WITCH are only up to a 1,000 GW higher than REF level in 2050, but results for 2030 indicate that low carbon policies accelerate diffusion, leading to around 2,000 GW more wind power capacity in the stringent policy scenario (W_CB710_OPT) relative to W_REF. Regarding TIAM-ECN scenarios with cost reductions from MERGE-ETL and GEM-E3 (panels d and f, respectively), a larger gap in capacity observed between CB and REF scenarios reflect the more conservative average capital cost reduction in REF derived from these models, as observed in Figure 1.
We observe that R&D policies in a perfect spill-over dynamics can significantly favour the expansion of VRE technologies (Figure 6). In scenarios with costs from WITCH and GEM-E3, installed capacity is higher in OPS than in OPT scenarios: G_CB710_OPS scenario, for example, shows an increase of 1,600 GW. In fact, the lowest capital costs observed for a region is a result from GEM-E3, which in OPS scenarios is spread globally, leading to a significant capacity expansion. In that context, solar and wind energy technologies seem to become more competitive under perfect spill-over assumptions, and they can even limit the expansion of CCS in the power sector.
Figure 7 shows TIAM-ECN results for FEC of electricity (first row), biofuels (second row) and hydrogen (third row) in scenarios with capital costs from WITCH (panels a, b and c), MERGE-ETL (d, e and f) and GEM-E3 (g, h and i). The higher level of electrification in WITCH-derived scenarios, which was observed in Figure 2, is also observed here. For each of the three models, electricity consumption levels are very similar among REF and CB scenarios, and only a slight increase is observed CB710 scenarios in 2050. This links with results shown in Figure 1 for technologies in both supply and end-use side (see, in special, G_CB710 results for CCS, VRE and EVs), in which, for instance, capital cost reductions for EVs in CB710 are similar or even smaller than in REF. This illustrates how cost increases incurred from mitigation policies might offset the effects of R&D investments on technology diffusion.
Biofuels consumption in final sectors declines over time in all cases (panels b, e and h), although CB scenarios present a less pronounced decrease due to the imposed carbon restrictions. This is an indirect effect of shifting biomass resources from final sectors to the power sector, which is a way to expand biomass-based CCS technologies in CB scenarios. Regarding hydrogen consumption in final sectors, it increases to over 30 EJ/yr by mid-century in all CB_710 scenarios (panels c, f, i). Consistent with previous results, the stringency of climate policy is the main differentiator among TIAM-ECN projections, while R&D strategy and choice of IAM with ITC model used to derive the cost assumptions have a smaller impact on the results. The slightly higher levels of hydrogen and biofuels consumption in projections based on MERGE-ETL may stem from the fact that this model has a very detailed set of technologies for advanced fuels production based on these two carriers, which is reflected in the capital cost reductions incorporated in TIAM-ECN.
When we add the assumption of perfect spill-overs to the OPT scenario, we can observe that electricity consumption is slightly favoured: CB1460_OPS and CB710_OPS scenarios inherited from both WITCH and GEM-E3 show a limited increase – inferior to 5% - in consumption of electricity relative to the corresponding OPT scenarios (Figure 8, panel a). This is an effect of the higher electricity production from VRE resulting from the perfect spill-over assumption of low solar and wind energy capital costs, which drives costs down. As consequence, consumption of biofuels and hydrogen is negatively affected, especially in CB710 scenarios in 2050, leading to less consumption in OPS than in OPT scenarios, as observed in panels b and c of Figure 8.
We also look at the impact of capital cost reductions driven by R&D and climate policy on energy systems costs. The energy system contains all energy conversion routes from resource to end-use and the corresponding energy extraction, conversion, transportation, and consumption costs. Hence, its costs include not only technology capital costs, but also fixed and variable operational and maintenance costs, trade costs, and commodity prices (when applicable). In Figure 9, we show the undiscounted annual energy system cost difference of CB scenarios relative to their corresponding REF in absolute terms (billion US dollars per year). The figure includes OPT, OPS and FIX scenarios. Scenarios derived from WITCH have the lowest additional cost, which is consistent with the fact that this model provides the most optimistic capital cost reduction ranges among the three IAMs with ITC. Aligned with what has been observed regarding technology diffusion, climate policies are the main driver of energy system cost additions, resulting in similar values in both OPT and FIX scenarios. Small negative values observed in 2030 and 2040 in CB1460 scenarios relate to lower costs from trade. Cost additions are clearly lowered in OPS scenarios, in which perfect spill-overs are possible. This is observed in both CB1460 and CB710 scenarios from WITCH and GEM-E3, and notably more prominent in the more stringent G_CB710_OPS scenario – around US$ 1,000 billion difference. The steep cost reductions derived from GEM-E3, especially in technologies that are currently already well consolidated, such as solar PV and wind energy, explain this result.