Systemwide energy return on investment in a sustainable transition towards net zero power systems

The study addresses how different decarbonisation pathways for the global power sector affect systemwide net energy performance, measured via energy return on investment (EROI). In light of the Glasgow Climate Pact and accelerated net‑zero pledges, the authors argue that anticipating biophysical constraints is crucial for selecting optimal transition pathways. EROI—defined as the ratio of energy output to energy invested—has been inconsistently applied in prior research, often with differing boundaries and incomplete treatment of enabling technologies. Recent work emphasizes systemwide analyses for energy transition scenarios; however, methodological gaps (e.g., energy quality standardization, boundary selection, interoperability of technologies) can bias conclusions about the feasibility of high‑renewable systems. This paper applies an improved, systemwide EROI framework to nine global transition scenarios to evaluate sustainability risks and to link EROI with levelised cost of electricity (LCOE) and CO2 emissions.

Prior EROI studies vary widely in concepts, boundaries, and comparators (fossil vs renewables, inclusion/exclusion of enabling technologies). Some questioned the plausibility of 100% renewable systems on net energy grounds, while more recent systemwide approaches suggest that methodological gaps (e.g., failing to capture optimal interoperability and standardize primary energy quality) may drive pessimistic conclusions. Integrated assessment model (IAM)-based EROI studies generally show decreasing net energy for society as fossil EROIs decline with depletion, while technology‑level EROIs for wind and solar tend to increase over time; gas and nuclear often exhibit lower and relatively stable EROIs. Comparisons across studies are complicated by differences in scope (whole energy system vs power sector), treatment of storage, and whether storage’s energy investment is integrated with renewables or treated separately. The paper positions its approach relative to this literature by adopting a harmonized LCA‑based cumulative energy demand (CED) framework, energy quality normalization, and integration with detailed energy system modeling.

The authors extend an Excel‑based LUT‑EROI model to compute systemwide EROI for nine global power sector transition scenarios from 2015–2050 in five‑year steps across nine world regions. The scenarios include five LUT Best Policy Scenarios (BPS: NWF, WF, plus2030, plus2035, plus2040), two IEA scenarios (SDS, STEPS), and two Teske/DLR scenarios (1.5°C, 2.0°C). Scenario inputs (technology portfolios, capacity additions, generation mixes, storage, curtailment, fuel usage) are taken from LUT-ESTM outputs (cost‑optimized energy system model) or replicated in the same modelling environment for the IEA and Teske/DLR cases. Key elements: (1) CED database: Technology‑level CED values are built from ecoinvent v3.7.1 inventories, with clear system boundaries (allocation, cut‑off by classification), disaggregated into construction/decommissioning (per kW capacity) and operation (per kWh) components. Primary energy is standardized to electricity quality via category‑specific conversion factors to ensure energy quality consistency across all inputs. (2) Technology evolution: Dynamic CED updates incorporate learning effects—PV ELR ~14% (decreasing over time), batteries with ELR moving from ~5% (2015) towards ~4% (2050). For nascent technologies (electrolyzers, DAC, methanation, CSP ST, nuclear), future CEDs consider maturity trajectories, planned capacities, and market shares. (3) Fuel embodied energy: Upstream energy for extraction, processing, and transport of fossil fuels is included (excluding energy content), with default trends from Sgouridis et al. and sensitivity using Delannoy et al. for oil and gas. Nuclear fuel CED is taken from ecoinvent (static). (4) Systemwide EROI: EROI is computed at the point of final consumption (F; including transmission/distribution losses) and at generation (G; in Supplementary Info). Annual net electricity supply accounts for self‑consumption, curtailment, storage and other process losses. Energy invested aggregates construction and operation CED over lifetimes using full‑load hours. (5) Annual Energy Investment Flow (AEF): Annualized energy investments are computed to assess sustainability risk during transition, allocating construction energy to the commissioning period without lifetime amortization. (6) Sensitivity and regional analysis: Scenario‑level EROIs are tested against alternative fuel embodied energy trends for gas/oil, and regional heterogeneity is evaluated due to resource diversity, technology mix, and storage needs.

- Across 2015–2050, all scenarios maintain global systemwide EROI above 16 (and above the net energy cliff upper limit ~10), indicating physical feasibility of the transitions considered. - Early transition (2015–2025): EROI rises from ~18.8 to >20 as fossil generation is displaced by renewables without yet requiring extensive enabling technologies. - Beyond ~2025 in BPS scenarios, EROI declines as VRE shares exceed ~50% and increased enabling technologies (batteries, hydrogen/methane storage, electrolysis, methanation) add to invested energy; accelerated BPS-plus scenarios show sharper temporary declines around their forced completion dates (2030/2035), then stabilize after achieving 100% RE. - IEA scenarios: EROI dips from ~20 to ~18 by 2025, then rebounds slightly to >19 by 2050 as coal is phased down and gas/nuclear/RE shares evolve; lower VRE penetration and inclusion of nuclear limit achievable EROI compared to BPS. - Teske/DLR scenarios: EROI trends are similar to IEA until ~2040, then decline due to capacity expansion of high‑CED, low‑efficiency technologies (geothermal, CSP ST). Nascent technologies (wave, fuel cell CHP) have negligible EROI impact at current scales. - VRE‑storage‑curtailment nexus: EROI increases with VRE penetration up to ~50%, then declines; the decline steepens approaching ~80% VRE as seasonal storage needs grow. Spatial aggregation smooths variability and can enhance EROI at higher VRE shares. - Cost and emissions: System LCOE falls from 70.9 €/MWh (2015) to 45.2–49.7 €/MWh (BPS), 53.9–54.1 €/MWh (Teske/DLR), and 59.2–69.5 €/MWh (IEA) by 2050. LCOE peaks 2020–2030 due to transition dynamics and CO2 costs (10–20% of LCOE depending on scenario). Low LCOE solutions correlate with lower EROI as fuel/CO2 costs are replaced by storage/enabling tech investments. Lower EROI paths tend to be cleaner (lower CO2 intensity), while higher EROI paths often retain higher CO2 emissions (e.g., IEA scenarios not net‑zero by 2050). - Investment flows: AEF never exceeds 16% of final energy consumption; systemwide energy investment flow upper limit estimated at ~7%. Annual EROI values remain ≥5 despite substantial RE buildout. Upstream fuel chain energy needs approach zero as fossil use declines, replaced by RE investment energy. - Sensitivity: IEA scenarios are highly sensitive to higher embodied energy of oil/gas (EROI reductions up to ~6 points for STEPS and ~3 for SDS under alternative fuel EROI trajectories). BPS/BPS-plus are largely unaffected; Teske/DLR show ~1‑point changes. - Regional heterogeneity: Regions with diverse resources and strong hydropower (e.g., South America) can reach EROI >26; reliance on high‑CED technologies (e.g., geothermal) or CCS integration can depress regional EROI. Resource diversity can reduce storage needs and curtail the EROI decline during rapid transitions. - Demand differences among scenarios (BPS ~48.38 PWh vs IEA/Teske ~45–46.5 PWh by 2050) showed no clear effect on EROI trends compared to system composition and technology choice.

The analysis shows that pathway design—particularly the VRE mix, the pace of transition, and the scale/type of enabling technologies—critically shapes systemwide EROI. Replacing fossil fuels with renewables initially raises EROI, but as VRE penetration increases, additional storage and flexibility investments raise energy inputs and EROI declines, with the steepest reductions in accelerated transitions and at very high VRE shares requiring seasonal storage. Nevertheless, EROI remains well above the net energy cliff, supporting physical feasibility of high‑renewable power systems. Lower EROI solutions often correspond to lower CO2 emissions and lower LCOE as expensive fuels and carbon costs are replaced by capital‑intensive but efficient renewable and storage assets. However, scenarios that rely on gas and nuclear can show higher EROI in the near term while risking sensitivity to fuel depletion and potentially forgoing achievable EROI and cost benefits attainable with optimized high‑VRE systems. Multicriteria planning (considering EROI, LCOE, emissions, and technology risk) is necessary, and regional resource diversity can mitigate EROI declines by reducing storage dependence.

All nine global power sector transition scenarios assessed maintain systemwide EROI safely above the net energy cliff (≥10), indicating that transitions to low‑carbon and even 100% renewable systems are physically feasible from a net energy perspective. Key contributions include: (i) demonstrating the EROI implications of different transition speeds and technology mixes; (ii) highlighting the VRE‑storage‑curtailment nexus that drives EROI trends; (iii) linking EROI to LCOE and CO2 outcomes; and (iv) quantifying sensitivity to fossil fuel embodied energy. Shortening the transition period causes sharper but temporary EROI declines that stabilize post‑100% RE, while diversification of renewable technologies and resource‑rich regions can smooth EROI trajectories. Future research should examine the impacts of higher electrification, energy efficiency, and sector coupling on systemwide EROI, improve CED estimates (including recycling and network expansion), and further elucidate the dynamic relationship between EROI and economic metrics to anticipate potential macroeconomic implications of declining EROI during rapid transitions.

Limitations arise from methodology and data: use of a single LCA database (ecoinvent v3.7.1) and linear/extrapolated CED projections; uncertainty in technology learning rates and future process improvements; exclusion of recycling of materials and wastes; limited treatment of transmission and distribution network expansion CED (assumed negligible due to limited LCI data); static nuclear fuel CED; potential aggregation bias in global/regional modelling; operational anomalies and seasonal/resource distribution uncertainties in LUT‑ESTM; and the relative trend analysis used to compare EROI and LCOE, which may not generalize across systems with different fossil/renewable shares. Socio‑economic‑political factors, climate impacts and extremes, and broader sector coupling are out of scope. Sensitivity was conducted for oil and gas embodied energy; further sensitivities (e.g., coal, nuclear) warrant investigation.