Is a 100% renewable electricity mix cost-efficient under weather uncertainty?
Job Market Paper
Renewables (wind and solar) do not emit greenhouse gases, but their output is highly uncertain and intermittent. This raises the question of how costly a 100% renewable electricity mix would be, and how this cost depends on the cost and efficiency of the available storage technologies. I study these questions in a dynamic stochastic model of wind and solar power generation and storage. The model is calibrated on German data and accounts for both flexible consumption and demand curtailment. I first solve numerically for the optimal operation of given generation and storage capacities, using a semi-Lagrangian scheme with relative value iteration. I then optimize over renewable and storage capacities to obtain the efficient electricity mix. With battery storage only and a 5% discount rate, the optimal 100% renewable mix requires multiplying the 2025 wind and solar capacities about sevenfold, at a welfare-adjusted levelized cost of 285 €/MWh. Allowing for lossy, yet inexpensive, hydrogen storage reduces the optimal renewable capacities sharply and lowers this cost to 142 €/MWh. Relative to perfect foresight over the observed decade, weather uncertainty mainly raises the optimal hydrogen storage capacity, from 23 to 35 TWh. With a fossil backup, a non-negligible storage capacity becomes optimal only at a positive carbon price, and marginal abatement costs rise sharply close to zero emissions if there are no other low-carbon alternatives.
Presented at (selection):
2025: EAERE conference (Bergen), FAERE conference (Nantes), ENTER Jamboree (Stockholm), UAB (Barcelona, ENTER exchange seminar), University of Leipzig lunchtime seminar
2026: EDF (Paris), Doctorissimes (Paris), HYRCE (Liège), EEA-ESEM Congress (Dublin), IAEE European Conference (Munich), IAEE COP Symposium (Antalya)
This work was supported by two French government grants managed by the Agence Nationale de la Recherche under the "Investissements d'avenir" program (references "ANR-18-EURE-0021", "ANR-21-ESRE-0051"). Under the latter, this work was granted access to the MesoNET resources center and the MesoNET Project under the allocation m26158. This work was also supported by the Agence Nationale de la Recherche through the project AVINEQ.