Assuming we could build a power grid that is mostly powered by solar, wind and storage, how much would it cost and how much more would it cost if we did it in deliberately sub-optimal ways?
Looking for an example of a very different climate zone within the European power grid, we can use the data from the Italy-South bidding zone (Calabria) with is fully contained within the mediterranean climate zone, where we can expect less variation and more sunny days and a higher capacity factor for solar.
To answer this question, we are using the configuration from a previous system cost optimization, based on a combination of wind and solar generation, backed by a hybrid storage in order to cover 100% of the hourly load. The hybrid storage consists of Li-ion style batteries for fast and efficient short duration storage combined with inefficient but cheap hydrogen underground storage.
The goal of this new simulation is to see how much the system cost increases if any of the production and/or storage technologies are removed from the optimal mix.
As an example, we are using current hourly load and Wind/PV generation forecast for the Germany bidding zone from the ENTSO-E transparency platform. Then we use a linear program to scale up the current production levels and add storage configurations to cover the load with the least amount of cost. This assumes that any additional wind and PV deployment would have roughly the same hourly production profile as the existing ones today.
Maybe the most surprising part about these results is the relatively small impact of the short-term storage (Li-ion style batteries) on the system cost, as the cost only increases slightly if they were left out. It seems that most of the heavy lifting comes from wind and long duration storage, which could be expected given Germany's windy north-sea coast as well as a variable and strongly seasonal climate.
Trying to cover the load with PV and batteries alone would be about 6 times as expensive as the optimal mix. As we can see from the table below, this would result in roughly a 4x PV overbuild with 85% of generated electricity discarded rather than stored.
| Cost E/MWh | Cost multiplier | Efficiency % | Curtailment % | |
|---|---|---|---|---|
| S/W + Li-Ion/H2 | 102.02 | 1.00 | 83.78% | 3.52% |
| S/W + H2 | 104.95 | 1.03 | 80.73% | 4.09% |
| W + Li-Ion/H2 | 126.66 | 1.24 | 75.81% | 7.18% |
| W + H2 | 127.91 | 1.25 | 74.68% | 7.44% |
| S + Li-Ion/H2 | 170.04 | 1.67 | 66.82% | 11.66% |
| S + H2 | 195.01 | 1.91 | 54.76% | 11.41% |
| S/W + Li-Ion | 214.39 | 2.10 | 61.29% | 38.04% |
| W + Li-Ion | 295.30 | 2.89 | 34.35% | 65.36% |
| S + Li-Ion | 586.92 | 5.75 | 13.11% | 86.23% |
We can indeed see solar-only plus storage configuration move up in the efficiency ranking compared to wind only, We can also see how battery short duration storage is a more effective complement to solar than to wind, as solar has an inherent 24 hour cycle while the variability of wind tends to be on a day to weeks timescale that is out of reach to be efficiently compensated by short duration storage.
| Cost E/MWh | Cost multiplier | Efficiency % | Curtailment % | |
|---|---|---|---|---|
| S/W + Li-Ion/H2 | 110.31 | 1.00 | 84.86% | 3.37% |
| S/W + H2 | 124.46 | 1.13 | 75.05% | 4.60% |
| S + Li-Ion/H2 | 130.56 | 1.18 | 78.49% | 7.45% |
| W + Li-Ion/H2 | 152.85 | 1.39 | 71.94% | 5.08% |
| W + H2 | 153.94 | 1.40 | 71.07% | 5.21% |
| S + H2 | 167.63 | 1.52 | 58.68% | 5.99% |
| S/W + Li-Ion | 181.10 | 1.64 | 67.41% | 30.09% |
| S + Li-Ion | 234.46 | 2.13 | 40.99% | 56.80% |
| W + Li-Ion | 442.88 | 4.01 | 29.12% | 70.44% |
While solar and wind are by now the cheapest source of new generation in most parts of the world, powering a grid with 100% renewable energy will require a careful mix of generation and storage technologies which complement each other.
What might be surprising is that short-duration storage (battery storage) might be playing the least significant role in this mix, despite batteries being the storage technology that is built out most rapidly today. Grid-scale battery storage plants are very effective in short-term balancing of the grid as well as increasing the market value of solar (intra-day arbitrage), which are value contributions ignored by this simulation.
Batteries will unlikely be able to displace gas or coal plants as the flexibility provider of last resort to balance out long-term or seasonal differences, as they lack the "long breath" and "deep pockets" of a very large and very cheap energy reservoir.
H2 or derivatives which are expensive to produce but cheap to store could in principle fill this role, but only if CO2 emissions were to become illegal or sufficiently expensive to displace gas and coal plants from this role.
Based on the cost assumptions below, carrying 100% of the load at all times with renewables and storage could be accomplished today with a unit cost per MWh that is less than double that of raw solar and wind generation.
Appendix: Cost Assumptions
Since predictions are hard, specially about the future, we are using as much as possible current cost estimates for our assumptions.
As cost estimate for wind and solar production, we are using a mid-point of the onshore wind and utility scale PV LCOE estimates from the 2024 edition of the renewable energy technology LCOE report by the German Fraunhofer Institute (67 Euro/MWh for wind and 55 Euro/MWh for PV).
For the storage cost estimates, we are using annual infrastructure cost based on the Technology catalog of the Danish Energy Agency. For the Li-ion grid-scale battery storage, we are assuming 143'000 E/MWh plus 16000 E/MW investment cost with operating costs of 540 E/MW over 15 years at 90% efficiency. These might be high estimates as they are extrapolated from data in 2018 while battery storage has seen a lot of progress since then.
For the hydrogen storage, we are assuming a capex/opex of 800000/16000 E/MW electrolyzer capacity at 70% efficiency and 900000/30000 E/MW for a combined-cycle gas turbine plant at 50% efficiency, both amortized over a lifetime of 25 years. The underground salt cavern storage is assumed at 2000 E/MWh over a lifetime of 100 years.
For the capital cost, we are assuming a 5% amortization rate which is nearly double the current German 30 year bond rate.
While we can all be nit-picking at any of these numbers, they seem reasonably in the ballpark and changing them ±2x has little impact on the big picture of the results in relative terms.