Can We Afford Large-scale Solar PV?

In the US solar energy has steadily risen in how much of our electricity it supplies. In 2024, it supplied just over 7% of total electricity generation, up from less than 1% in 2015. Because most planned US electricity generation projects are solar photovoltaics (PV), this fraction will almost certainly continue to rise.

But how much can it continue to rise? Is it feasible for solar power to meet most of our electricity demand? In the essay Understanding Solar Energy, we used some simple simulations of solar power to understand how much electricity demand solar PV can supply under different conditions. We found that due to solar’s intermittency, supplying large fractions of electricity demand requires a fair degree of “overbuilding” (solar panel capacity well in excess of total electricity demand), as well as a fair amount of storage. For a single family home where power demand never exceeds 10 kilowatts (and most of the time is below 2 kilowatts), supplying 80% of annual electricity consumption requires at least 13.7 kilowatts of solar panels, and 40 kilowatt-hours of storage. And supplying even higher fractions of electricity demand — 90%, 95%, 99% — the infrastructure requirements gets even more burdensome. Going from 80 to 90% of electricity supplied requires nearly doubling solar panel capacity.

However, we also found that the falling costs of solar PV will make it feasible for solar to supply large fractions of electricity demand cost-effectively. Reaching 90 or 95% is indeed costly, but 70-80% appears to be well within the realm of possibility.

One common objection to the large-scale use of solar PV is that, even with batteries, it effectively requires an entire parallel energy system to meet electricity demand when there’s little or no sun and no battery charge. Even if the solar plus battery system can meet demand 99.9% of the time, supplying that last 0.1% of demand might theoretically require a large amount of expensive infrastructure, making effective costs much higher.

For instance, if power demand in a state is a constant 20,000 megawatts, I might need a 20,000 megawatt solar and battery system, plus an entire other 20,000 megawatt energy system (say, 40 large gas turbines) to turn on in emergencies. Once you take these extra costs into account, so the theory goes, the economics of solar PV look much worse.

It is true that meeting 100% of electricity demand with large-scale solar PV requires some amount of parallel infrastructure, such as gas turbines, to fill in the gaps. But analysis suggests that at low enough solar PV and battery costs, this isn’t all that burdensome. Partly this is because in many cases (such as with gas turbines), a large fraction of electricity costs are due to fuel costs, which aren’t incurred when the system isn’t running. But it’s also because batteries can help reduce the need for this parallel infrastructure. Batteries aren’t just a complement to solar PV: they’re a complement to any energy generation system. We can use our gas turbines to charge the batteries too, helping us smooth out the peaks in demand and reducing how much “extra” infrastructure is required to meet demand.

Meeting electricity demand with solar PV

To understand what sort of electricity generation infrastructure is required to meet large-scale demand, we’ll use electricity data from the state of California. The graph below shows hourly electricity supplied by CAISO (California’s independent system operator) for the first week of January 2024.¹ Data is from Gridstatus.io.

And this graph shows the daily maximum demand over the entire year of 2024.

Day to day, we see a pattern of electricity demand that’s similar to our single family home simulation. We see electricity demand rise in the morning, decline around 9-10 AM, rise again in the evening around 5-6 PM, then decline again late at night. If we look at seasonal variation, we see electricity demand rise steeply in the summer, then decline in fall.²

Right away, we can see that the problem of needing “extra” infrastructure isn’t simply a solar PV and battery problem. During most of 2024, electricity demand didn’t exceed 30,000 megawatts, but during the summer months it regularly exceeded 40,000, and for one brief window it approached 50,000 megawatts. Meeting 100% of California’s electricity demand will require a lot of infrastructure that most of the time will sit idle, regardless of what system of electricity generation you use.

If we want to meet this demand with solar PV, we need to understand what sort of capacity factor (the ratio of actual solar PV power supplied to maximum theoretical capacity) we can expect to achieve. Gridstatus.io breaks down electricity supplied by the source of generation, letting us see how much electricity is being supplied by solar PV in California at any given time. By combining this with the total nameplate solar PV capacity in California, we can calculate an hour-by-hour capacity factor.³ Here’s CAISO’s solar PV capacity factor for the first week of January 2024.⁴

And here's the daily maximum capacity factor for the entire year of 2024.

As with our single family home simulation, we see that solar PV generation drops to zero at night, and rises to its maximum during the day. That maximum changes over the course of the year: during the summer months, California’s solar PV panels regularly generate 90% or more of their nameplate capacity, but in January that maximum is closer to 60%. And day to day, output might be reduced even more due to things like cloud cover. On some summer days in California, solar PV generation doesn’t exceed 70% of its theoretical maximum, and on some winter days it never cracks 20%.⁵

Using this electricity demand and solar PV capacity factor data, we can calculate what fraction of California’s electricity demand can be met by different combinations of solar and storage. The graph below shows the fraction of electricity supplied by solar PV systems ranging from 20,000 megawatts (slightly less than California’s current 22,000 megawatt capacity) and battery storage systems ranging from 50,000 megawatt-hours to 800,000 megawatt-hours. (If we simplify and assume that California’s maximum electricity demand is 50,000 megawatts, this is roughly equivalent to having 1, 2, 4, 8, and 16 hours of battery storage for the entire state.)

We see here the exact same diminishing returns that we saw when simulating a single family home. Each additional watt of solar PV capacity and watt-hour of battery storage gets used less and less frequently, and contributes a smaller amount to meeting overall electricity demand. And as we approach 100% electricity demand, the infrastructure requirements become enormous to deal with the occasional long stretch of cloudy days and low panel output. Getting through that brief period in late January where maximum daily solar PV output never exceeds 20% requires a lot of extra panels and batteries.

These diminishing returns mean that it’s infeasible to meet 100% of electricity demand with just solar PV and batteries, and we need to make up the difference with some other energy generation technology. Let’s assume that we’ll meet any additional demand with combined cycle gas turbines, and that these turbines aren’t used to charge the batteries. The graph below shows how much gas turbine capacity is needed to meet 100% of electricity demand for different combinations of solar PV and battery storage.

We can see here the dynamic that we were worried about initially — while energy supplied by our gas turbines declines precipitously as our solar PV and battery system gets larger and larger, power demand (and thus infrastructure requirements) declines much more slowly. A small number of cloudy stretches, like that period in late January where PV capacity factors are under 20%, require a large amount of gas turbine capacity to deal with. Even with a huge 160+gigawatt solar PV + 16-hour battery system, capable of supplying more than 95% of our energy, we still require more than 27 gigawatts of gas turbines on top of that (more than 50% of maximum electricity demand) to meet 100% demand.

However, we can alter this calculus significantly by allowing our gas turbines to charge our batteries. By letting turbines charge batteries in periods of low demand (such as at night when demand drops), we can further smooth demand peaks in cloudy stretches when there’s little sun, and reduce the gas turbine capacity we require. The graph below shows how much gas turbine capacity is needed to meet 100% of electricity demand at different combinations of solar PV and storage, if we allow gas turbines to charge our batteries too.⁶

Without gas turbine charging, our 160+gigawatt solar PV and 16-hour battery system still required over 27 gigawatts of gas turbine capacity to meet demand. With gas turbine charging, that falls to 8 gigawatts. And those turbines run a lot more often: the 27-gigawatt system only generated around 1.4 million megawatt hours, or less than 1% of total demand. The 8-gigawatt, battery-coupled system generated almost 10 times that. (For reference, California currently has about 39 gigawatts of gas turbine capacity, though this will be both combined cycle plants and simple cycle plants used for peaking.)

Cost of meeting demand

So by using gas turbines (or some other energy generation system) to charge batteries, we can dramatically reduce how much extra generation capacity we need to pick up the slack for solar PV. But what does this do to our overall costs for generating electricity?

The chart below shows the levelized cost of electricity (LCOE) for meeting 100% of electricity demand at different combinations of solar PV and battery storage, with combined cycle gas turbines picking up the slack. Costs are based on approximate current US costs (see Appendix below for details).

And here’s how much electricity is provided by the solar panels (either directly or via batteries) for each combination of solar PV and battery storage capacity.

Putting these together, this graph shows the generation costs required to meet 100% of electricity demand at different fractions of solar PV capacity.

We can see that at current US costs, it does indeed get very expensive to meet large fractions of electricity demand with solar PV. Up to around 40% of demand, costs remain fairly low, but beyond that they rise quickly. The current costs for meeting 80% of electricity demand with solar PV are more than three times the costs of meeting 40% of electricity demand.

But what happens if costs for solar PV and batteries continue to fall? The graph below shows current US costs against current worldwide averages, and some speculative future costs. (Current average US nuclear costs and combined cycle gas turbine costs are per Lazard.)

Current worldwide average costs for solar and battery storage let us get to around 50% of total electricity demand with solar PV for around the current LCOE of gas turbines in the US. If we can hit $400 kw solar and $100 kw-h storage, we can push that percentage to around 80%. And while these costs are substantially lower than current US costs, they’re not outside the realm of possibility (they’re probably roughly the lowest costs currently achievable in China).

Conclusion

These are still somewhat simplified simulations. For one, they only look at generation costs, and totally ignore transmission or distribution.⁷ They also only look at costs, and not prices: what generators would actually charge, which could easily be much higher. (More generally, they ignore electricity market dynamics, which are very complex.) And they totally ignore the fact that in the real world states can share electricity, further damping demand peaks. But they can nevertheless help us understand the mechanics of large-scale solar PV deployment. My main takeaways:

Solar has a lot of room to grow at current prices. The simulations above suggest solar PV can meet 30-40% of electricity demand without requiring burdensome additional infrastructure.

Cheap batteries are as big a deal as cheap solar PV. Low cost batteries are transformative, not simply because of how they complement solar but because of how well they complement other energy generation technologies. Batteries act as a buffer to overcome a mismatch between solar PV supply and electricity demand, and that buffer works just as well to smooth out variation in demand more generally.

If solar PV and battery costs continue to fall, supplying very large fractions of electricity demand with solar PV becomes feasible. At $400 a kw solar and $100 a kwh batteries (costs China is probably achieving right now), we could meet 80% of electricity demand with solar PV for roughly current US average combined cycle gas turbine costs. If, like some folks, you think solar PV and batteries will get even cheaper than this, the path to almost total solar and battery dominance is very clear.

Concerns that large-scale solar PV requires a lot of parallel infrastructure aren’t unreasonable, but large-scale storage deployment dulls them significantly.

Appendix: LCOE calculation assumptions

• Current solar PV capital costs: $1100/kw (source).

• Current battery capital costs: $475/kw (source).

• Current combined cycle gas turbine capital costs: $1000/kw

• Gas turbine fuel costs: $23/mwh (source)

• Solar PV maintenance costs: 1.5%/year (source)

• Battery maintenance costs: 2.5%/year (source)

• Gas turbine maintenance costs: 4%/year (source)

• Solar PV output degradation rate: 0.5%/year (source)

• Discount rate: 7%

• Solar PV lifetime: 30 years (source)

• Battery lifetime: 15 years (source)

• Gas turbine lifetime: 25 years (source)

For gas turbines, there’s a lot of variation in costs. $1000/kw would put this somewhat on the lower end of Lazard’s cost estimate. It’s on the high-end of what Gas Turbine World estimates for current large, combined-cycle plants, but well below smaller, simple cycle peaker plants. But it’s much lower than what power companies are currently paying for combined cycle plants, which seems to be closer to $2400/kw (possibly due to high demand driving up prices). Similarly, maintenance costs varied from source to source (these are somewhat higher than what Gas Turbine World estimates.)

Thanks to Austin Vernon and Nathan Iyer for reading a draft of this. All errors are my own.

1. CAISO manages around 80% of California’s electricity. For simplicity, I’ll refer to this as “California’s electricity demand,” even though it’s not all of California.

2. This seasonal variation is somewhat different than our single family home simulation, presumably due to a combination of California having mild winters and few heat pumps.

3. Nameplate capacity is given for the end of each year listed. To get an average value for 2024, I averaged 2023 and 2024 capacity.

4. I believe these values from Gridstatus.io are the full power generated, before any sort of curtailment.

5. If we average the hourly capacity factors over the course of the year, we get just over 27%, almost exactly California’s average solar PV capacity factor.

6. Turbines are set to charge the batteries if there’s excess turbine capacity and battery charge falls below 70%. If you spent time optimizing when to charge the batteries, you might be able to get away with even less turbine capacity.

7. This might make solar PV look even better if you think that large-scale solar PV deployment will reduce the need for transmission.