The Real Cost of Solar Energy

Linkedin

Facebook

Twitter

Email

Print

Note:

This post is a late-stage draft and is not yet meant to be fully public. It will likely be revised over the coming weeks after we receive comments and suggestions from early readers. Once all revisions are made, it will be republished and made more publicly available—likely in September or October of 2024.

Thanks,

Doug

You’re standing in a crowd, waiting for the start of a major marathon. You have money to burn and want to place a bet on who’ll win the race. You scan the runners. Suddenly, you spot a familiar face. It’s Usain Bolt, the fastest man alive!

Bolt has a smile on his face and a swagger in his step. He should, he’s at his competitive peak. You think to yourself, this is it… the sure thing you’ve been waiting for. You rush to the nearest bookie and put your money on Bolt to win it all.

You don’t know it yet, but you’ve just made what’s likely the worst bet of your life, having fallen for one of the oldest mistakes in the book—confusing speed with endurance.

You see, Bolt is a sprinter, not a marathoner. And running 26.2 miles is a wholly different challenge than running 100 meters. Your fast-twitch expert self has duped you into making a bad decision by placing the “fastest man alive” label in the left-side lobes of your brain, while ignoring the course to be run.

This is the same blunder many make when they buy the notion that solar and wind energy are the “cheapest” forms of electricity on our grids. They are led to believe “cheap” in the short run (which doesn’t matter) must also be cheap in the long-run—and thus, the best solution to our energy needs; that switching to solar and wind power as soon as possible makes sense. Save the planet and our wallets at the same time, the argument goes.

But is it correct?

Wind and solar energy have big limitations and drawbacks that make seem to make them unsuitable for what many of their advocates are attempting to do with them. To wit, our energy system is an always-on, always-needed economic asset. Optimizing it requires more than just piling on solar panels and wind turbines, and crossing our fingers that the stars align day-in and day-out.

When it comes to energy, getting correct answers requires work. And developing a deeper understanding of the complex and dynamic nature of electricity production and consumption is essential to getting the grid right.

In this post, we’ll perform a deep dive analysis into the real-world economics and implications of today’s anointed “green” energy source—solar. In the process, we’ll expose what we see as omitted truths behind claims that solar is the “cheapest” form of electricity.

We’ll do this by methodically analyzing a hypothetical grid powered solely by utility scale solar energy + batteries (a PV-hybrid system). In the process, we’ll come to understand the real costs of such a system.

Scenario A: “Highly Idealized” System (Baseline)

Let’s start with Scenario A, a highly idealized “baseline” system that operates at 100% daytime utilization and seamlessly services all nighttime demand via a perfectly sized utility-scale battery energy system (BESS). The supply and demand profile for this scenario is below.

It’s all pretty straightforward. To wit, a full 24-hour day divides into 12 hours of sunlight, with perfectly level end-user demand of 100 GW over the course all 12 hours. Nighttime divides into 12 hours of no sunlight and perfectly level demand of 50 GW (exactly half that of daytime demand) across all 12 hours.

We assume the sun rises and sets fully and immediately at 6 AM and 6 PM, respectively. During all daylight hours, the sun is perfectly positioned for maximum PV-solar panel efficiency and power generation. All days are 100% cloudless.

These idealized conditions repeat each day—without fail.

To service daytime demand, we install exactly 100 GW of PV-solar panels (panels), perfectly matching the system’s perfectly predictable daytime demand. We install an additional 50 GW of panels and 50 GW of lithium-ion battery capacity (and the necessary ancillary equipment). This 100% solar + battery energy storage system (BESS) is what’s referred to as a “PV-hybrid” system” (the system).

Actual operations are straightforward as well. During daylight hours the system’s 150 GW worth of panels perfectly and fully meet daytime demand, as well as fully charge the BESS. At nighttime, the BESS evenly and fully discharges at 50 GW for 12 hours—also exactly meeting demand.

We assume neither the panels nor BESS incur any operating or maintenance (O&M) expense, energy leakage, degradation or replacement costs for the full 20-year horizon. That is, the cost of the system is limited to the cost of installed capital equipment and required investor returns on capital.

For the system’s panel capital costs, we assume an average of $1MM per MW. This cost seems pretty well within the range of estimates for 2024. BESS costs are assumed to be $300 per KWhr for 4-hour duration batteries, or $3.6 MM per MW for the necessary 12-hour nighttime duration.

Determining a reasonable capital cost for BESS in 2024 took some effort. Fortunately, we found—and leaned heavily upon—a study published in 2023 by National Renewable Energy Laboratory (NREL), a unit of the US Department of Energy.

The study, which analyzed battery-cost estimates from 16 reputable sources, acknowledges the wide range of estimates and the fast pace of change with BESS. It also acknowledges the variation among studies in terms of elements included in BESS costs. For example, inverters might be included in one estimate, but not another.

Ultimately, NREL settled on a simple average of all 16 estimates for lithium-ion batteries with four-hour durations and 15-year lives. The study also includes an O&M cost estimate for BESS, which proved useful for incorporating real-world costs in later scenarios.

Finally, we’ll select a (generous) 7.00% as the pre-tax weighted average cost of capital (WACC).

With these assumptions in place, we can now calculate the following outputs…

Daily power consumption totals 1,200 GWhrs (100 GW x 12 hours) during daylight hours, plus 600 GWhrs (50 GW x 12 hours) in the nighttime hours. This sums to 1,800 GWhrs per day, or 657 TWhrs annually (1,800 GWhrs x 365 days). Note that we hold this level of annual consumption steady across all scenarios.

The system’s panels enjoy a capacity factor of 50% (1,800 GWhrs / (24Hrs x 150 GW)). The comparable BESS capacity factor (based on hours of discharge) is 16.7% (4Hrs / 24Hrs).

The system’s breakeven cost of generation is a function of the average annual power revenues collected over the 20-year life of the project needed to achieve the required ROIC of 7.00%. After discounting the result back to 2024 dollars, we get $39.89 per MWh as the system breakeven. We have our baseline.

Scenario B: Adding Intraday Variability

Now that we know the breakeven under highly idealized conditions, we’ll begin methodically layering in certain conditions the system will face in the real world. In doing so, we’ll gain a better understanding of the incremental effects of various factors on system breakeven.

We’ll start by adding intraday (ie, hourly) variability to demand. Below is a chart showing the effects of daily variability on both supply and demand.

Here are the changes in assumptions driving the new demand profile—

In daytime hours, rather than coming online at 100 GW at 6 AM, demand now starts at 65 GW at 6 AM, rising to 125 GW by 11 AM. It stays at 125 GW for two hours. At 1 PM, demand begins a drift back to 65 GW by the 5 PM hour, effectively reversing out the AM ramp-up. Total daytime demand remains at 1,200 GWhrs.

Nighttime demand begins at 60 GW in the 6 PM hour, drifting to a low of 40 GW at 11 PM. It holds at 40 GW for two hours, then drifts upward beginning at 1 AM to hit 60 GW in the 5 AM hour. Total nighttime demand remains steady at 600 GWhrs.

Both the new daytime and nighttime demand pattern repeat each day, every day.

Although considerable hourly demand variability has been added to the system, note that neither daytime nor nighttime demand has changed. That is, total daily and annual demand remain steady at 1,800 GWhrs and 657 TWhrs, respectively.

Because a daytime peak of 187.5 GW—compared to 150 GW in the prior scenario—must now be accommodated , panel capacity must also rise to 187.5 GW. This pushes overall capital investment to $367.5B, up from $330B in the prior scenario, even as the capacity factor of the panels falls to 40%.

BESS capital costs and capacity factor show no change because cumulative nighttime demand shows no change (ie, we assume BESS can vary discharge rates as needed to perfectly follow hourly nighttime demand).

The changes assumed in Scenario B push the breakeven price of power up slightly to $43.31 per MWhr from $38.89.

Scenario C: Adding Seasonal Variability

We’ll now add “peak” and “off-peak” months (ie, seasonality) to the system’s demand profile. This proves to be a more consequential type of variability—at least when it comes to affecting breakeven.

Virtually all systems must deal with seasonality. In this case, we assume an extra 15 GW of demand during the peak months of July, August and September, and 15 GW less demand in the off-peak months of January, February and March During the remaining six months of the year—ie, the shoulder months—both hourly and daily demand remain the same as in the prior scenario.

With changes in peak demand perfectly offsetting those of off-peak months, total power consumption remains unchanged at 657 TWhrs per year.

We now need to resize the system’s panel and BESS capacities to meet system demand. To do so, we must determine peak demand—ie, consumption during the peak hours of peak months. It comes out to 235.7 GW of panels coupled with 65 GW of BESS. This new configuration lifts capital costs to $469.7 B.

Because capacity has risen, even though overall demand hasn’t changed, the capacity factor for panels falls from 40.0% to 31.8%. BESS’s capacity factor falls from 16.6% to 12.8%. The extra panel and BESS capacity needed to service just three months of peak-season demand acts as costly dead weight the remaining nine months of the year.

The higher capital costs and lower capacity factors mean the breakeven must rise. And rise it does—to $55.36 per MWhr—up $16.47 from Scenario B.

Scenario D: Addition of Basic Weather

Having accounted for systematic daily and seasonal variability, we now introduce some idiosyncratic variability to the system in the form of basic weather.

Prior to now, we’ve assumed power generation rose and fell consistently as the sun moved across the sky—but the skies themselves were cloudless. We’ll change that in this scenario. Specifically, for each of the four seasons, we assume one of four levels of the sunlight reach the system’s panels each day over an average 10-day period.

For example, Winter (our off-peak season) is assumed to enjoy three perfectly clear days (ie, 100% sunny), three days of modest cloud cover (ie, 75% sunny), three very cloudy days (ie, 50% sunny), and one day with rain, snow or storms (ie, 25% sunny). This gives us an average of 70% sunny for the season.

We repeat this exercise assuming slightly different frequencies of various weather conditions for Summer (peak season) and Spring and Fall (shoulder-month seasons). These produce averages of 80% sunny for Summer, and 77.5% and 75% for Spring and Fall, respectively.

Since we must design for peak demand, we’ll use 80% sunny in the Summer for sizing the system’s capacities. The calculations are relatively straightforward. The derating of panels from weather simply increases the number of solar panels needing to be installed to meet system demand during peak hours of Summer. This comes out to 287.8 GW of panel capacity (235.7 GW / 80%).

With $528.6 B in total capital now employed as a result of the increase in panel capacity, the breakeven cost of generation increases to $62.30 per MWhr—up $6.94 from the prior scenario. The capacity factor of the system’s panels falls from 31.8% to 25.5%. BESS’s capacity factor remains steady.

We’ve now accounted for how weather—ie, cloud cover, storms, rain and snow—can affect performance of the system.

Scenario E: Addition of BESS O&M and Inefficiencies

Until now, we’ve assumed the system’s BESS was able to charge and discharge at 100% efficiency every day for 20 straight years. This isn’t realistic. So, we’ll now assume the system’s BESS capacity discharges at 90% of its capacity on average after energy losses and other inefficiencies.

The first effect of this adjustment to our assumption set is to increase required nameplate BESS to 72.2 GW (65 GW / 90%). The second is to increase panel capacity dedicated to charging this new BESS by 7.2 GW (72.2 GW – 65.0 GW).

Another adjustment we make in this scenario is to remedy the mismatch in assumed life-expectancy between the panels (20 years) and its BESS (15 years). We’ll reflect this adjustment in the amount of capital required to be invested in BESS at the project’s start.

Specifically, for the replacement BESS in year 15, we assume NREL’s costs (in 2024 dollars) of $225 per KWhr. This cost is below the $300 per KWhr we’ve assumed for the original BESS installation at the beginning of the project, reflecting NREL’s expectations that battery costs will decline over time.

To account for BESS replacement costs, we allocate one-third of these forecasted costs to the final five years (years 15 thru 20) of the project, assuming the replacements also have a 15-year life. This increases the BESS capital costs in 2024 terms from $300 to $375 per KWhr ($300 + ($225 / 15 years x 5 years).

All totaled, BESS’s inefficiencies and replacement costs increase total capital investment by $98.2 B, or 19.1%. This breaks down to an $91.0 B for additional BESS assets and $7.2 B for additional panels. The capacity factors for panels and BESS fall to 24.8% and 11.5%, respectively. Breakeven increases by $11.58 per MWhr to reach $73.88.

Scenario F: Addition of Solar Panel Fixed O&M

It’s now time to account for O&M. So far, we’ve assumed variable O&M for both the system’s panels and BESS to be very low (zero, in fact). We’ll keep this favorable assumption in place. However, fixed O&M is too material to ignore. For our estimate, we (again) depend on findings from NREL.

NREL does a good job framing what costs would typically be included in fixed O&M. For panels, the larger administration and service-line costs are assumed to be insurance, land lease payments, property taxes, cost of inverter specialists, transformers, connection fees, major and minor replacement part costs, inspection, cleaning, electricians, PV array and module specialists, general maintenance, plus other miscellaneous costs.

NREL’s list of costs for the BESS portion of the system is similar to that of panels, but includes battery augmentation costs and excludes inverter costs.

Because the NREL resource allowed, we selected separate mid-point fixed O&M estimates for both panels and BESS. Once again, we chose something more favorable to the economics of the system than what a typical best-guess or mid-point assumption might look like.

As the charts below from NREL show, our annual fixed O&M assumptions of $25 and $15 per KW of nominal capacity per year for the panels (PV) and BESS, respectively, are on the low end of NREL’s midpoint estimates.

Note—NREL pre-selects reasonable default values for several variables when calculating fixed O&M costs. For example, for panels estimated costs are for a “mid (moderate) case over 20 years and mature technology.” For BESS, the assumption is “4-hour utility scale storage and a 20-year life.”

After assuming 2% escalation annually, annual fixed O&M for panels comes out to 2.4% ($7.5 B / $301.8 B) and 0.6% ($1.1 B / $325.0 B) of gross panel and BESS capex on average, respectively. We’d guess most investors would be pleased with overall system-wide annual fixed O&M averaging less than 1.4% of initial capex over the life of a project.

As would be expected, layering in fixed O&M has a deleterious effect on the breakeven of generation, pushing it to $86.55 per MWhr from $73.88 previously.

Note—For economic evaluation purposes, all O&M costs accounted for in PV terms as additional capital investment at construction.

Scenario G: Addition of Realistic Pre-tax WACC

In this scenario, we’ll adjust our assumed pre-tax cost of capital from 7.00% to something more realistic.

First, let’s be clear what we mean by cost of capital. In short, it’s the assumed pre-tax weighted average cost of capital (WACC) consistent with the valuation principles of the Capital Asset Pricing Model (CAPM). It represents the overall cost of capital from all sources—including all forms of equity and debt—needed to finance the system’s construction and operation.

WACC can be thought of as the average blended rate that system developers can expect to pay for all capital employed. It also serves as the “hurdle” rate for discounting unlevered pre-tax cash flows of the system, which are derived via expected-value calculations.

So, what should the WACC be? The first step is to estimate a reasonable capital structure for a PV-hybrid system. Despite our suspicions otherwise, we’ll assume (favorably to the system) that a system of this size can be financed with 60% non-recourse senior debt. The remaining 40% comes in the form of equity.

To determine the rate on the debt portion of our financing, we look to US treasury bonds with 20-year maturities (the life of our project). As of the date of our calculations in the spring of 2024, these are trading at about 4.80%. If we (again generously) assume just a 75-bps spread—or risk premium—to treasuries, this gives a cost of debt for our project of 5.55%.

For their part, we assume equity investors require a basic 10% annual return over the life of the assets.

These two assumptions—after weightings of 60% and 40%, respectively—yield an after-tax WACC of 7.33%. If we assume the project’s cash flows (which are expressed in pre-tax values) are only 10% lower after paying taxes, this suggests a grossed-up (ie, pre-tax) WACC of 8.14% (7.33% /  0.9 ). We’ll round this up to 8.25% as our after-tax WACC, which we suspect is still too low.

The upshot? Adjusting pre-tax WACC to 8.25% pushes the system’s breakeven up $7.06 per MWhr to $91.44.

Scenarios H1 & H2: Addition of Overbuild/Reserve Generation

The final adjustment we must make to our assumption set is a big one—adding sufficient reserve capacity (ie, overbuild) to the system to ensure reasonable reliability under worst-case conditions. As it stands now, there’s no overbuild of either panels or BESS in our assumptions.

First, we determine what a realistic worst-case scenario looks like. We’ll settle on a scenario in which a prolonged period of inclement weather—ie, five consecutive days of very cloudy, rainy and/or snowy daytime conditions—reduces by half the amount of solar power the system generates during a typical peak-demand month. Obviously, real-world conditions could be worse than this, but we’ll stick with this assumption in the spirit of generosity toward solar.

We further assume such an event might occur once every five or ten years—frequently enough for grid stakeholders to agree the frequency and duration of blackouts that would result absent the overbuild would be intolerable.

To simplify the analysis, we assume—again, favorably to solar—that a full third of any investment in overbuild capacity pays for itself via arbitrage, export, or other opportunities arising from having the extra capacity on the system. This means we only burden the project economics with two-thirds of the capital and O&M costs of the overbuild.

We first run the numbers as if only panel capacity was added to the system as overbuild, enough to ensure all demand—including recharging of BESS—each of the five days could be met during the period of inclement weather.

Before excluding 100.6 GW of panels based on the assumption it pays for itself, this scenario—which we’ll label as H1—requires an additional 301.8 GW of panels. Net of the 100.6 GW, the extra cost of the overbuild comes to $201.2 B. Total capital investment thus rises to $828.0 B, pushing breakeven to $127.16 per MWhr.

We then compare this to the breakeven if we only add BESS capacity sufficient to meet the cumulative unmet demand of 3,600 GWhrs that would occur in peak season for the five consecutive days. We’ll label this Scenario as H2. The total comes to 332.8 GW of additional BESS capacity, after excluding 167.2 GW that that is presumed to pay for itself. This leads to total capital costs for BESS of $1.82 trillion, and a total system capital investment of $2.1 trillion. H2’s Breakeven is an astronomical $292.74 per MWhr.

We’ll go with Scenario H1 as the final assumption set for the system.

If 72.2 GW of charged BESS capacity perfectly services 780 GWhrs of demand each night after leakage and inefficiencies, 332.84 GW of charged capacity can be expected to discharge 3,600 GWhrs over five days (3,600 / 780 x 72.2) as needed. This means total BESS required on the system is 72.2 GW + 332.84 GW, or 405.04 GW.

Conclusion
Given what we calculate as a real-world breakeven for generation of almost $130 per MWhr for a PV-hybrid system—even after favorable assumptions—we find claims that solar power is the “cheapest” form of power very likely to be false. How could they be anything but when coal and gas-fired generation could serve this same grid at anywhere from half to a quarter of the breakeven for a PV-hybrid system?

We conclude renewables are like Usain Bolt—a performer that can impress in short bursts, but not in the long run. That means solar’s not the answer to solving climate change or meeting our energy needs any more than sprinting is the right choice for running a marathon. Pretending otherwise can only lead to serious problems.

Linkedin

Facebook

Twitter

Email

Print

Doug Sheridan

Doug Sheridan is Managing Director and Founder of EnergyPoint Research.

See Full Bio