Solar photovoltaics (PV) are taking the world by storm – 75 gigawatts (GW) of global installations in 2016 was the latest record in a decade of exponential growth averaging 34% per year. But PV estimates have consistently and severely underestimated growth, while future estimates vary widely, making planning decisions almost impossible and risking billions in the wrong type of power sector investments.
If the question today is “will this solar surge continue,” and the forecast answers are consistently wrong, what’s a utility executive or investor to do? The already significant levels of PV deployed today mean mistaken expectations about the near-term future of solar can lead to faulty planning, continued investment in imprudent assets that won’t be able to compete, or missed opportunities to invest in the wake of a winning technology.
Fortunately, one simple method of data extrapolation – the logistic function, or S-Curve – improves future planning accuracy. Applying the S-curve predicts much faster near-term growth than current expectations, which will challenge utilities, grid planners, and competing generation sources. But, while PV installers and vendors will see a faster boom than expected, they must also prepare for “peak solar” installation point occurring between 2023 and 2028, albeit at a rate of two to ten times current installations (145 to 720 GW/year).
Two Common Biases Cloud Solar PV Predictions
Recent research in the journal Nature Energy found that a combination of strong policy support and steep technological learning were crucial to historical PV deployment consistently exceeding forecasts.
This gives the lie to two common biases in various outlooks about future PV deployment. The first stems from thinking that solar had its early run, but installations must now stabilize as multiple challenges to growth appear. These challenges include diminishing subsidies, the need to integrate variable generation resources into the existing grid, displacing existing generation instead of meeting new demand growth, and the significant need for cheap capital – especially in developing countries.
This view encourages conservative linear growth forecast like the International Energy Agency’s (IEA) Medium-Term Renewables Energy Market Report 2016 – without much room for installation costs to drop.
The second bias is that annual solar deployment will keep deploying exponentially with a fixed compound annual growth rate (CAGR). Last decade’s 34% CAGR for total installations imply similar CAGR for annual installations (i.e. the nature of exponential growth). Instead, modelers tend to switch to the future CAGR of annual demand (market-size analysis) with values around 5%-10%.
This is well-suited for bottom-up outlooks like GTM Research’s Global Solar Demand Monitor (projectong 6.7% Annual Install CAGR for 2017-2022) which provides good short-term intelligence but fails to account for how that short-term success enables additional success further on. The rapid change in assumptions for annual installation CAGR has two problems: It underestimates medium-term growth (3-12 years out), and it maintains a rosy view of endless steady growth out into the distant future for solar installers.
Introducing The Solar Deployment S-Curve
One way to solve the extremely difficult problem of forecasting solar deployment is the logistic function or S-curve – a simple method that enables learning by interpolating between past exponential growth during PV’s early adoption phase and its eventual saturation. For example, a 2016 report on the “Current and Future Cost of Photovoltaics” by the Fraunhofer institute (FI) used S-curves to correct for the distortion of annual install CAGR.
The S-curve can be used directly as a simple and powerful way for thinking about the future of solar, but don’t worry, this won’t require parsing any equations! The S-curve describes a very common pattern of deployment among many technologies with initial exponential growth, then steady growth and eventual saturation, with three parameters. For PV’s purposes, two parameters are fixed by the initial exponential curve for solar deployment, and the last one is set by our final expectation for the saturation level of solar – how many GWs of deployment it will roughly reach.
Where Solar PV Goes Depends On Where You Think It Will End Up
So, to peer into the future with S-curves trend analysis, we must estimate how much PV will meet world electricity demand in 2050. Fraunhofer’s analysis averaged IEA scenarios to reach total global electricity demand of 43 million gigawatt-hours (GWh). Assuming PV is only 25% efficient, that’s what you would expect to generate from just under 20,000 GW of solar.
Our back-of-the-envelope analysis looks at three scenarios for future growth. The most aggressive has solar provide half of total generation: 10,000 GW. Keep in mind that with Fraunhofer’s “high electrification” scenario – where other fossil demand like transportation fuels is replaced by electricity for a total demand of 100 Million GWhs – the high solar scenario could end up as much less than half of all generation.
Our two other scenarios peg solar at 10% and 25% of eventual global generation by 2050, with global installation numbers at 2,000 GW and 5,000 GW respectively. Analyzing these 2050 possibilities with the S-curve yields the following extrapolations to 2030.
What The Solar Deployment S-Curves Tell Us
Looking at our S-curves, we immediately see that all scenarios quickly exceed IEA’s flat outlook to 2021. The lowest scenario (green line with 2,000 GW in 2050) also surpasses GTM’s 2017-2022 bottom-up outlook. Meanwhile the other two scenarios take off, with roughly double total global PV installations in five years as GTM’s outlook and almost triple IEA’s numbers.
Extending the GTM outlook (gray dashed-line) with the same 6.7% CAGR in annual installs points to another important feature of the S-curves: The extended GTM outlook eventually catches up with the lowest scenario in 2027, because this scenario maxes out annual solar deployment in 2023 at 145 GW/year (a little more than twice 2016’s total). S-curves always slow down and cannot provide the constant CAGR in sales that most companies desire. The top scenario envisions break-neck growth with 20-80 GW of new panel fabrication capacity coming on every year for a decade – but even that peaks in 2028 at just under 10 times the current global solar installation market. This means market participants should brace themselves for “peak solar” within the next decade.
Faulty Solar Forecasts Create Risks For Developers, Utilities, And Governments
A global PV deployment forecast like the S-curves above has important implications for investors, regulators, planners, and customers. The first implication is cost. Lawrence Berkeley National Lab’s Utility Scale Solar 2016 report shows U.S. PV prices are dropping fast. Depending on the experience curve you choose (15%-30% drop per doubling), we can expect 2023 PV prices to drop another 50%-80% on the most aggressive scenario and 40%-70% in the least aggressive scenario.
The cheapest unsubsidized levelized cost for PV in top locations today is just below $60/megawatt-hour (MWh) in the U.S. and $24.2/MWh in the United Arab Emirates. Price declines like those above imply that in the next 5-10 years we are likely to see widespread unsubsidized prices for solar in the $20-$30/MWh range, easily outcompeting the marginal cost of power for coaland natural gas generation. This should concern any investor in new fossil generation, especially in sunnier locations!
Planners and regulators should not drive blind with outdated forecasts, but should consider how current and likely low prices will affect their resource portfolio in the near future, and use market-based solicitations like reverse-auctions to integrate the latest pricing drops into their decisions.
Even investors and companies riding high on the PV tide should heed the S-curve lessons. While faster near- to mid-term growth will improve business opportunities, accelerating price declines will also squeeze profit margins. Innovation leading to a 1% improvement in solar capital costs or performance translates into 4-5 cent per watt (c/W) of improved financial performance in 2010, 1c/W today, and even less tomorrow.
Eventually the PV tide will no longer lift all boats. Companies that live or die based on annual revenue growth will face difficulty in an environment with the sharp cost declines and flat or declining installation the S-curves predict in just a decade: Fat times today, lean ones tomorrow.
By Eric Gimon, Energy Innovation’s Senior Fellow