Overestimation of solar output
The solar industry has anecdotally begun raising concerns about whether solar power plants are underperforming compared to their P50 output forecasts.
What began as hushed conversations at industry conferences is now widely discussed and analyzed. Individual engineering firms and asset owners are beginning to review their portfolios to assess whether or not their original P50 forecasts were accurate.
DNV GL published a piece in the annual “Solar Risk Assessment” report identifying a 3% to 5% overestimation bias in P50 forecasts, even after adjusting for weather. NextEra published a technical discovery around biases in hourly-resolution energy predictions that overestimate solar resource availability. Behind closed doors, asset owners will also acknowledge struggles to hit P50 figures as consistently as the definition attributes.
Under a P50 forecast, a project is supposed to have a 50% chance of performing at least as forecast. This figure is the base case for the project and is generally the most optimistic projection used in financings. Financiers also run sensitivities by looking at other forecasts — for example, P99 and P90 — as well. A project should have a 99% chance of performing at least at the P99 forecast, if not better.
Generating a production estimate integrates weather forecasting and equipment performance expectations into complex physics models. As with any technical model, results vary based on the assumptions used.
kWh Analytics collaborated with 10 of the top 15 asset owners in the United States to conduct the industry’s largest cross-sectional energy validation study, quantifying the accuracy — or inaccuracy — of solar projects’ P50 estimate. We looked at data from 30% of the operating utility-scale and distributed solar capacity. The results are reported in an inaugural “2020 Solar Generation Index” report.
Projects on average underperformed by 6.3%, even after adjusting for weather.
This means that actual performance of the US solar fleet is closer to P90 expectations than the P50 definition used by project stakeholders.
It is important to note that while 6.3% underperformance is the average, there is a wide distribution that highlights significant variability among projects. In the bottom quartile, projects are falling more than 10% below forecast while the top quartile performers are meeting their P50 expectations. As a result, we can see that each project is indeed unique, even if the general trend points towards a 6.3% bias.
The issue of energy estimation is not unique to solar. The wind industry similarly struggled to align lenders, owners and operators on expectations around energy output and is still developing tools to address accuracy and biases.
Implications for Shareholders
If unaddressed for solar, systemic asset underperformance can have serious implications for the equity holder cash flows, investor returns and the long-term financeability and credibility of solar as an asset class.
The impacts are discernible from day 1 of operation.
For an equity investor or sponsor who sits last in line behind the tax equity and debt, P90 performance realities mean equity cash yields are cut in half for the life of the asset. For lenders, given the prevalence of P90 scenarios, underproduction poses a risk to debt coverage.
As a risk management company that enables insurers to provide all-risk production coverage to solar assets, kWh Analytics is also observing this trend firsthand through claims against a “solar revenue put” product that actual output will be at least at a guaranteed level. (For more information about solar revenue puts, see “New product: solar revenue puts” in the October 2016 NewsWire.)
To date, insurers have continued to pay all claims in full within 30 days and remain committed to providing sponsors with credit-enhancing insurance products.
However, if unaddressed, inaccurate production estimates and return uncertainty will have long-term consequences for the solar industry.
Every major asset class leverages market data to improve the accuracy and certainty of investment returns. If we look at other mature asset classes like consumer credit or mortgages, companies like Experian and CoreLogic exist to provide market data to validate asset performance and modeling assumptions for investors. Solar is at an inflection point now where we have more than a decade of asset performance data that can be leveraged to inform diligence and improve operating assumptions.
kWh Analytics is using its industry database to offer objective market comparables to evaluate expected yield and performance estimates for pre-construction and operating plants. This new offering, the Solar Technology Asset Risk (STAR) Comparables Report, equips deal teams with historic performance of similar plants to help evaluate performance and financial risk of their projects. In addition, it has helped asset management teams contextualize their portfolio’s performance against projects in the field to improve O&M and asset management strategies.
The solar industry has generated the data required to improve the forecasts. The next step is to leverage that data in investment decisions.