Originally posted on S&P Platts.

A study of approximately 30% of utility-scale and commercial and industrial solar generation operating in the US between 2016 and 2019 showed the average system underperforming by 6.3% compared to what the performance was expected to be during financing, according to solar risk management company kWh Analytics.

The San Francisco-based consulting firm noted that reliable production forecasts "are the cornerstone to solar financeability." The report said 30% of the fleet that was studied, "a quarter of the projects missed their production targets by over 10%, even after weather-adjustment."

"Fundamentally, it is our hope that this report serves as a platform to discuss the data-driven approaches to inform deployment of capital,"said Richard Matsui, CEO of kWh Analytics, in a statement accompanying the report.

The kWh Analytics report is titled "2020 Solar Generation Index," and was released Oct. 5. It noted that every asset class is governed by market cycles and modeling assumptions, and those assumptions "naturally swing between optimism and conservatism."

There is "optimism that naturally emerges from market growth [that] can inadvertently undermine the long-term stability of the industry as a whole."

The solar asset class has now generated a decade of actual data, kWh Analytics said, that can be used to guide sustainable growth, and some of the data "affirms positive attributes."

But some of the data also reveals "that we have collectively turned a blind eye to realities of solar asset performance," the report said.

P50 expectations

The consulting firm said its study was a "coordinated initiative" with 10 of the 15 largest asset owners in the US.

"Combining the contributed data and kWh Analytics' HelioStats database, this analysis encompassed over 30% of the industry's non-residential solar projects across more than 30 different asset owners," the report said.

It said that all projects analyzed were larger than 1 MW in DC capacity.

The analysis compared data relative to P50 expectations. The P50 estimate is a statistical measure that indicates the base case of predicted energy yield.

"P50 expectations were degraded annually based on the annual degradation factor assumed by the asset owner," the report said. "The results do not incorporate system losses due to utility curtailment."

Stakeholders in a project can be "financially motivated" to increase production estimates, as it is directly correlated to the amount of capital that can be raised. Moreover, developers may be concerned about the near-term impact on solar asset valuations resulting from adjusted P50 estimates.

The analytics group said it believes a "course correction" will enable the solar industry "to continue the structural trend of lower-cost capital entering into the industry."

Real-world data

The report contends that most in the solar industry have used the same software tools and engineering firms to generate production estimates.

"The lack of accepted standards means that production forecasts can vary dramatically depending on who is running the model," it said.

Until now, data has been unavailable to validate production forecasts with real-world operating data at scale, the report said.

"This research highlights the need to bring this real-world data into the project evaluation process to meet investment return expectations," Matsui said.

Predicting energy yields for generation sites did not begin with solar: other generating assets like natural gas and wind also rely on modeling by engineering firms to assess energy yield potential.

"When lenders began doing their due diligence on solar projects, they often borrowed the same 'playbook' to model their exposure and return expectations off of predicted production estimates," Hao Shen, head of data products at kWh Analytics, said in an interview Oct. 7.

"What started as the concerns of a few asset managers has evolved to be the topic du jour at conferences and in the boardroom," Shen contends. "However, these conversations to date have largely been based on anecdotal evidence from single projects or portfolios."