1603 Data Lifts the Veil On PV System Performance

In the world of photovoltaic (PV) finance, historical performance data—or the lack thereof—stands as an obstacle to the risk assessment and credit evaluation of solar installations. Investors eyeing solar-backed securities or other investment vehicles might view the asset class cautiously without a firm, data-driven understanding of how solar assets perform (i.e., generate power) in the field and how this affects the cash flows that would ultimately backstop their investment.

However, the National Renewable Energy Laboratory (NREL) has a trove of data which suggest that PV systems are, in the aggregate, performing better than expected in the field. These data are from systems installed using "grant" money from the 1603 Section of the American Recovery and Reinvestment Act which allowed for the conversion of investment tax credits into cash rebates. The data describe nearly 50,000 PV systems from 0.5 kW to 25 MW in size (with a median of 6 kW), representing a total of approximately 1.7 GW of installed capacity (see Figure 1 for the geographic distribution of these installations). Data points for each system include predicted yearly production values, actual yearly production values, zip code location, and comments regarding performance issues for each year of production.

map showing geographic distribution of 1603 systems in this analysis

Figure 1. Geographic distribution of analyzed installations

How much better are these systems performing in the field than they were predicted to by independent engineers and developers? As shown in Table 1, they are producing on average 2% - 5% more energy than their targets. 

Table 1. Median Ratio of Measured Over Predicted Values for All Data
Median Measured/Predicted Year 1 Year 2 Year 3 Year 4
All Data 1.04 1.05 1.04 1.02

Figure 2 displays the proportion of normal systems to systems that experienced identifiable issues. NREL has classified these issues into four types:

  • Project: Includes systems that incurred site-related issues, such as permitting delays, incidents where the PV system had to be removed because the roof was damaged (but not the PV installation), or the sale of a property
  • Hardware: Denotes systems with equipment problems, such as inverter outages.
  • Weather: Indicates problems associated with storm-related shutdowns, snow remaining on the systems, etc.
  • Data Collection: Comprises systems that had issues with the transmission of data and thus do not reflect the actual performance of the installed system.

As shown in Figure 2, these issues affected less than 10% of projects in any given year, and became less prevalent as projects got older. Typically, the incidence of problems—e.g. project delays, out-of-the-box hardware flaws, and others—is highest in the initial months of operation. Actual performance-related issues –those that arose from system functioning and not as a result of initial failures—total only 1%–3% of all systems for each year.

Note: the data points for this analysis decrease with each year, which reflects the newness of many of the 1603 systems. Only about 480, or less than 1% of the 48,259 systems in the dataset, have been operational for over four years.

Four Pie charts showing the percent of analyzed systems without any issues, and the percent of systems that experienced, Project, Hardware, Weather, or Data Collection issues by operational year

Figure 2. Charts of data for systems with and without issues partitioned by type of issue and years after installation

Figure 3 shows measured production data versus predicted production data for normal systems (black-filled circles) and for systems with issues (colored symbols). Because data collection issues do not necessarily reflect the performance of the PV system and can have a broad impact range depending on how soon the issue is resolved, they have been excluded. A unity line for the normal-system data is shown as a guide to the eye.

Four scatter plots showing the distribution of system production in analyzed systems and the effect that issues have had on that production

Figure 3. Measured production versus predicted production for normal systems (red-filled circles) and for systems with issues (colored symbols). The "Data Collection" category has been excluded from this figure.

Project-related data points tend to lie farther away from the unity line. Therefore, although project- or site-specific issues may not occur as frequently as other categories, when they do occur they tend to have a larger impact on production than, for example, weather-related underperformance. In the first year, two-thirds of the project category anomalies are due to delays in permitting and constructions where cash flow has not yet commenced. In subsequent years, delays decrease to inconsequential levels, and the project category becomes dominated by operational issues.

In conclusion, the data indicate that the large majority of systems operate under "normal" conditions, and to date these systems overproduce relative to the predicted value. And, at 1.7 GW of capacity—which is 13% of the roughly 13 GW of total PV U.S. capacity at the end of 2013—these 1603 projects can reasonably be viewed as a representative cross-section of the U.S. PV market.