Solar photovoltaic (PV) module and system degradation continues to be an important and unresolved variable in the PV industry.* One of the principal uses for degradation data is the calculation of levelized cost of energy (LCOE)—a metric used by actors across the PV value chain for decision-making purposes (e.g., benchmarking PV against other sources of generation, and setting power purchase agreement [PPA] rates). One of the most common ways to input degradation in an LCOE model is via a flat rate (0.5 percent per year is a typical default value), which discounts the nameplate capacity cumulatively each year of project life. However, the available degradation data do not necessarily privilege such an approach, and this begs the question: is the industry standard for degradation accounting in project finance models delivering the most accurate LCOE?
A new journal article produced by the National Renewable Energy Laboratory with the Colorado School of Mines and DNV GL offers some perspective on this question. The article—”Compendium of Photovoltaic Degradation Rates,” published in Progress in PV—is an update on degradation research that has been ongoing for several years. But in the appendix, the authors introduce the concept of the “degradation curve,” the shape of the decline in panel performance over time and its potential effects on solar LCOE. Traditional degradation rate modeling assumes a linear curve over time, with the same chunk of energy loss occurring year after year, in stepwise fashion. But what if the curve were not linear? What if it were exponential or stepwise? Moreover, what if the assumption that the nameplate capacity serves as the baseline for the degradation rate is problematic? Perhaps it would be appropriate to discount the nameplate—say 5 percent because of consistently occurring losses from, say, manufacturing defects or workmanship flaws—and base the degradation rate and curve off of that figure?
In the appendix to the article, the authors detail the results of a Monte Carlo analysis they conducted to study the effect of various degradation rates, curves, and baselines on the LCOE for a 100-kW plant in Colorado. The four scenarios were:
- Linear degradation at 0.5 percent per year of nameplate capacity (this represents a sort of industry “standard”)
- Exponential degradation
- Two-step degradation, where performance remains flat until year 12 and then drops off linearly
- Linear degradation at 0.16 percent per year of 90 percent of the nameplate capacity
Figure 1 displays each of these curves, illustrating their effect on nameplate capacity over time. Note, the curves all meet at the same point by year 25 (at around 87 percent of nameplate capacity).
Figure 1. Different degradation rate input curves as a function of time. Linear degradation at a rate of 0.5%/year (red circles), an exponential decline (blue diamonds), a two-step profile (green crosses) and a linear decline (orange crosses) starting at 90% of nameplate rating are shown.
Figure 2 shows the most important factors affecting LCOE in a sensitivity analysis. The point at which the curves converge in each degradation scenario represents the analysis mean, or the most likely cost of energy given those particular degradation conditions. Two-step degradation (the green curve in Figure 1) produced the lowest LCOE at 10.2 cents/kWh, with the standard 0.5 percent linear curve coming in second at 10.5 cents/kWh. The most significant factors in the analysis are the discount rate and the initial cost, assumed to have an average of $1.60/watt for a 100-kW utility-scale installation.
Figure 2. Spider plot of the impact of all input parameters on LCOE. The four different compartments represent the four different degradation curves that are labeled on top. In addition, the mean for each compartment is given.
These plots raise a host of questions: How would stakeholders in the PV value chain be affected if field experience tells us that degradation behaves, say, more exponentially than linearly? Will solar project sponsors experience compressed margins as their production falls out of line with anticipated levels and their per-kilowatt-hour revenue is not sized to cover the gap? Will manufacturers have to alter their warranties, insurance coverage, marketing, etc.? Will operations and maintenance service providers retool their scopes and cost structures? Will lenders and other financiers reconcile a variety of degradation rates, curves, and baselines and size their investments accordingly (similar to the way lenders base debt size on a spread of exceedance probabilities)? In the competitive power business, a delta of 0.8 cents/kWh—as shown between the highest and lowest cost scenario means in Figure 2—could make the difference between a winning bid and a losing bid in a market where cheap gas is on the margin.
While this analysis may offer more by way of questions than answers, the questions are provocative and present something of a challenge to conventional practice of PV project modeling. The linear method of calculating the effects of degradation may provide a reasonable initial approximation of LCOE, but how should alternative methods be accounted for? And, as we learn more about how PV generators actually perform and degrade over time, will compensation rates for those generators change accordingly? Accommodating such changes could entail several shifts within the PV value chain. More data are needed before any such shifts would be required, but it is something the industry want to have on their radar.
*The impacts from degradation in larger PV systems have been somewhat mitigated in recent years as installers increasingly oversize the DC side. This means that the inverters (and not the module capacity) limit output—a practice known as “clipping.” With a system overbuild on the DC side, the effects of degradation are masked (because there is additional backup capacity compared to a 1:1 DC to AC build) though it does not affect the degradation rate.
This article was originally published by the National Renewable Energy Laboratory and was republished with permission.
Lead image credit: NREL