Utility-Scale Solar PV Statistical Analysis
Originally published on CleanTechnica.
Deployment of utility-scale PV power in the United States in recent years has been rapid and noteworthy. It has resulted very dissimilar photovoltaic operating projects whose empirical AC capacity factors differ by more than a factor of two, however. Understanding the performance of utility-scale PV power is important because profitability depends directly on how well projects perform over time. The sector can only raise investment capital if the long-term profitability factors are favorable.
Under the title ”Maximizing MWh,” scientists from the Lawrence Berkeley National Laboratory and the University of California at Berkeley’s Goldman School of Public Policy contributed a valuable statistical analysis of American utility-scale photovoltaic projects this month. The report is the first known used of multivariate regression techniques to analyze empirical variation in project-level performance.
The 128-project (3201 MW) sample split almost evenly between fixed-tilt (63 projects, 1776 MW) and tracking (65 projects, 1776 MW) projects. Its findings give solar project developers and investors a good indication of what they can expect from the different project configurations used in different regions of the country. Also, through this model’s tight relationship between actual and fitted capacity factors, investors can gain confidence that the projects in this sample have largely performed as expected.
Authors Mark Bolinger, Joachim Seel, and Manfei Wu analyze the independent variables and other factors responsible for the variations:
“The regression models developed for this analysis find 92% of this variation caused by only three highly significant independent variables”:
- Solar resource strength, in terms of average annual global horizontal irradiance (“GHI”) estimates (note: GHI alone explains 71.6%);
- Tracking, which increases “plane of array” irradiance; and
- Inverter Loading Ratio (“ILR”), which boosts AC capacity factor.
Adding a fourth independent variable (project vintage, or COD Year) and 3 interactive terms (Tracking x GHI, Tracking x ILR, GHI x ILR) improves the model further. Good data on power temperature coefficients and module operating temperatures might also improve calculations.
The study did not examine orientation (tilt and azimuth) and temperature because of limited reliable data on coefficients and temps. It’s also worth noting that 83% of MW and 66% of projects are in the “top 5” states for NCF and GHI; California alone accounts for 49% of capacity and 32% of projects; and single-axis tracking prevails in the high-GHI states.