Description
Transposition (forward) is the calculation of global irradiance on a tilted surface (GTI), which is also known as plane-of-array (POA) irradiance. The calculation requires the standard values of diffuse horizontal irradiance (DHI) and direct normal irradiance (DNI) or global horizontal irradiance (GHI) in addition to the solar position angles and orientation of the tilted surface. The Perez diffuse sky model [1] is one of the more well-known models used for transposition calculations in PV system simulations and yield assessments. There are other situations where a POA measurement is available but a GHI value is needed, for example for comparisons with long-term averages or other systems. This occurs more frequently in PV system monitoring applications. The calculation of GHI from POA is called reverse transposition and usually requires some kind of search or optimization algorithm. A simple one is built into PVsyst for evaluations when only POA data are available. GTI-DIRINT is another example of a somewhat more complex algorithm for reverse transposition implemented in other software [2]. The strength of the Perez model lies in the table of empirical factors that summarize the full range of anisotropic sky conditions that can occur. The use of discrete empirical factors results in discontinuities or small jumps in irradiance values at the transition between sky condition categories. These are primarily visible in high-resolution data found in monitoring applications, where similar jumps in performance ratio can then be observed. The discontinuities have a more adverse effect in the reverse transposition application, where they cause conventional optimization algorithms to fail. In this work we iron out these wrinkles in the Perez model by making the entire parameter space of sky conditions continuous (including the first derivative) using mean-preserving quadratic splines. We validate that the performance of forward transposition calculations remains as before, allowing the new version to be used as a plug-in replacement for the original. Then we demonstrate that reverse transposition can be reliably done using unsophisticated optimization algorithms like the bisection search, and clearly delineate the narrow range of conditions where reverse transposition is impossible without additional information (or assumptions). Both convergence and accuracy of the reverse model are substantially better than GTI-DIRINT. GHI derived from GTI measurements is also significantly more accurate than GHI from satellite source–and it can be obtained at higher time resolutions as well. In addition to applications in PV system monitoring and performance evaluation, new opportunities for site adaptation and satellite model validation, as well as processing multi-sensor data are being investigated. The algorithm will be implemented in pvlib-python in summer 2023.| Period | 21 Sept 2023 |
|---|---|
| Event title | 40th European Photovoltaic Solar Energy Conference and Exhibition |
| Event type | Conference |
| Conference number | 40 |
| Location | Lisbon, PortugalShow on map |
| Degree of Recognition | International |