In-situ Atmospheric Intelligence for Hybrid Power Grids: Volume 5 (Ambient vs. Panel Temperature in Photovoltaic Power Modeling)
Abstract
Reducing the vulnerability of remote electrical resources can be advanced through power diversity. To optimize this diversity, one strategy is to secure advanced knowledge of key atmospheric parameters. When solar energy (photovoltaic [PV] panels) is integrated into a hybridized grid, the critical meteorological elements include 1) solar radiation, which defines the maximum potential PV power production, and 2) PV panel temperature, which impacts power production efficiency. Examining temperature as a function of PV power generation, the study asked 1) Can ambient temperatures be used for PV panel temperatures? and 2) What is the optimal data reach-back and reach-forward for a neural network PV panel-temperature forecast model? Results showed that under overcast sky, the two temperatures performed equivalently. Under clear sky, grid optimization routines will need to assimilate the potential for midday excursions from actual power generation the model using ambient temperature overestimated measured power, whereas using PV panel temperatures underestimated measured power. For neural network PV panel-temperature forecast modeling, the optimal reach-back/-forward was defined by temperature input thresholds. The next steps are to test the temperatures in other PV power models, and experiment with the hidden layer number, layer population, and/or neutral net architecture.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 20, 2022
- Accession Number
- AD1169411
Entities
People
- Gail Tirrell Vaucher
- Jessica Whitaker
- Robert A Jane
Organizations
- Howard University
- United States Army Research Laboratory