SPLTRAK Abstract Submission
Evaluation of Solar Capacity Factor of ~2000 Solar Plants Across the United States Using Multilayer Perceptron Regressor Models
Samantha S Wilson, Stephen Lightfoote, Stephen Voss, Brian Joseph
Power Factors, Brossard, QC, Canada

Capacity factor is the ratio of a solar plant's theoretical maximum energy output to its actual energy production over the course of some measure of time. Solar capacity factor is also fundamentally limited by the availability of solar power. Since plants cannot produce power at night it is generally impossible for them to have capacity factors exceeding 35%.  This also means capacity factor is limited by anything that impacts a plant's solar resource including location and plant design.
Before a plant is built its capacity factor is assessed through pre-build analysis based on projected understandings of local weather conditions and plant design. However, pre-build modeling is based on our understanding of basic climate data and the physics of specific plant design. A detailed understanding of the local micro-climate is difficult to capture in modeling.
Due to the limitations of physical models for assessing microclimate, we have attempted to generate a model for capacity factor entirely based on measured capacity factor for ~2000 existing sites and climatology data from the NASA POWER API. We have limited our analysis to the continental United States and have attempted to generate a capacity factor prediction algorithm using machine learning.
We have also compared the model fit of measured capacity factor to the modeled fit of budged capacity factor. By comparing these two models we conclude that the budgets for the Eastern United States may be more inaccurate than budgets for the Western United States.