Compare the right alternatives
The analysis compared linear irradiation, quadratic irradiation and a model that also includes temperature interactions. The quadratic comparison matters because it asks whether additional temperature terms help beyond a more flexible irradiation curve.
A result that changes by plant
Plant 1’s final-week nMAE was 1.97% for quadratic irradiation and 1.94% with temperature interactions. Plant 2’s was 7.28% and 7.57%, respectively. Lower is better. The extra features offered little improvement in the first case and worsened error in the second.
The transferable research habit
Use a baseline that represents a serious alternative. Keep evaluation separate, report the unsuccessful comparison and state the limits of the data. This habit is relevant to SBX’s future diagnostic study, but these generation results do not validate that different task.
Continue the research.
Explore the chronological comparison, model errors and selected evaluation code.
Explore the solar models and results