The idea
Solar generation changes with weather conditions. This undergraduate research project investigates how historical energy and meteorological features can support useful power-output forecasts.
The approach
I worked through the complete modeling pipeline: cleaning and preparing data, exploring relationships, engineering features, comparing regression models, and tuning hyperparameters.
My contribution
- Compared LightGBM, ExtraTrees, and Ridge regression models.
- Evaluated predictions using MAE, RMSE, and R².
- Examined how weather variables, including solar radiation and temperature, relate to generation.
- Presented the research at a student research conference.
What I learned
The work connects machine learning with an energy-system problem, emphasizing evaluation and interpretation alongside predictive performance.