A project comparing the forecasting capabilities of various statistical and machine learning methods applied to renewable energy data
Specifically, I plan to compare methods available in Darts with NGBoost, NBeats, NeuralProphet, and Gaussian processes with spectral kernels.
I will use solar irradiance data in 4 locations in the US: Seattle WA, Los Angeles CA, Denver CO, Rochester NY
I will build the models on a training/test set of data, and compare models using a validation set.
The model comparison will be MSE, MAE, as well as PPC where model uncertainties are provided.