Short-Term Solar and Wind Energy Forecasting Using Machine Learning and Deep Learning
Developed a multivariate time-series forecasting pipeline using weather and SCADA variables from two solar PV plants and one wind turbine. I compared statistical, machine-learning, and deep-learning approaches, including ARIMA, Random Forest, XGBoost, LSTM, and GRU.
I also evaluated model behavior under missing-data and Gaussian-noise scenarios to examine how dataset size, temporal structure, and data quality affected forecasting performance. The project provided the starting point for my current UROP work on larger, multi-site forecasting problems.