Arya Sankhalkar completed Energy and Renewable: Machine Learning for the Power Grid

Built a turbine TEY predictor end to end with leakage-free inputs, a genuine time-based holdout, and a baseline-vs-forest comparison. The final RandomForest model improved held-out MAE from 2.90 to 1.53 MWh and was evaluated with MAE, RMSE, R2, and residual analysis on unseen later-period data.

• Leakage-Free Feature Selection
• Time-Based Holdout Validation
• Honest Held-Out Evaluation
• Residual Pattern Reading