Build a regression model for wind power output

Build a non-time-series regression model to predict turbine-level wind power output for individual snapshots using available weather and turbine features. Describe: candidate features (e.g., wind speed/direction, air density, temperature, humidity, turbulence intensity, turbine specs) and preprocessing; handling missing or noisy sensors and outliers; model choices (regularized linear models, gradient boosting, random forest, shallow MLP) and how you would encode known physics (e.g., approximate power curve features) without sequence modeling; validation strategy across sites and wind-speed regimes to ensure generalization; evaluation metrics (RMSE, MAE, MAPE) and treatment of heteroscedastic errors and caps at rated power; methods for uncertainty estimation and calibration; and safeguards for extrapolation and curtailment scenarios.