AI FORECASTING TERMINAL XGBoost v3.2

Prediction Engine

MODEL ACTIVE
current

72-Hour Demand & Emissions Forecast

48H HISTORY + 72H AHEAD
Historical (solid)
AI Forecast (dashed)
95% Confidence Band
Peak forecast: 22,140 MW @ 20:30
Current Model
XGBoost Regressor v3.2
256 trees depth: 8 lr: 0.05
Model Accuracy
94.2 %
R² on 30-day holdout validation
MAE 312 MW
MAPE 1.8%
Last Retrained
2 hours ago
Auto-retrain on fresh CAMMESA dispatch
Ingest
Clean
Train
Valid
Deploy

Feature Importance — Drivers

SHAP VALUES
Temperature Ambient temperature drives heating/cooling load
45%
Industrial Activity Manufacturing output and heavy industry cycles
30%
Seasonality Day-of-week and monthly seasonal patterns
25%
Wind Generation Patagonia wind output offsets thermal dispatch
18%
Grid Frequency System frequency deviations indicate stress
12%