Load Forecasting

Short-Term Electricity Load Forecaster

Interactive day-ahead load forecasting model using the key features identified in the literature: weather, seasonality, and historical load patterns.

Architecture based on Willingham (MathWorks, 2011). Feature selection informed by recent LSTM/CNN-LSTM literature (2024–2025).
Weather Forecast
Dry Bulb Temperature22 °C
Dew Point14 °C
Cloud Cover40%
Seasonality & Calendar
Day of Week
Month
Holiday
No
Historical Load Context
Base Load Level5000 MW
Load Growth Trend0%
Forecast Metrics
24-Hour Load Forecast
Forecast (MW)
Previous Day Actual (MW)
Temperature (°C)
Temperature Sensitivity — Forecast vs Temperature at Peak Hour
Predicted Peak Load (MW) vs Dry Bulb Temperature
Feature Importance — Contribution to Load Prediction
Feature importance is based on coefficient magnitude in the linear regression model. In production LSTM/CNN-LSTM models, SHAP values are used for interpretability. Temperature and hour-of-day are consistently the most impactful features across all methods in the literature (DWT-LSTM achieves MAPE 0.3–4.2%).