Electric Load Forecasting Model

This repository contains a robust electric load forecasting pipeline built with Python and Scikit-learn.

Features

• Data Integration: Merges load data with external weather factors and holiday events.
• Feature Engineering: Includes 24h/7d lags and rolling averages.
• Cost-Aware Optimization: Uses a custom asymmetric penalty function (4x underforecast, 2x overforecast).
• Multiplier Sweep: Automatically finds the optimal forecast adjustment to minimize total penalty.
• Baseline Comparison: Compares model performance against a naive lag-based baseline.

Installation

pip install pandas scikit-learn numpy

Usage

Simply run the main script:

python load_forecasting.py

Performance

The model consistently outperforms the naive baseline (Lag-96) by significant margins in terms of total penalty.