Fortum Energy Consumption Prediction

LightGBM models for predicting electricity consumption:

• 48-hour ahead predictions (48hr_LightGBM.ipynb)
• 12-month ahead predictions (12m_LightGBM_model.ipynb)

Setup

Install dependencies:

pip -m venv .venv
pip install -r requirements.txt
.venv/Scripts/activate

Data

Ensure that the CSVs are in the project root.

Running

48-Hour Prediction Model

jupyter notebook 48hr_LightGBM.ipynb

Run all cells sequentially. The notebook:

1. Loads and preprocesses data from merged.csv
2. Creates time and lag features
3. Tunes hyperparameters with Optuna
4. Trains final LightGBM model

12-Month Prediction Model

jupyter notebook 12m_LightGBM_model.ipynb

Run all cells sequentially. The notebook:

1. Aggregates hourly data to monthly
2. Creates monthly features and lags
3. Trains 12 separate models (one for each month ahead)
4. Models saved in 12 month prediction model weights/

Model Weights

• 48hr_weights.txt - 48-hour model weights
• 12m_weights/ - 12 monthly model files (lightgbm_model_1.txt through lightgbm_model_12.txt)