alanoee/energy_prediction_model

Energy Consumption Prediction Model

A Random Forest model for predicting household energy consumption patterns and costs.

Model Description

This model predicts monthly energy consumption in kWh and associated costs in PLN (Polish Złoty) based on historical consumption patterns and seasonal features.

Model Type: Random Forest Regressor
Framework: scikit-learn
Performance: R² = 0.848

Features

The model uses 17 engineered features including:

• Moving averages (3-month and 6-month windows)
• Lag features (1, 2, 3 months back)
• Seasonal indicators (winter, summer, transition periods)
• Temporal features (month, year, day of year, quarter)
• Cyclical encoding (sin/cos transforms for monthly patterns)

Usage

import sys
sys.path.append('.')  # Add current directory to path

from model import EnergyConsumptionPredictor

# Load the pre-trained model
model = EnergyConsumptionPredictor.from_file('energy_model_latest.joblib')

# Make predictions for next 6 months
predictions = model.predict_future(months=6)

# Display results
print(predictions[['Date', 'Predicted_Consumption', 'Predicted_Cost']])

Output Format

The model returns a pandas DataFrame with columns:

• Date: Month start date
• Predicted_Consumption: Predicted consumption in kWh
• Predicted_Cost: Predicted cost in PLN
• Month: Month number (1-12)
• Year: Year

Requirements

pandas>=2.0.0
scikit-learn>=1.3.0
numpy>=1.24.0
joblib>=1.3.0

Model Training Data

The model was trained on residential energy consumption data with:

• 17 data points spanning multiple months
• Features include seasonal patterns, consumption history, and temporal indicators
• Target variable: Monthly energy consumption in kWh

Performance Metrics

• R² Score: 0.848
• Model Type: Random Forest (100 estimators)
• Cross-validation: 3-fold CV used for model selection

Feature Importance

Top 5 most important features:

1. consumption_ma_3 (3-month moving average)
2. consumption_ma_6 (6-month moving average)
3. consumption_lag_1 (1-month lag)
4. consumption_lag_3 (3-month lag)
5. month_sin (seasonal encoding)

Cost Calculation

The model calculates costs using Polish energy pricing structure:

• Energy rate per kWh
• Distribution fees
• VAT (Value Added Tax)

Limitations

• Model is trained on Polish residential data
• Cost calculations use Polish energy pricing
• Designed for monthly predictions
• Performance may vary for different consumption patterns

Example Output

        Date  Predicted_Consumption  Predicted_Cost
0 2025-06-01                    191              216
1 2025-07-01                    135              153
2 2025-08-01                    199              224

License

This model is provided as-is for demonstration and educational purposes.