Pafkun333/Smart_Energy_Twin

Smart Energy Twin – Forecasting Residential Energy Demand

This repository contains the report and accompanying materials for the Smart Energy Twin project, a minor project within the Master Engineering Systems program at HAN University of Applied Sciences.

Project Overview

The project develops machine learning models to forecast both electricity consumption and heat demand for eight residential households in the Smart Energy Twin initiative. By applying a combination of tree-based algorithms and neural networks, the study addresses the challenges of household energy forecasting and supports sustainable energy management strategies.

Key Highlights

• Electricity Consumption Forecasting

  • Algorithms: Random Forest, XGBoost, CNN-GRU
  • Feature engineering: lag variables, rolling windows, cyclical encoding
  • Best-performing model: XGBoost (Test RMSE ≈ 1769 W, R² ≈ 0.86)

• Heat Demand Forecasting

  • Hourly and daily models using XGBoost and Neural Networks
  • Daily delta-prediction model achieved the lowest RMSE (~0.39 kW)
  • Demonstrated improved generalizability for practical deployment

Repository Structure

├── report/ │ └── MP_smart_energy_report.pdf # Project report ├── src/ # (to be added: scripts, notebooks, training pipeline) ├── data/ # (to be added: sample/preprocessed datasets) └── README.md

Methods

• Data preprocessing: anomaly detection, temperature mapping, normalization
• Models: Random Forest Regressor, XGBoost, CNN-GRU hybrid networks
• Metrics: RMSE, MAE, R²
• Feature engineering: lagged features, rolling statistics, temperature interactions, cyclical encodings

Results

• Robust and interpretable forecasting models for minutely and hourly consumption data
• Demonstrated the importance of feature engineering and hyperparameter tuning
• Provided insights into household-level energy management

Contributors

• Melika Mirmohammad
• Pavel Petrovski
• Khashayar Amir Hosseini
• Mohammad Eghbalighahyazi

Supervised by: Aishwarya Aswal
Client: Trung Nguyen

Date

Arnhem, June 2025

Citation

If you use this work, please cite as:
MP-Smart Energy Twin Project Report, HAN University of Applied Sciences, 2025.