Deep Learning for Time-Series Forecasting with Exogenous Variables in Energy Consumption: A Performance and Interpretability Analysis

Project Overview

This repository contains the code and data for the paper titled "Deep Learning for Time-Series Forecasting with Exogenous Variables in Energy Consumption: A Performance and Interpretability Analysis." The study evaluates different deep learning architectures for short-term load forecasting (STLF) in smart grids, focusing on their ability to integrate exogenous variables for improved predictive accuracy and interpretability. doi={10.1109/ACCESS.2025.3570618}

Language

• Python

Algorithms and Models

• Long Short-Term Memory (LSTM)
• Deep Autoregressive (DeepAR).
• Temporal Fusion Transformer (TFT).
• TimeLLm.

Libraries and Tools

• Data Analysis: pandas, numpy, statsmodels
• Machine Learning: keras, TensorFlow, PyTorch, PyTorch Lightning, PyTorch Forecasting, gluonts
• Model Interpretability: SHAP, Feature Permutation, Attention-based Analysis
• Hyperparameter Optimization: Optuna
• Data Visualization: matplotlib, seaborn, TensorBoard
• Other Tools: scikit-learn
• Data Processing & Analysis: pandas, numpy, statsmodels

Dataset

SmartMeter Energy Consumption Data in London Households -Original Dataset: https://data.london.gov.uk/dataset/smartmeter-energy-use-data-in-london-households -Refactorized Dataset : https://www.kaggle.com/jeanmidev/smart-meters-in-london