Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhanced Load Forecasting with GAT-LSTM: Leveraging Grid and Temporal Features

Ugochukwu Orji, Çiçek Güven, Dan Stowell

TL;DR

This paper tackles short-term load forecasting in power grids with high renewable penetration by introducing GAT-LSTM, a hybrid model that fuses Graph Attention Networks with LSTM and incorporates edge attributes into attention, plus an early fusion of spatial and temporal features. It demonstrates that the approach yields substantial accuracy gains on the Brazilian Electricity System dataset, achieving reductions of 21.8% in MAE, 15.9% in RMSE, and 20.2% in MAPE over strong baselines. The results underscore the value of integrated spatio-temporal modeling that respects grid topology and constraints for grid management and energy planning. The work also discusses limitations such as computational complexity and graph-quality sensitivity, and suggests future directions including adaptive graph refinement and uncertainty quantification to enhance robustness and interpretability.

Abstract

Accurate power load forecasting is essential for the efficient operation and planning of electrical grids, particularly given the increased variability and complexity introduced by renewable energy sources. This paper introduces GAT-LSTM, a hybrid model that combines Graph Attention Networks (GAT) and Long Short-Term Memory (LSTM) networks. A key innovation of the model is the incorporation of edge attributes, such as line capacities and efficiencies, into the attention mechanism, enabling it to dynamically capture spatial relationships grounded in grid-specific physical and operational constraints. Additionally, by employing an early fusion of spatial graph embeddings and temporal sequence features, the model effectively learns and predicts complex interactions between spatial dependencies and temporal patterns, providing a realistic representation of the dynamics of power grids. Experimental evaluations on the Brazilian Electricity System dataset demonstrate that the GAT-LSTM model significantly outperforms state-of-the-art models, achieving reductions of 21. 8% in MAE, 15. 9% in RMSE and 20. 2% in MAPE. These results underscore the robustness and adaptability of the GAT-LSTM model, establishing it as a powerful tool for applications in grid management and energy planning.

Enhanced Load Forecasting with GAT-LSTM: Leveraging Grid and Temporal Features

TL;DR

This paper tackles short-term load forecasting in power grids with high renewable penetration by introducing GAT-LSTM, a hybrid model that fuses Graph Attention Networks with LSTM and incorporates edge attributes into attention, plus an early fusion of spatial and temporal features. It demonstrates that the approach yields substantial accuracy gains on the Brazilian Electricity System dataset, achieving reductions of 21.8% in MAE, 15.9% in RMSE, and 20.2% in MAPE over strong baselines. The results underscore the value of integrated spatio-temporal modeling that respects grid topology and constraints for grid management and energy planning. The work also discusses limitations such as computational complexity and graph-quality sensitivity, and suggests future directions including adaptive graph refinement and uncertainty quantification to enhance robustness and interpretability.

Abstract

Accurate power load forecasting is essential for the efficient operation and planning of electrical grids, particularly given the increased variability and complexity introduced by renewable energy sources. This paper introduces GAT-LSTM, a hybrid model that combines Graph Attention Networks (GAT) and Long Short-Term Memory (LSTM) networks. A key innovation of the model is the incorporation of edge attributes, such as line capacities and efficiencies, into the attention mechanism, enabling it to dynamically capture spatial relationships grounded in grid-specific physical and operational constraints. Additionally, by employing an early fusion of spatial graph embeddings and temporal sequence features, the model effectively learns and predicts complex interactions between spatial dependencies and temporal patterns, providing a realistic representation of the dynamics of power grids. Experimental evaluations on the Brazilian Electricity System dataset demonstrate that the GAT-LSTM model significantly outperforms state-of-the-art models, achieving reductions of 21. 8% in MAE, 15. 9% in RMSE and 20. 2% in MAPE. These results underscore the robustness and adaptability of the GAT-LSTM model, establishing it as a powerful tool for applications in grid management and energy planning.

Paper Structure

This paper contains 19 sections, 7 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Review of Key Concepts
  3. Traditional Energy Forecasting Models
  4. Deep Learning-Based Energy Forecasting Models
  5. Graph Neural Networks (GNNs) for Energy Forecasting
  6. Hybrid Models for Energy Forecasting
  7. Methodology
  8. Data
  9. Data Source and Description:
  10. Data Preprocessing:
  11. Problem Definition and Model Architecture
  12. Training Process
  13. Experiments and Results
  14. Experimental Setup
  15. Baseline Models
  16. ...and 4 more sections

Figures (5)

  • Figure 1: Neighborhood Aggregation in GNNs. Source Jin2022GNNLens.
  • Figure 2: Model Architecture
  • Figure 3: Learning curve for all models.
  • Figure 4: Mean Actual vs Predicted Load Values for all Models.
  • Figure 5: Peak vs Off-peak Performance for GAT-LSTM.