Deep learning-driven hybrid model for short-term load forecasting and smart grid information management

Area of Science:

• Electrical Engineering
• Computer Science
• Artificial Intelligence

Background:

• Accurate power load forecasting is essential for smart grid sustainability.
• Challenges include complex load dynamics, uncertainty, and high-dimensional energy data.
• Existing methods struggle with intricate features and long-term dependencies.

Purpose of the Study:

• To develop a computational approach for precise short-term power load forecasting.
• To improve energy information management in smart grid systems.
• To enhance the prediction of future load demand.

Main Methods:

• A hybrid deep learning model combining Gated Recurrent Unit (GRU) and Temporal Convolutional Network (TCN).
• Incorporation of an attention mechanism to focus on relevant input features.
• Evaluation on public datasets, including GEFCom2014.

Main Results:

• The proposed algorithm outperforms baseline models in prediction accuracy, efficiency, and stability.
• On GEFCom2014: FLOP reduced by 48.8%, inference time by 46.7%, and MAPE improved by 39%.
• Demonstrated significant enhancements in smart grid reliability, stability, and cost-effectiveness.

Conclusions:

• The hybrid GRU-TCN model with attention effectively addresses challenges in power load forecasting.
• The method improves operational planning and risk assessment for smart grids.
• Enhances the overall performance and economic viability of smart grid systems.