Performance analysis of deep learning-based electric load forecasting model with particle swarm optimization.

Area of Science:

• Artificial Intelligence
• Electrical Engineering
• Data Science

Background:

• Traditional power load forecasting methods struggle with the complexity and uncertainty of modern power grids.
• Deep learning offers new potential for improving forecasting accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for enhanced power load forecasting.
• To address the limitations of traditional methods in handling complex, time-series power data.

Main Methods:

• The proposed model, PSO-BiTC, integrates Temporal Convolutional Networks (TCN) for sequence processing and Bidirectional Long Short-Term Memory (BiLSTM) for dependency capture.
• Particle Swarm Optimization (PSO) is employed to optimize the model's parameters, enhancing predictive performance and generalization.
• The TCN component effectively processes long time-series data, identifying patterns, while BiLSTM captures both short-term and long-term dependencies.

Main Results:

• The PSO-BiTC model demonstrated superior performance across four datasets, significantly reducing Mean Absolute Error (MAE) compared to traditional methods.
• Achieved MAE values of 20.18, 17.57, 18.61, and 16.7 on the tested datasets.
• The model exhibits excellent performance metrics, a low parameter count, and reduced training time.

Conclusions:

• The PSO-BiTC model represents a significant advancement in deep learning-based power load forecasting.
• Its effectiveness contributes to improved power system operation, optimized resource allocation, and supports urban carbon emission reduction goals.