TCN-QRNN model for short term energy consumption forecasting with increased accuracy and optimized computational...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Energy Systems

Background:

• Traditional Recurrent Neural Networks (RNNs) struggle with computational complexity for real-time energy forecasting.
• Effective energy system management is hindered by limitations in current forecasting methods.
• Growing data volume and complexity necessitate advanced forecasting solutions.

Purpose of the Study:

• To propose a novel hybrid model combining Temporal Convolutional Networks (TCN) and Quasi-Recurrent Neural Networks (QRNN) for energy consumption forecasting.
• To address the computational complexity and real-time application limitations of existing forecasting models.
• To improve the accuracy and efficiency of energy consumption predictions.

Main Methods:

• Developed a hybrid TCN-QRNN model integrating TCN's long time-series processing with QRNN's computational efficiency.
• Evaluated the model against established methods like Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU).
• Assessed performance using metrics including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE).

Main Results:

• The TCN-QRNN model demonstrated a 40% accuracy improvement over LSTM and an 8% improvement over TCN-LSTM.
• Achieved a 30% reduction in data processing time compared to existing models.
• Exhibited a significantly smaller parameter count than LSTM and GRU, enhancing suitability for resource-constrained environments.

Conclusions:

• The TCN-QRNN model offers a promising solution for accurate and efficient energy consumption forecasting.
• The model's reduced computational demands and high accuracy make it suitable for real-world energy management.
• This hybrid approach overcomes the limitations of traditional RNNs in handling complex, large-scale energy data.