Deep learning-based forecasting of electricity consumption.

Area of Science:

• Building Science
• Energy Systems Engineering
• Artificial Intelligence in Energy

Background:

• Building energy management systems (BEMS) are crucial for monitoring and controlling energy consumption in modern buildings.
• Effective BEMS implementation reduces energy usage, enhances energy supply quality, and improves occupant comfort.
• Understanding building energy dynamics is key to identifying effective energy-saving strategies.

Purpose of the Study:

• To address model optimization and electricity consumption forecasting for BEMS.
• To develop and validate an LSTM-based time series approach for future energy usage prediction.
• To test the proposed methodologies on real-world hospital facility energy consumption data.

Main Methods:

• Utilized Long Short-Term Memory (LSTM) neural networks for time series forecasting of electricity consumption.
• Applied model optimization techniques to enhance the performance of the forecasting methods.
• Collected and analyzed actual electricity consumption data from a hospital facility.

Main Results:

• The LSTM-based time series approach accurately predicted electricity consumption trends on actual data.
• Model optimizers significantly improved the performance of the proposed energy management strategies.
• Achieved a 95% accuracy for the objective function gain, demonstrating the effectiveness of the methods.

Conclusions:

• The developed BEMS strategies are effective for accurate electricity consumption forecasting.
• LSTM-based time series analysis is a viable method for predicting energy usage in buildings.
• The study demonstrates the potential for significant energy savings and improved efficiency in hospital facilities through advanced BEMS.