Deep Learning Assisted Buildings Energy Consumption Profiling Using Smart Meter Data.

Area of Science:

• Energy Management Systems
• Data Science
• Artificial Intelligence

Background:

• Rising global populations and increased reliance on electronic devices drive demand for efficient energy production.
• Precise energy management systems are crucial for forecasting consumer usage and informing future energy policies.
• Smart sensors on meters and appliances facilitate detailed energy usage analysis by power suppliers.

Purpose of the Study:

• To propose a clustering-based analysis for categorizing consumer electricity usage levels.
• To develop a method for effective energy utilization analysis and resource allocation.
• To provide energy providers with a compact overview of energy consumption patterns.

Main Methods:

• Training a deep autoencoder to transform low-dimensional energy consumption data into high-level representations.
• Applying an adaptive self-organizing map (SOM) clustering algorithm to the high-level representations.
• Performing statistical analysis on clustered data to define electricity consumption levels.

Main Results:

• Successful categorization of consumer electricity usage into distinct levels.
• Visualization of energy consumption patterns through graphs, calendar views, and city maps.
• Generation of a comprehensive overview for energy providers on energy utilization.

Conclusions:

• The proposed clustering-based analysis effectively categorizes energy consumption patterns.
• The methodology provides valuable insights for optimizing energy distribution and management.
• Visualizations enhance the understanding of energy utilization for informed policymaking.