An efficient electricity theft detection based on deep learning.

Area of Science:

• Electrical Engineering
• Data Science
• Artificial Intelligence

Background:

• Electricity theft is a global problem causing significant financial losses for utility companies and increasing costs for consumers.
• Smart grids generate extensive data, offering opportunities for advanced analytics to detect energy theft.
• Existing methods for electricity theft detection face challenges with data quality and class imbalance.

Purpose of the Study:

• To develop and evaluate a novel deep learning approach for accurate electricity theft detection.
• To address data limitations, such as incompleteness and imbalance, using data augmentation techniques.
• To compare the proposed method's performance against existing state-of-the-art techniques using real-world data.

Main Methods:

• Utilized a hybrid model combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks.
• Employed LoRAS data augmentation to overcome dataset shortcomings like incomplete and imbalanced data.
• Validated the approach using authentic power consumption data from the State Grid Corporation of China.

Main Results:

• Achieved a validation accuracy of 97%, surpassing previous studies by 1%.
• Reported high performance metrics, including accuracy (98.75%, 95.45%, 97.7%), recall, and F1 scores.
• Demonstrated superior performance compared to other methods evaluated on the same dataset.

Conclusions:

• The proposed CNN-LSTM approach with LoRAS augmentation effectively detects electricity theft.
• The method significantly improves accuracy and overcomes common data challenges in smart grid analytics.
• This research offers a competitive and accurate solution for identifying electricity theft in smart grids.