GrAb: A Deep Learning-Based Data-Driven Analytics Scheme for Energy Theft Detection.

Area of Science:

• Electrical Engineering
• Computer Science
• Data Science

Background:

• Smart grids (SG) integrate information and communication technology (ICT) with energy infrastructures for efficient energy management.
• Partial SG deployment leads to energy losses, including technical and non-technical (energy theft), impacting grid reliability and demand fulfillment.
• Accurate energy theft detection is vital for reducing grid burden and ensuring energy availability.

Purpose of the Study:

• To propose GrAb, a novel deep learning (DL)-based scheme for detecting energy theft in smart grids.
• To leverage a data-driven analytics approach for identifying various forms of energy theft, such as data manipulation and clandestine connections.

Main Methods:

• Utilized a deep learning (DL) long short-term memory (LSTM) model to predict energy consumption from smart meter data.
• Implemented a threshold calculator to determine energy consumption benchmarks.
• Employed a support vector machine (SVM)-based classifier, integrating predicted consumption and threshold values, to categorize energy losses (technical, non-technical, normal).

Main Results:

• The GrAb scheme demonstrated high accuracy in identifying energy theft.
• Experimental results confirmed GrAb's superior performance compared to existing state-of-the-art energy theft detection methods.

Conclusions:

• The proposed GrAb scheme effectively detects energy theft using a data-driven DL approach.
• GrAb offers a promising solution for enhancing smart grid security and operational efficiency by accurately identifying non-technical energy losses.