Advanced Electricity Theft Detection Using Deep Learning and Smart Grid Data
Author(s):
Famina F F, Dr.L.Nisha Evangelin
Keywords:
Electricity Theft, Smart Grids, Deep Learning, Data Interpolation, Energy Efficiency
Abstract
Electricity theft poses a global challenge, jeopardizing the financial stability of utility providers and compromising safety. This project introduces an advanced electricity theft detection system leveraging smart grids and deep learning techniques. The study utilizes comprehensive datasets, employing deep neural network-based classification models trained on customer consumption information. Addressing challenges like missing data and class imbalance, the project enhances robustness through data interpolation and synthetic data generation. Feature engineering incorporates time and frequency domain analyses, with experiments employing principal component analysis for a reduced feature space. Optimizations via Bayesian and adaptive moment estimation optimizers improve model accuracy. Benchmarking against existing methods demonstrates competitiveness, achieving a 97% area under the curve and 91.8% accuracy. Beyond economic stability, the system enhances safety, consumer benefits, and energy efficiency. The methodology and findings establish a foundation for advancing electricity theft detection, providing a powerful tool for utility companies. Keywords: Electricity theft, smart grids, deep learning, data interpolation, feature engineering, optimization, benchmarking, energy efficiency.
Article Details
Unique Paper ID: 162206
Publication Volume & Issue: Volume 10, Issue 8
Page(s): 314 - 319
Article Preview & Download
Share This Article
Join our RMS
Conference Alert
NCSEM 2024
National Conference on Sustainable Engineering and Management - 2024