Year: 2024
Type: article
Abstract: Electricity theft detection (ETD) plays a crucial role in reducing nontechnical losses of power transmission and distribution in smart grids. However, accurate ETD remains challenging due to the presence of missing values and highly unbalanced data. To further improve the accuracy of ETD, an advanced electricity theft detection system (ETDS) containing five modules is newly proposed. In detail, the data imputation module is designed to intelligently recover the missing information in the studied dataset; the feature extraction module is to eliminate redundant noise and extract the key components from the dataset; the deep learning module is developed based on the newly proposed regularized attention mechanism-based bidirectional gate recurrent unit; and the parameter optimization module is dedicated to acquiring the optimal parameters. Experiment results based on the performance evaluation module indicate that the ETDS not only outperforms its benchmarks in forecasting accuracy but also exhibits higher convergence rate and robustness.
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4405303380