Open AccessStudy on User Fraud Identification of PV Expansion based on a Bottom-up Approach of a DELM Algorithm improved by SSA for a Power Distribution Network
Author(s) -
Wang Jinpeng,
Wei Haojie,
Dou Shunyao,
Jeremy-Gillbanks,
Zhao Xin
Publication year - 2025
Publication title -
ieee access
Language(s) - English
Resource type - Magazines
SCImago Journal Rank - 0.587
H-Index - 127
eISSN - 2169-3536
DOI - 10.1109/access.2025.3574167
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
In order to accurately identify users engaged in the fraudulent expansion of illegally distributed photovoltaic (PV) capacity, this paper initially leverages the similarity of the PV power generation output in the same region. It preprocesses the reference power station and the site to be tested using cosine similarity. Next, a Sparrow Search Algorithm (SSA) was applied to optimize the weight parameters of the Deep Extreme Learning Machine (DELM). And then, this study proposes DELM Network model improved by the SSA to detect fraudulent behavior among domestic customers, which is the primary cause of non-technical losses in distribution networks. Meanwhile, we utilized a bottom-up approach to determine the normal behavior pattern of household loads with and without PV sources. Customers suspected of energy theft are identified by calculating an anomaly index for each user. Finally, the proposed approach is validated both in simulation and using measurement data from real networks. The algorithm’s performance in detecting fraudulent behavior in outdated electromagnetic meters is evaluated and verified. The results demonstrate that the proposed SSA-DELM model achieves significant optimization, increasing the fitness degree by 3.89% and reducing the overall detection error by 67.12% compared to other single deep learning models without data preprocessing.