Improving Electricity Theft Detection using Combination of Improved Crow Search Algorithm and Support Vector Machine

Abstract

Advanced Metering Infrastructure (AMI) is an essential segment of the smart grids that is responsible for gathering, measuring and analyzing the electricity demand. Energy losses in the electricity distribution and transmission network and electricity theft detection are major challenges of electricity suppliers around the world. The analysis of consumption data related to the customers is one of the essential resources to identify electricity thieves. In this paper, the Crow Search Algorithm (CSA) is improved and the factors weight (𝑤) and awareness probability (𝐴𝑃) are obtained dynamically and used to adjust the parameters 𝐶 and 𝛾 related to the Support Vector Machine (SVM). The results illustrate that the ICSA-SVM framework has acceptable performance and detects fraudulent customers with high accuracy.

Application of SSA-KNN Algorithm in Fault Analysis of Power Metering Automation System

“…The monitoring agent module is responsible for scheduling and scheduling the work of the whole system, and can monitor the operation of the agent in time. After the load imbalance is detected, it is changed from a busy agent to an idle agent to achieve load balancing [14]. When the monitoring Agent finds a busy Agent, it sends a query message to the agent.…”

Machine learning-based electricity theft detection using support vector machines

“…To identify electricity theft, several research employ machine learning methods such as artificial neural networks (ANNs), decision trees (DTs), SVMs, and random forests (RFs). For example, a study by [15] proposed model uses features extracted from monthly consumption data to segregate normal electricity consumption (non-theft)and theft customers, selecting the most relevant features using the Pearson's chi-square feature Another study by Ghaedi et al [16] proposed a method for electricity theft detection based on a combination of an improved crow search algorithm and support vector machines. The proposed method analyzed customer electricity consumption data provided by meters and combined the algorithms to classify electricity consumption patterns as normal or theft.…”