Scholarlink

<p>In response to the demand for fine-grained management of power consumers by electric grid companies, a composite model is proposed to achieve stratified classification of four types of electricity usage behavior: ‘normal,’ ‘theft,’ ‘leakage,’ and ‘metering anomalies’. Firstly, a three-dimensional feature representation is obtained by dimensionality reduction of high-dimensional electricity usage data using features such as voltage imbalance. Subsequently, a class of support vector machine, KH-OC-SVM (krill-herd optimized one-class support vector machine), optimized by the krill algorithm, is introduced to automatically classify the feature vectors into ‘normal’ and ‘abnormal’ categories. Finally, a density-based K-means clustering algorithm is utilized to analyze the ‘abnormal’ data, automatically categorizing them into three types of abnormal electricity usage behavior: ‘theft,’ ‘leakage,’ and ‘metering anomalies.’ The experimental results demonstrate the effectiveness of the proposed method in achieving automatic classification of power consumers’ electricity usage behavior.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Stratified classification method for power consumers’ electricity usage behavior based on composite models

  • 期刊论文

  • 发表日期：

    2025年5月17日

查看全文
  • Jun Yang,
  • Yi He,
  • Qi Meng,
  • Xixiang Zhang

摘要

In response to the demand for fine-grained management of power consumers by electric grid companies, a composite model is proposed to achieve stratified classification of four types of electricity usage behavior: ‘normal,’ ‘theft,’ ‘leakage,’ and ‘metering anomalies’. Firstly, a three-dimensional feature representation is obtained by dimensionality reduction of high-dimensional electricity usage data using features such as voltage imbalance. Subsequently, a class of support vector machine, KH-OC-SVM (krill-herd optimized one-class support vector machine), optimized by the krill algorithm, is introduced to automatically classify the feature vectors into ‘normal’ and ‘abnormal’ categories. Finally, a density-based K-means clustering algorithm is utilized to analyze the ‘abnormal’ data, automatically categorizing them into three types of abnormal electricity usage behavior: ‘theft,’ ‘leakage,’ and ‘metering anomalies.’ The experimental results demonstrate the effectiveness of the proposed method in achieving automatic classification of power consumers’ electricity usage behavior.