Scholarlink

<p>As China’s electricity spot market accelerates, ensuring its smooth operation requires the timely and accurate identification of power generators’ market power abuse. Traditional, expert-based judgements cannot keep pace with the expanding volume of trades. This paper proposes a novel approach that combines cost-sensitive learning (CSL) with an improved CatBoost model. First, we analyse the mechanisms by which generators may abuse market power and provide a quantitative definition. Next, CSL is employed to address the resulting class imbalance. We then construct an ensemble framework using CatBoost as the base classifier and apply a hunter–prey optimisation (HPO) algorithm to fine-tune its initial parameters and boost classification performance. Finally, we validate the method on spot-market data from a regional Chinese electricity market, achieving an identification accuracy of 98.29%. The results demonstrate the effectiveness of the proposed approach in recognising generators’ market power abuse and providing robust technical support for regulating electricity spot markets.</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

Identification of generators’ market power abuse based on hunter–prey optimisation and CatBoost

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月13日

查看全文
  • Qian Sun,
  • Shan Deng,
  • Yunyong Zhang,
  • Jiayu Zhong,
  • Yansong Wu

摘要

As China’s electricity spot market accelerates, ensuring its smooth operation requires the timely and accurate identification of power generators’ market power abuse. Traditional, expert-based judgements cannot keep pace with the expanding volume of trades. This paper proposes a novel approach that combines cost-sensitive learning (CSL) with an improved CatBoost model. First, we analyse the mechanisms by which generators may abuse market power and provide a quantitative definition. Next, CSL is employed to address the resulting class imbalance. We then construct an ensemble framework using CatBoost as the base classifier and apply a hunter–prey optimisation (HPO) algorithm to fine-tune its initial parameters and boost classification performance. Finally, we validate the method on spot-market data from a regional Chinese electricity market, achieving an identification accuracy of 98.29%. The results demonstrate the effectiveness of the proposed approach in recognising generators’ market power abuse and providing robust technical support for regulating electricity spot markets.