Scholarlink

The annual compliance cycle of the carbon trading system allows GenCos to buy allowances at different time than when the emissions happen, which gives GenCos the possibility to use a proper carbon trading strategy to lower their emission costs by taking advantage of the carbon price fluctuations. In this paper, the carbon trading problem considering the compliance cycle is formulated as a Markov Decision Process. The uncertain future carbon prices make it difficult to choose the right trading time and the uncertain future emissions make it difficult to choose the right trading amount, so the Twin Delayed Deep Deterministic Policy Gradient algorithm is used to make carbon trading decisions for the GenCos. Instead of using forecasted values, the algorithm learns the environment dynamics and make optimal carbon trading decisions only based on the information available currently. Numerical simulations show that the proposed method can help to lower the emission cost for a GenCo.

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

A Deep Reinforcement Learning Based Carbon Trading Strategy for Generation Companies Under Uncertainties

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shouyuan Shi,
  • Tao Yu

摘要

The annual compliance cycle of the carbon trading system allows GenCos to buy allowances at different time than when the emissions happen, which gives GenCos the possibility to use a proper carbon trading strategy to lower their emission costs by taking advantage of the carbon price fluctuations. In this paper, the carbon trading problem considering the compliance cycle is formulated as a Markov Decision Process. The uncertain future carbon prices make it difficult to choose the right trading time and the uncertain future emissions make it difficult to choose the right trading amount, so the Twin Delayed Deep Deterministic Policy Gradient algorithm is used to make carbon trading decisions for the GenCos. Instead of using forecasted values, the algorithm learns the environment dynamics and make optimal carbon trading decisions only based on the information available currently. Numerical simulations show that the proposed method can help to lower the emission cost for a GenCo.