|
| 基于多智能体深度强化学习的绿电制氨多能微网能量管理方法 |
| Low-Carbon Energy Management Approach for Multi-Energy Microgrids with Renewable Power-to-Ammonia Using Multi-Agent Deep Reinforcement Learning Algorithm |
| 投稿时间：2026-03-19 修订日期：2026-05-22 |
| DOI： |
| 中文关键词: 绿电制氨 多能微网 碳交易 多智能体深度强化学习 |
| 英文关键词: power-to-ammonia multi-energy microgrid carbon trading multi-agent deep reinforcement learning |
| 基金项目::国家自然科学基金资助项目(52507160); 山东省自然科学基金资助项目(ZR2023QE006) |
|
| 摘要点击次数: 14 |
| 全文下载次数: 0 |
| 中文摘要: |
| 可再生能源制氨作为零碳燃料载体在能源系统中具有重要潜力。本文针对含可再生能源制氨的多能微网系统，提出一种基于阶梯式碳交易的能量管理模型。首先，构建了涵盖光伏发电、电解制氢、储氢、合成氨、储氨及热电联供的全系统模型，精细刻画氨合成过程的热电化学动态特性。其次，为应对可再生能源出力波动，将能量管理问题建模为状态转移函数未知的马尔可夫决策（Markov decision process，MDP）过程。再次，基于“集中训练-分散执行”的多智能体深度强化学习（multi-agent deep reinforcement learning，MADRL）架构，提出一种无模型的数据驱动实时优化算法，并采用柔性演员-评论家（soft actor-critic，SAC）框架设计智能体学习机制。算例结果表明，所提模型能够实现多微能网在可再生能源波动下的低碳经济调度，所提算法在求解速度与精度上均优于传统优化方法。 |
| 英文摘要: |
| Ammonia production using renewable energy holds significant potential as a zero-carbon fuel carrier in energy systems. This paper proposes an energy management model based on tiered carbon trading for multi-energy microgrid that incorporate re-newable energy-based ammonia production. First, a full-system model was developed that encompasses photovoltaic power generation, electrolytic hydrogen production, hydrogen storage, ammonia synthesis, ammonia storage, and combined heat and power (CHP), providing a detailed description of the ther-mo-electrochemical dynamics of the ammonia synthesis process. Second, to address fluctuations in renewable energy output, the energy management problem is modeled as a Markov decision process (MDP) with unknown state transition functions. Fur-thermore, based on a “centralized training–decentralized execu-tion” multi-agent deep reinforcement learning (MADRL) archi-tecture, a model-free, data-driven optimization algorithm is pro-posed, and the agent learning mechanism is designed using a soft actor-critic (SAC) framework. Case studies demonstrate that the proposed model can achieve low-carbon economic dispatch for multi-microgrids under fluctuating renewable energy conditions. The proposed algorithm outperforms traditional optimization methods in both solution speed and accuracy. |
|
View Fulltext
查看/发表评论 下载PDF阅读器 |
| 关闭 |