Intelligent Solution Algorithm for Unit Commitment Based on Deep Reinforcement Learning

MU Huanhuan; YU Ling; XIA Fan; YUAN Ye

doi:10.13296/j.1001-1609.hva.2025.05.021

您当前的位置：

首页 >

文章列表页 >

Intelligent Solution Algorithm for Unit Commitment Based on Deep Reinforcement Learning

The Key Technologies for Energy Conservation and Carbon Reduction in New Power Systems | 更新时间：2025-05-25

- Intelligent Solution Algorithm for Unit Commitment Based on Deep Reinforcement Learning
- High Voltage Apparatus Vol. 61, Issue 5, Pages: 197-207(2025)
- 作者机构：
  
  华电（林周）新能源有限公司，拉萨　 851600
- 作者简介：
- 基金信息：
- DOI：10.13296/j.1001-1609.hva.2025.05.021
  CLC：
- Received：03 December 2024，
  
  Revised：18 February 2025，
  
  Published：16 May 2025
- 稿件说明：
移动端阅览
MU Huanhuan, YU Ling, XIA Fan, et al. Intelligent Solution Algorithm for Unit Commitment Based on Deep Reinforcement Learning[J]. High voltage apparatus, 2025, 61(5): 197-207.
DOI：

MU Huanhuan, YU Ling, XIA Fan, et al. Intelligent Solution Algorithm for Unit Commitment Based on Deep Reinforcement Learning[J]. High voltage apparatus, 2025, 61(5): 197-207. DOI： 10.13296/j.1001-1609.hva.2025.05.021.

摘要

随着中国能源结构持续深化转型，高比例波动性新能源接入使得现有的机组组合（unit commitment，UC）理论已经无法适应新型电力系统日前市场决策的发展需要。为此文中结合深度强化学习（deep reinforcement learning，DRL）技术提出了一种UC智能求解算法。首先引入DRL算法，对UC问题进行马尔科夫决策过程（Markov decision process，MDP）建模并给出对应的状态空间、转移函数、动作空间及奖励函数；然后采用策略梯度（policy gradient，PG）算法进行求解；在此基础上采用Lambda迭代求解该启停状态下机组的出力方案，并最终提出一种基于DRL的UC智能求解算法。基于仿真算例验证文中方法的适用性和有效性。

Abstract

With the continuous and deep transformation of energy structure in China

the access of the high proportion of volatile new energy makes the existing unit commitment theory unable to suit the development demands of the present market decision-making in the new power system. Therefore

a kind of UC intelligent solution algorithm in combination with the deep reinforcement learning (DRL) technology is proposed. Firstly

the DRL algorithm is introduced to model the Markov decision process of UC problem and the corresponding state space

transfer function

action space and reward function are given. Then

the strategy policy gradient (PG) algorithm is adopted for solution and

on this basis

the Lambda iteration are adopted to solve the output scheme of the unit under the startup and shutdown state respectively. Finally

a DRL-based UC intelligent solution algorithm is proposed. The applicability and effectiveness of this method are verified based on simulation examples.

关键词

Keywords

references

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Study on Automatic Control Method of Substation Inspection Robot Based on Deep Reinforcement Learning

Related Author

DONG Shihui

NIU Caiwen

DAI Kun

Related Institution

Tangshan Polytechnic College

⁰