首页|融合OMAAPS算法的现代智慧配电网技术研究与应用

融合OMAAPS算法的现代智慧配电网技术研究与应用

扫码查看
针对未知环境下智慧配电网电压越限求解问题,研究选择有源配电网与多智能体深度强化学习算法作为基础.同时使用集中式训练分布式执行框架、裁剪函数和注意力机制进行优化,并引入迭代策略提取验证算法对训练进行进一步改进,得到多智能体注意力近端策略优化算法,最后设计融合多智能体注意力近端策略优化算法的现代智慧配电网技术.研究结果显示,在实际应用中,与主流的下垂控制方法相比,研究方法的用电成本降低了 1.75%,最小为 582.955 6 元,同时平均室内温度偏离与平均电压偏离结果均最小,分别对应0.001 1℃与1.37*10-7 p.u..上述结果说明研究方法能帮助用户有效节约用电成本,有效提升现代智慧配电网对分布式光伏等新能源的消纳能力.
Research and application of modern smart distribution network technology with OMAAPS algorithm
For the problem of voltage exceeding limits in intelligent distribution networks in unknown environments,research is conducted on selecting active distribution networks and multi-agent deep reinforcement learning algorithms as the basis.Simultaneous-ly using a centralized training distributed execution framework,pruning function,and attention mechanism for optimization,and intro-ducing an iterative policy extraction verification algorithm to further improve the training,a multi-agent attention proximal strategy op-timization algorithm can be obtained.Finally,a modern intelligent distribution network technology that integrates multi-agent attention proximal strategy optimization algorithms is designed.The research results show that in practical application,compared with the main-stream sag control method,the electricity cost of the research method is reduced by 1.75%,and the minimum is 582.955 6 yuan.Meanwhile,the average indoor temperature deviation and average voltage deviation are the smallest,corresponding to 0.001 1℃and 1.37*10-7 p.u.,respectively.The above results show that the research method can help users effectively save electricity costs and effectively improve the absorption capacity of modern smart distribution network for new energy such as distributed photovoltaic.

OMAAPS algorithmsmart distribution networkvoltage regulationcivil buildingsenergy system

郑悦

展开 >

国网天津市电力公司,天津 300000

OMAAPS算法 智慧配电网 电压调节 民用建筑 能源系统

2024

自动化与仪器仪表
重庆工业自动化仪表研究所,重庆市自动化与仪器仪表学会

自动化与仪器仪表

CSTPCD
影响因子:0.327
ISSN:1001-9227
年,卷(期):2024.(11)