| 引用本文: | 王光华,李晓影,宋秉睿,张沛.基于深度强化学习的配电网负荷转供控制方法[J].电力自动化设备,2022,42(7): |
| WANG Guanghua,LI Xiaoying,SONG Bingrui,ZHANG Pei.Load transfer control method of distribution network based on deep reinforcement learning[J].Electric Power Automation Equipment,2022,42(7): |
|
| 摘要: |
| 随着城市规模的快速扩张以及电能替代的不断推进，配电网节点数大量增加，结构愈加复杂，发生故障后拓扑变化不确定性较大，传统负荷转供方法难以在短时间内给出高质量的解决方案。为此，提出基于深度强化学习的配电网负荷转供控制方法。将负荷转供过程视为一个马尔可夫决策过程，与配电网实时电气、拓扑数据进行交互，对联络开关与分段开关进行控制。为了提高算法的精度与泛化能力，针对算法动作策略加入了预模拟机制，调整了动作与学习的比例并采用自适应优化算法进行求解。算例分析表明，所提方法能够应对不同故障下配电网的拓扑变化，即时给出负荷恢复量、电网损耗、开关动作次数多方面最优的转供控制方案，这对于减小故障后的停电损失与提高用户满意度有着重要意义。 |
| 关键词: 深度强化学习 配电网 负荷转供 马尔可夫决策过程 |
| DOI：10.16081/j.epae.202204015 |
| 分类号:TM761 |
| 基金项目:国网河北省电力有限公司科技项目(kj2021-014) |
|
| Load transfer control method of distribution network based on deep reinforcement learning |
|
WANG Guanghua1, LI Xiaoying1, SONG Bingrui2, ZHANG Pei2
|
|
1.Baoding Power Supply Company, State Grid Hebei Electric Power Co.,Ltd.,Baoding 071000, China;2.Tianjin Xianghe Electric Co.,Ltd.,Tianjin 300042, China
|
| Abstract: |
| With the rapid expansion of urban scale and the continuous advancement of electric energy substitution, the number of buses in distribution network increases greatly, the structure becomes more complex, the uncertainty of topology changing after fault becomes larger, and it is difficult for the traditional load transfer methods to provide high-quality solution schemes in short time. For that, the load transfer control method of distribution network based on deep reinforcement learning is proposed. The load transfer process is regarded as a Markov decision process, which interacts with the real-time electrical and topological data of distribution network, and controls the tie switch and subsection switch. In order to improve the accuracy and generalization ability of the algorithm, the pre-simulation mechanism is added to the action strategy for the algorithm, the proportions of action and learning are adjusted and the adaptive optimization algorithm is adopted to solve the problem. The case analysis shows that the proposed method can deal with the topology variation of distribution network under different faults, and provide the optimal transfer control scheme of load recovery amount, power grid loss and times of switching operation immediately, which is of great significance for reducing the outage loss and improving users’ satisfaction. |
| Key words: deep reinforcement learning distribution network load transfer Markov decision process |
