发电技术 ›› 2023, Vol. 44 ›› Issue (6): 865-874.DOI: 10.12096/j.2096-4528.pgt.22165

• 智能电网 • 上一篇    下一篇

堆栈式集成学习驱动的电力系统暂态稳定预防控制优化方法

潘晓杰1, 徐友平1, 解治军2, 王玉坤1, 张慕婕1, 石梦璇1, 马坤2, 胡伟2   

  1. 1.国家电网公司华中分部公司,湖北省 武汉市 430077
    2.电力系统及发电设备控制和 仿真国家重点实验室(清华大学电机工程与应用电子技术系),北京市 海淀区 100084
  • 收稿日期:2023-10-05 出版日期:2023-12-31 发布日期:2023-12-28
  • 通讯作者: 胡伟
  • 作者简介:潘晓杰(1976),男,博士后,研究方向为电力系统稳定与控制,pxj76@163.com
    胡伟(1976),博士,研究员,主要研究方向为综合能源系统运行、电力系统建模与安全分析中的深度学习、智能控制等,本文通信作者,huwei@mail.tsinghua.edu.cn
  • 基金资助:
    国家电网公司华中分部科技项目(5214DK210014)

Power System Transient Stability Preventive Control Optimization Method Driven by Stacking Ensemble Learning

Xiaojie PAN1, Youping XU1, Zhijun XIE2, Yukun WANG1, Mujie ZHANG1, Mengxuan SHI1, Kun MA2, Wei HU2   

  1. 1.Central China Branch of State Grid Corporation of China, Wuhan 430077, Hubei Province, China
    2.State Key Lab of Control and Simulation of Power Systems and Generation Equipments (Department of Electrical Engineering and Applied Electronic Technology, Tsinghua University), Haidian District, Beijing 100084, China
  • Received:2023-10-05 Published:2023-12-31 Online:2023-12-28
  • Contact: Wei HU
  • Supported by:
    Science and Technology of Central China Branch of State Grid Corporation of China(5214DK210014)

摘要:

针对暂态稳定预防控制在线计算的快速性要求和时域方程计算复杂性之间的矛盾,提出一种堆栈式集成学习驱动的电力系统暂态稳定预防控制优化方法。首先,构建了基于堆栈式集成深度置信网络的暂态稳定评估器,用以代替暂态稳定判定所需的非线性微分代数方程求解过程;其次,将训练好的暂态稳定评估器作为暂态稳定约束判别器,嵌入帝企鹅启发式优化算法的迭代寻优过程中;最后,以预防控制代价最小为目标,建立集成学习驱动的电力系统暂态稳定预防控制启发式优化算法,该算法实现了预防控制中暂态稳定约束的高效判断,提高了发电再调度预防控制决策水平。基于IEEE39节点系统对所提预防控制优化方法进行实验验证,结果表明,该方法在评估准确率和计算效率上都具有良好的效果。

关键词: 电力系统, 堆栈式集成学习, 帝企鹅启发式优化算法, 暂态稳定, 预防控制

Abstract:

Aiming at the contradiction between the rapidity requirement of online calculation of transient stability preventive control and the computational complexity of time-domain equations, a stacking ensemable learning driven optimization method for power system transient stability preventive control was proposed. Firstly, a transient stability estimator based on a stacking ensemble deep belief network was constructed to replace the nonlinear differential algebraic equation solution process required for transient stability determination. Secondly, the trained transient stability estimator was used as a transient stability constraint discriminator, which was embedded in the iterative optimization process of the Aptenodytes Forsteri optimization algorithm. Finally, with the goal of minimizing the cost of preventive control, a stacking ensemble learning driven power system transient stability preventive control optimization algorithm was established. The algorithm realized the efficient judgment of transient stability constraints in preventive control, and improved the decision-making level of preventive control for power generation rescheduling. Based on the IEEE39 bus system, the proposed preventive control method was verified by experiments. The results show that the method has achieved good results in both evaluation accuracy and calculation efficiency.

Key words: power system, stacking ensemble learning, Aptenodytes Forsteri optimization algorithm, transient stability, preventive control

中图分类号: