半月刊

ISSN 1000-1026

CN 32-1180/TP

+高级检索 English
基于在线强化学习的风电系统自适应负荷频率控制
作者:
作者单位:

武汉大学电气与自动化学院,湖北省武汉市 430072

作者简介:

通讯作者:

基金项目:

国家重点研发计划资助项目(2018AAA0101501);国家自然科学基金资助项目(51707136);湖北省杰出青年基金资助项目(2018CFA080)。


Adaptive Load Frequency Control of Wind Power System Based on Online Reinforcement Learning
Author:
Affiliation:

School of Electrical Engineering and Automation, Wuhan University, Wuhan 430072, China

Fund Project:

This work is supported by National Key R&D Program of China (No. 2018AAA0101501); National Natural Science Foundation of China (No. 51707136) and Excellent Youth Foundation of Hubei Scientific Committee (No. 2018CFA080).

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
    摘要:

    大规模风电接入给系统带来新的不确定性,影响系统频率响应特性,从数据驱动的角度出发,提出了一种基于自适应动态模型的在线强化学习方法,用于系统的负荷频率控制。建立低秩自编码器特征提取网络,从所量测的低维数据中发现隐藏特征;基于特征网络,建立非线性动态系统稀疏辨识学习模型,感知系统动态模型的潜在物理状态,提升模型在线学习效率;通过结合模型预测控制,进行实时决策控制。所提出方法能够有效解决传统模型预测控制对系统全局模型准确性的依赖问题,加强控制器对系统动态模型的自适应性,且能有效跟踪风电输出功率的随机波动。最后,以接入四型风机的负荷频率控制模型为例,验证所提方法的有效性。

    Abstract:

    Large scale of wind power connected to the gird brings new uncertainties which affects the frequency response characteristics of the system, from the perspective of data driving, an online reinforcement learning method based on adaptive dynamic model for load frequency control is proposed. A low rank autoencoder feature extraction network is established to discover hidden features from the measured low-dimensional data. Based on the feature network, a nonlinear dynamic system sparse identification learning model is established to sense the potential physical state of the dynamic model of the system and improve the data efficiency of online learning. By combining model predictive control, real-time decision control is implemented, which solves the problem of traditional MPC’s dependence on the accuracy of the system global model, enhance the controller’s adaptability to the system dynamic model and effectively track the random fluctuation of wind power. Finally, the validity of the proposed control method is illustrated by load frequency control model integrated type VI wind turbine.

    表 1 重构相平面平均DTW距离Table 1 Average DTW distance of embedded phase-space
    表 3 LFC物理模型参数学习结果Table 3 Learning results of parameters in LFC physical model
    表 5 Table 5
    图1 基于在线强化学习的负荷频率控制原理图Fig.1 Principle diagram of load frequency based on online reinforcement learning
    图2 无LLAE的模型学习曲线Fig.2 Model learning curve without LLAE
    图3 有LLAE的模型学习曲线Fig.3 Model learning curve with LLAE
    图4 不同风电渗透率下系统频率响应Fig.4 Frequency responses with different wind penetration ratios
    图 风功率与预测值间偏差标幺值曲线Fig. Wind power deviation (pu)
    图 原系统庞加莱截面图Fig. Poincare Sections of original LFC system
    图 重构相平面庞加莱截面图Fig. Poincare Sections of embedded phase-space
    图 样本效率Fig. Data efficiency
    表 4 Table 4
    表 2 LLAE隐藏层平均DTW距离Table 2 Average DTW distance of hidden layer in LLAE
    图 LFC仿真系统Fig. Diagram of single area power system
    图 自编码器原理图Fig. Schematic diagram of Autoencoder
    图 SINDy原理图Fig. Schematic diagram of SINDy
    图 模型预测控制原理图Fig. Schematic diagram of model predictive control
    图 基于强化学习的LFC控制流程图Fig. Flow chart of Model-based Reinforcement Learning for LFC
    图 LFC系统状态空间演变图Fig. evolution of LFC system over time
    图 LLAE隐藏层演变图Fig. evolution of hidden layer over time
    图 SINDy对LFC系统的离线训练效果Fig. SINDy prediction performance for LFC
    图 SINDy对LFC系统的学习效果Fig. SINDy prediction performance for LFC
    图 在线强化学习效果Fig. Online reinforcement learning performance
    图 有、无模型参数误差系统频率响应Fig. Frequency responses for single area integrated wind power with or without model errors
    图 考虑模型误差风功率随机扰动下系统频率响应Fig. Frequency responses for single area integrated wind power with model errors
    参考文献
    相似文献
    引证文献
引用本文

杨丽,孙元章,徐箭,等.基于在线强化学习的风电系统自适应负荷频率控制[J/OL].电力系统自动化,http://doi.org/10.7500/AEPS20190706006.
YANG Li,SUN Yuanzhang,XU Jian,et al.Adaptive Load Frequency Control of Wind Power System Based on Online Reinforcement Learning[J/OL].Automation of Electric Power Systems,http://doi.org/10.7500/AEPS20190706006.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2019-07-06
  • 最后修改日期:2020-04-27
  • 录用日期:2019-10-10
  • 在线发布日期:
  • 出版日期: