汤迎春, 晏光辉, 张雅婷, 刘书池, 刘颂凯, 张磊. 基于约束投票极限学习机的在线静态电压稳定评估[J]. 现代电力, 2022, 39(5): 521-528. DOI: 10.19725/j.cnki.1007-2322.2021.0162
引用本文: 汤迎春, 晏光辉, 张雅婷, 刘书池, 刘颂凯, 张磊. 基于约束投票极限学习机的在线静态电压稳定评估[J]. 现代电力, 2022, 39(5): 521-528. DOI: 10.19725/j.cnki.1007-2322.2021.0162
TANG Yingchun, YAN Guanghui, ZHANG Yating, LIU Shuchi, LIU Songkai, ZHANG Lei. Online Static Voltage Stability Assessment Based on Constrained Voting Extreme Learning Machine[J]. Modern Electric Power, 2022, 39(5): 521-528. DOI: 10.19725/j.cnki.1007-2322.2021.0162
Citation: TANG Yingchun, YAN Guanghui, ZHANG Yating, LIU Shuchi, LIU Songkai, ZHANG Lei. Online Static Voltage Stability Assessment Based on Constrained Voting Extreme Learning Machine[J]. Modern Electric Power, 2022, 39(5): 521-528. DOI: 10.19725/j.cnki.1007-2322.2021.0162

基于约束投票极限学习机的在线静态电压稳定评估

Online Static Voltage Stability Assessment Based on Constrained Voting Extreme Learning Machine

  • 摘要: 快速准确的在线静态电压稳定评估是规模化互联电网安全稳定运行的重要保障。针对传统神经网络学习模型调参繁杂、训练时间长、样本需求数量庞大等缺点,提出了一种基于约束投票极限学习机(constrained voting extreme learning machine,CV-ELM)的在线静态电压稳定评估模型。CV-ELM基于类间样本差值构建差向量集计算输入层对隐藏层的权值及隐藏层节点偏置项,并引入多数投票机制,通过集成学习的方式进行分类决策。此外,CV-ELM可自适应确定网络参数,在分类准确率、鲁棒性及泛化能力方面均优于传统的ELM。最后,基于新英格兰10机39节点系统的算例仿真结果证明了所提模型的有效性。

     

    Abstract: Rapid and accurate online static voltage stability assessment is important guarantee to ensure secure and stable operation of large-scale interconnected power grids. In allusion to such defects existing in traditional neural network learning model as many and diverse parameter invocation, long training time and large number of sample requirements, based on constrained voting extreme learning machine (abbr. CV-ELM) an online static voltage stability assessment model was proposed. Based on the sample differences among classes the difference vector set was constructed by CV-ELM to calculate the weights of the input layer to the hidden layer and the bias item of the hidden layer nodes, and the majority voting mechanism was led in to conduct the classification decision by ensemble learning. Furthermore, the network parameters could be determined by CV-ELM adaptively and in the aspects of the classification accuracy, robustness and generalization ability were better than those by traditional ELM. The effectiveness of the proposed model is proved by simulation results from the computing example based on New England 10-machine 39-bus system.

     

/

返回文章
返回