摘要
针对数据驱动的电力系统暂稳分析，考虑到安全域概念下的暂稳预测和预防控制对输入特征的不同要求，以及充分兼顾数据挖掘模型的直观性与准确性，该文提出两阶段支持向量机用于暂稳预测及预防控制。在第一阶段，采用可控特征和直观模型挖掘运行方式与暂态稳定的内在联系，并用于制定预防控制策略；第二阶段，采用复杂模型构建准确率高的暂稳预测模型；此外，第一阶段模型能够为第二阶段的暂稳预测模型的训练提供样本筛选的依据，从而缩短了预测模型的训练时间。仿真分析表明，经过数据缩减后，预测模型的训练时间大大缩短，所得模型仍具有较高的准确率；当系统被判断为不安全时，可提供发电机有功调整的预防控制措施，以保证电力系统安全稳定运行。
This paper focused on the data driven transient stability analysis of power systems. Considering different requirements of stability analysis and preventive control for input features, and in order to balance the accuracy and transparency of the data mining models, a two-stage support vector machines was presented for transient stability prediction and preventive control. In the first stage, a preventive control model was built using the controllable variables and simple model. In the second stage, a complex but accurate model was built for transient stability prediction. Moreover, the generated preventive control model can also provide a basis for the instance selection for the second stage. The test results show that the proposed prediction model can predict the transient stability accurately and with shorter training time; when the power system is predicted as unstable, preventive control measures can be provided to ensure the security and stability of power system operation.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2018年第1期137-147,共11页
Proceedings of the CSEE
基金
国家自然科学基金项目(51577009)~~
关键词
暂态稳定预测
预防控制
两阶段
支持向量机
特征选择
样本筛选
数据挖掘
transient stability prediction
preventivecontrol
two-stage
support vector machine
feature selection
instance selection
data mining
