用于在线预测静态电压稳定性的SIPSS-Lasso-BP网络

发布时间：2018-05-13 08:41

本文选题：静态电压稳定 + 神经网络　；参考：《中国电机工程学报》2014年34期

【摘要】：快速求解系统负荷能力极限是在线评估电力系统电压稳定性的基本要求。提出一种用于离线拟合并在线预测负荷能力极限的SIPSS-Lasso-BP网络。该网络由基于电网状态相似度指标(similarity index of power system state,SIPSS)的样本筛选方法、最小绝对值收缩选择(least absolute shrinkage and select operator,Lasso)方法和BP(back propagation)神经网络共同组成。基于SIPSS的样本筛选方法以样本负荷能力极限值和电网状态相似度量化指标为依据,对训练样本进行筛选。Lasso方法对训练样本进行回归分析,确定各状态量中对负荷能力极限最具有解释性的系统状态量。BP神经网络通过精简后的训练样本来离线拟合负荷能力极限并用于在线预测。通过新英格兰39节点算例和某省实际算例对SIPSS-Lasso-BP网络的测试结果表明,该方法能够在保证预测精度的情况下明显提高BP神经网络的离线训练效率。
[Abstract]:The rapid solution of system load capacity limit is the basic requirement for on-line evaluation of power system voltage stability. A SIPSS-Lasso-BP network is proposed for off-line combining on-line prediction load capacity limit. The network is selected from the sample screening method based on the similarity index of power system state, SIPSS. The small absolute value contraction selection (least absolute shrinkage and select operator, Lasso) method and BP (back propagation) neural network together. Based on the SIPSS sample screening method, based on the limit value of the sample load capacity and the quantization index of the grid state similarity, the training sample is screened for the training sample by screening.Lasso methods. By regression analysis, the.BP neural network, which is the most explanatory system for the load capacity limit of each state, is used to fit the limit of load capacity off-line and be used for online prediction through the reduced training samples. The test results of the SIPSS-Lasso-BP network through a new England 39 node example and a provincial actual example show that this method is used. It can significantly improve the off-line training efficiency of BP neural network in ensuring the accuracy of prediction.

【作者单位】：浙江大学电气工程学院;国网浙江省电力公司电力科学研究院;
【基金】：国家863高技术基金项目(2011AA05A119)~~
【分类号】：TM712;TP183

【参考文献】