基于SVM-SMOTE算法的一维卷积神经网络电力系统暂态稳定评估模型

A 1D convolutional neural network-based transient stability assessment model for power systems based on the SVM-SMOTE algorithm

为了提高电力系统运行稳定性,降低大停电事故发生的概率,本文提出了一种基于SVM-SMOTE算法的一维卷积神经网络暂态稳定评估模型。为了避免人工特征选择引入的主观偏差对模型预测性能的影响,本文选择来自PMU的底层量测数据作为输入特征,并采用一维卷积神经网络(1D-CNN)捕捉输入特征的时序信息;考虑数据集样本不平衡带来的预测精度下降问题,采用SVM-SMOTE算法对样本进行均衡化。算例仿真结果表明,本文所提出的模型实现了端到端的时序特征提取和暂态稳定评估,可满足在线评估准确性、快速性和可靠性的要求,且有效解决了不平衡数据集中失稳样本漏判率高的问题。

In order to improve the operational stability of power systems and reduce the probability of major outages,this paper proposes a 1D convolutional neural network transient stability assessment model based on the SVM-SMOTE algorithm.To avoid the impact of subjective bias introduced by manual feature selection on model prediction performance,the underlying measurement data from PMU is selected as input features and a one-dimensional convolutional neural network(1D-CNN)is used to capture the time-series information of the input features.At the same time,the SVM-SMOTE algorithm is used to equalize the samples,considering the degradation of prediction accuracy due to the imbalance of samples in the data set.The simulation results of the algorithm show that the proposed model achieves end-to-end timing feature extraction and transient stability evaluation,which can meet the requirements of online evaluation accuracy,rapidity and reliability,and effectively solve the problem of high omission rate of unstable samples in unbalanced datasets.