基于半监督学习的输电线路状态预测

State Prediction of Transmission Lines Based on Semi-supervised Learning

输电线路状态评估及预测对于合理制定运维策略、提高运维水平具有重大意义。针对现有模型无法兼顾鲁棒性和数据需求量的问题,本文提出一种基于半监督学习的状态预测方法。首先,对拓展后的特征向量,利用正则矩阵填补缺失数据,并通过表征学习解决稀疏编码问题。然后,借助少量标注样本初步确定线路区段在不同缺陷状态下的类别中心。最后,使用未标注样本对模型估计参数进行修正。算例分析表明,该方法与现有模型相比,识别准确率大幅提升且数据使用效率更高。

The state evaluation and prediction of transmission lines are of significance for the formulation of operation and maintenance strategies and the improvement of operation and maintenance levels.Aimed at the problem that the ex⁃isting models cannot take both the robustness and data demand into account,a state prediction method based on semisupervised learning is proposed in this paper.For the extended feature vectors,regular matrix is used to fill the missing data,and the sparse coding problem is solved by means of embedding learning.A small number of labeled samples are used to preliminarily determine the class centers of line sections in different defect states,and then the estimated pa⁃rameters are modified using the unlabeled samples.The results of an example show that compared with the existing mod⁃els,the proposed method has a much higher recognition accuracy and a higher data utilization efficiency.