基于多源数据融合的变电运维误操作风险识别研究

Study on Risk Identification of Misoperation in Substation Operation and MaintenanceBased on Multi-source Data Fusion

现有变电运维误操作风险识别方法的综合评判结果与实际值存在出入,识别精准性差,为此研究基于多源数据融合的变电运维误操作风险识别方法。首先选择合适的滑动窗口,结合抽象数据进行融合,对数据进行周期性行为的特征提取,当需要融合的向量维数不相同时,需要对低维的特征添加0进行补齐,直到所有向量的最高维特征相同。然后用非线性映射将数据映射到可分的特征空间中,并运用分类学习器将特征空间中的线性可分数据进行分类。最后运用判别函数对所有样本进行训练,得到最大K值,对其进行二进制编码,计算编码距离并找出距离最短的一行,从而完成分类识别。实验结果表明,实验组的综合评判结果为0.07,与风险评价论域中实际结果一致。这表明实验组能够精准识别风险,并达到了较好的识别效果。

data to extract periodic behavior features from the data.When the dimensions of the vectors to be fused are different,it is necessary to add 0 to the low dimensional features for completion until the highest dimensional features of all vectors are the same.Then,by nonlinearly mapping the data into a separable feature space,the linearly separable data in the feature space is classified using a classification learner.Finally,the discriminant function is used to train all samples to obtain the maximum K value,which is binary encoded.The encoding distance is calculated and the nearest row is found to complete classification recognition.The experimental results indicate that the comprehensive evaluation result of the experimental group is 0.07,which is consistent with the actual results in the risk assessment domain.This indicates that the experimental group can accurately identify risks and achieve good recognition results.