基于马尔可夫转移场和深度残差网络的电能质量复合扰动多标签分类OA北大核心CSTPCD

The disturbance of power quality in modern power system becomes complicated and diversified.Traditional classification methods are difficult to adapt to complex and diverse perturbations.The traditional single-label classification method is used in the research of recognition and classification based on neural networks.When there are compound disturbances outside the label set,the classification method can not be used.If the label set is to be updated,the whole classification model should be retrained.Therefore,this paper uses deep residual network to construct a more adaptive multi-label classification system,which can accurately identify the power quality disturbances(PQDs)of unknown tag combinations outside the training sample tag set.First,a Markov transition field(MTF)is used to transform the disturbance signal into a two-dimensional visual image,and a deep residual network(ResNet)is used to build nine binary classifiers to extract the disturbance features covered by the two-dimensional image.The disturbance classification is carried out by a multi-label classification system composed of 9 binary classifiers.The classification accuracy of the training samples in the label set is 97.58%,and the average accuracy of the disturbance signals outside the doping label set is 97.67%,which is much higher than the classification system of the same level.