Cable Situation Awareness and Early Warning Method for Multi-source Input Fusion

Current cable situational awareness methods are often judged through data from a single source, and still require manual analysis and prediction based on monitoring information, which cannot fully meet the requirements of cable monitoring. Based on multi-source data fusion technology, this paper finally formed a multi-dimensional sequence feature sample by correlating and mining cable alarm information, operating voltage and current data characteristics, partial discharge signal map, partial discharge pulse signal and signal harmonic monitoring data. In order to effectively represent the sample Relationships, design composition rules based on sequence similarity, model data to map temporal features and potential relationship information into graph space, and use a classification model based on graph convolution to capture potential relationships that are beneficial to perception by aggregating data features Then, the cable situation can be sensed and predicted. Experimental results show that this model can obtain more accurate status evaluation values. Compared with the less-dimensional data source status evaluation model, the prediction accuracy is increased by 11.5% on average, providing sufficient time for monitoring personnel to deploy cable operation and maintenance work in advance.