Multi-target prediction model of urban distribution system rainfall-caused outage based on spatiotemporal fusion

Abstract

Improving the prediction accuracy of weather-related outages in the distribution system is crucial for disaster prevention and mitigation. Due to the vulnerability of the distribution system in rainfall, we propose a multi-target prediction model of the rainfall-caused outages of the distribution system based on the fusion of spatiotemporal attributes. In this article, the ground subdivision model based on element reconstruction is firstly proposed to establish a spatial grid. Then the rainfall waterlogging model considering seepage, drainage and confluence is established to provide a targeted reference for distribution system outage prediction. Then the multi-target random forest based on superposition features is proposed to deal with the meteorological data, geographic environment data, power grid data and waterlogging data. The model outputs the rainfall outage probability and visualization results of ground and overhead equipment branded with time and space labels. Finally, the model is tested on the instance data and compared with other models.