Prediction for Power Supply Area Interrupted by Heavy Rainfall in Guangxi Power Grid

Abstract

In view of the problem of flood disaster caused by heavy rainfall, which affected power supply reliability, a dynamic prediction approach for power supply area interrupted by heavy rainfall is proposed. First, under the spatial correlation rules, the power supply area within the surface rainfall boundary is judged by multidimensional data, such as digital geography. Secondly, with the distribution transformer of the power supply area as the geometric center, the area and the scale of power outage users are solved. Thirdly, the supervised learning algorithm is used to predict the disaster area out of service, on the basis of the binary classification of prior data. Finally, the dynamic changes of the predicted results are calculated according to the intensity changes of the rainfall process. Using the confusion matrix and the receiver operation characteristic curve to verify the prediction method. The numerical results show that this approach can dynamically predict the power supply area interrupted by heavy rainfall, which can be used for the production command in the process of “water rise and power failure, water back and power recovery”.