Multi-point synchronous temporary grounding wire detection device based on convolutional neural networks

Abstract

To prevent grounding accidents caused by forgetting to remove the temporary grounding wires and the contact of trees with the distribution network, a multi-point synchronous temporary grounding wire detection device based on convolutional neural networks is developed. The developed device can judge the type of groundings on the test line quickly and locate the ground points accurately during the cut-out. Three-phase alternating current (AC) signals with a specific frequency are injected at multiple points on the line by multiple groups of the device controlled by Raspberry Pi 3B + and embedded Android. The device can achieve the approximate location through the impedance ranging method and the accurate location by integrating the data from multiple points. The trained EfficientNet-B0 algorithm is utilized to predict the grounding positions. The tested results of the prototype verify the accuracy and rapidity of grounding diagnosis and positioning.