Fault Early Warning of Composite Insulator Based on Time-Domain Temperature Field

The real-time temperature field of composite insulators can better characterize their operating conditions and deterioration trends. Currently, the temperature rise of insulators usually relies on manual experience to set the temperature rise warning threshold, lacking scientific monitoring methods and criteria. This paper proposes a new fault diagnosis method for composite insulators based on multi-dimensional data fusion combined with neural network analysis, that is, using RFID passive temperature measurement technology to accurately monitor the temperature distribution of each key part and the whole composite insulator, obtain massive real-time temperature field data, and combine with the multi-dimensional information such as environmental temperature, humidity, wind speed and mounted wire temperature, and then fed into a Layered Multi-level Neural Network, which will ultimately output predicted temperature and warning level. Through on-site experiments and simulation calculations, the neural-network designed in this article can achieve accurate temperature prediction and warning level output from on-site monitoring data, and its F1 score of characterization error rate has reached above 0.912. Due to the use of the layered model, compared with existing models, the iteration times are fewer and the speed is faster, which can more accurately and quickly achieve insulator fault monitoring and warning.