Data-Driven short-term prediction for top oil temperature of Residential Transformer

Abstract

Focusing on the background that the state promotes the development of electric distribution network towards digitalization and intellectualization, this paper presents a model for predicting transformer oil temperature by using the electrical data collected by TTU without input parameters, and KS test is used to analysis the sensitivity relationship between predicted oil temperature and the TTU collected data. A comparative analysis of the multi-parameter sensitivity of the factors affecting the oil temperature rise of oil-immersed distribution transformers is carried out by using KS test.The LSTM algorithm is used to overcome the timeliness of the missing information of the GRNN algorithm in predicting the oil temperature, to solve the defect of data processing without timeliness information in the top oil temperature prediction of the oil immersed electric distribution transformer, and to analyze the internal thermal effect of the oil immersed transformer. The heating and heat dissipation mechanism of the transformer is described. The top oil temperature of the transformer is used to replace the hot spot temperature of the transformer to reflect the overall insulation of the transformer.