Design of Transformer Substation Fault Prediction Algorithm Based on Modified ESN

Temperature variation of some important nodes in the substation can reflect the danger operation of substation equipment. By predicting the temperature we can discover the potential hidden fault in the substation. An improved algorithm for predicting the temperature of the substation in this article is presented in this paper, which combines Echo State Network (ESN) and Local Average Denoising Method (LADM) in the phase space of time series. Simulation results show that this new algorithm acts well in predicting both noise-free time series and time series with high-level noise collected in the field.