Real-Time Evaluation System of Transformer Winding Short-Circuit Withstand Capability Based on Symplectic Geometry Mode Decomposition and Extreme Learning Machine

Aiming at the difficulty of real-time monitoring of winding during transformer operation and insufficient methods for evaluating the short-circuit withstand capability of winding, a real-time evaluation system for the short-circuit withstand capability of transformer winding is proposed. The system uses the symplectic geometry mode decomposition(SGMD) method of the transformer vibration signal to obtain several components, and evaluates the short-circuit wtih stand capability of the winding by analyzing the eigenvector formed by the kurtosis factor of each symplectic geometry component. Measure the data under normal operation and different failure states many times to construct an intelligent decision-making system of extreme learning machine (ELM). Experiments show that, compare with the ensemble empirical mode decomposition (EEMD) method and complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) method, this method can better reflect the transient process of winding subjected to short-circuit shock, and has better real-time performance, and the model has higher correct rate.