Research on Partial Discharge Identification of Power Transformer Based on Chaotic Characteristics Extracted by G-P Algorithm

In the past 20 years, China's power industry has developed rapidly, which has put forward higher requirements for the reliability of electrical equipment. One of the important factors to ensure the normal operation of electrical equipment is the insulation condition of electrical equipment. Partial discharge is the main cause of power insulation deterioration, and it is also the main characteristic of power transformer insulation degradation. Experience shows that partial discharge will not cause severe insulation damage temporarily, but in a long time, partial discharge will gradually develop, leading to more serious insulation failure. Therefore, it is necessary to monitor the power transformer in operation and diagnose the type of insulation fault by identifying PD mode. In recent decades, researchers have done a lot of research on the statistical characteristics of partial discharge signals and proposed a variety of analytical methods. The most commonly used method is statistical feature extraction based on phase resolved partial discharge (PRPD). What's more, fractal geometrical analysis, Characteristic of pulse waveform and wavelet analysis were used in PD analysis. However, if the correlation between and the interaction between different PD pulses are not used, the traditional method can only obtain the image feature patterns, which leads to the lack of PD recognition. Chaos theory is a branch of mathematics, which mainly studies the behavior of dynamical systems highly sensitive to initial conditions. Chaos theory involves deterministic systems whose behavior can be predicted in theory. A chaotic system can predict a period of time, and then it seems to become random. After decades of development, chaos theory has become a hot topic. There are complex chaotic phenomena in PD process. When partial discharge occurs in a power transformer, external factors, such as the PD type, temperature, water and sometimes even voltage of the partial discharge, remain unchanged, so the system is a deterministic system free from external interference. However, the generation of partial discharge, partial discharge waveform and partial discharge pattern still have strong randomness. Therefore, the randomness of partial discharge is inherent randomness, that is, the partial discharge signal under the same voltage and phase has chaotic characteristics. One of the characteristics of the chaotic system is the chaotic attractor which reflects the regularity of the chaotic system. The purpose of phase space reconstruction is to restore chaotic attractors in high dimensional phase space. In 1983, Grassberger and Procaccia proposed a G-P algorithm for computing attractor associated dimensions from time series, which is used to reconstruct phase space and extract chaotic characteristics. This paper introduces a method of PD signal recognition based on chaotic characteristics of PD signal in power transformer. Lyapunov exponents are extracted from different sequences to construct feature groups. This method can achieve pattern recognition well.