The invention relates to a fault
feature extraction method of weak transient zero-sequence current on the background of strong
noise. The method comprises the following steps: firstly, transient
frequency compression ratio CR is determined, compressive sampling frequency fsc is determined based on the CR, and compression numerical value is determined according to the fsc so as to calculate step length hsc, wherein
numerical range of the CR is 1000-2000; secondly, transient zero-sequence current izy(t) under the background of strong
noise is input into a bistable
system, numerical calculation of the bistable
system is carried out by a fourth-order Runge-Kutta method, and an output
signal is is(t); thirdly, an initial period of the is(t) is extracted as a characteristic
signal ic(t); and finally, a
cross correlation coefficient pcz between the characteristic
signal ic(t) and an initial period izo(t) of
noise-free transient zero-sequence current is calculated, and fault
feature extraction of the transient zero-sequence current is finished if the pcz is greater than or equal to 0.7; and if the pcz is less than 0.7, the CR is adjusted until the pcz is greater than or equal to 0.7, so as to finish fault
feature extraction of the transient zero-sequence current.