The invention provides a distribution network fault line selection method based on a cascade bistable system. The method comprises the steps that 1 a potential function parameter in the cascade bistable system is set; 2 the acquired transient zero sequence current in(t), which includes noise, of each line is normalized to acquire normal zero sequence current gn(t); 3 s(t)+gamma(t) in the potential function parameter y(t) is set as gn(t); a fourth-order Runge-Kutta algorithm is used to solve yn(t); and yn(t) is reverse-normalized to acquire process characteristic zero sequence current rn(t); 4 after yn(t) is acquired, the fourth-order Runge-Kutta algorithm is used to solve X, and the acquired X is called as second-order current sn(t); 5 sn(t) is reverse-normalized to acquire the transient characteristic zero sequence current cn(t) of each line; 6 the cross-correlation coefficient matrix rho n of each line cn(t) is calculated; 7 a polarity parameter Jn and an energy parameter En are acquired; and 8 the line selection distance dn is defined, and the maximum line of dn is determined as a faulty line. According to the invention, in a strong noise background, fault line selection is accurate, and the method has a good performance in various different fault situations.