The invention discloses an intelligent substation relay protection real-time reliability prediction method, which comprises the following steps: S1, establishing a two-parameter LN distribution modelof the failure rate of a single device, taking the life history data of the single relay protection device as input, and taking the failure rate of the single relay protection device as a cumulative density function; S2, carrying out linear conversion on the nonlinear LN distribution model, carrying out linear fitting, and solving a time-varying failure rate; and S3, establishing a time-varying Markov model of the protection system, determining the real-time transition probability between different states by a protection device LN distribution model, and respectively analyzing a transition probability matrix P and a transition density matrix A to obtain a Markov real-time state probability. According to the method, the real-time protection system state probability can be obtained by calculating the product of the state transition matrix and the initial state, short-time (one hour) state prediction is provided, state evaluation is achieved, and reference is provided for state maintenance.