The time-varying failure rate（TFR） has advantages over the mean failure rate（MFR） in accuracy for reliability assessment. The estimation approach for the TFR of the relay protection device（RPD） with the field data is proposed. According to the feature of RPD failure， the data is first divided into random failure data and aging failure data. And then for the random failure data， the constant random failure rate（RFR） of RPD is drawn up with estimating exponential distribution function， for the aging failure data， aging failure rate（AFR） is estimated with the Weibull， normal， lognormal and Rayleigh distribution functions， and the best estimation is determined by the coefficient of determination. Simulation results show that the aging failure data of real power grid obeys the Weibull distribution law. Furthermore， the general TFR is， obtained with the addition of the RFR and AFR， estimated with the field data in a regional power grid. Finally， considering the geographical features of the power grid， environmental and regional factors， failure rate is calculated and the conclusion is drawn: MFR is basically consistent; AFR depends on climate.