Abstract: Due to the influence of meteorological changes and human activities, the consistency of runoff sequences is often destroyed, and it is necessary to directly study the non-consistent hydrological sequences. Based on the Generalized Additive Model (GAMLSS), the effective runoff area of the basin under the control of the reservoir and dam was selected as covariates for simulation analysis, and the average design life level method (ADLL) and equivalent reliability method (ER) were used to estimate the design annual runoff and peak flood flow of the non-stationary series. The results show that the two hydrological sequences are non-consistent, and the fitting effect of the low quantile curve and high quantile curve of the constructed multi-covariate annual runoff model is good, and the overall simulation effect of the flood peak flow model is good. The GAMLSS model and the corresponding ADLL and ER methods can provide higher accuracy of the design values of peak flood flow and annual runoff in a specific return period or probability interval, and better reflect the influence of the hydrological sequence change trend on the design value.