In order to make better use of historical flood information and improve the parameter estimation precision, this paper studied the expected probability weighted moments method for general extreme value distribution when historical flood information is presented. The statistical performance of the expected probability weighted moments for estimation of population parameters by Monte-Carlo method was performed, and the method of the partial probability weighted moments was compared. The results show that the expected probability weighted moments ha stable statistical characteristics, and it is a simple, stable and effective method when historical flood information is available, and can be extended to other hydrological fields.