One of the important parameters in the design of flood control structures is to determine flood peak discharge for various return periods. A primary issue of planners in the face with flood is lack of data or insufficient data. One of the most reliable strategies is generalizing the results from sites with observed data to ungauged locations. The main goal of this study is regional flood frequency analysis using multiple regression method for Qazvin province of Iran. 8 out of 23 existing hydrometric station were removed because of the short-term statistics and construction of storage dam at upstream. The results of factor analysis showed that perimeter, equivalent diameter, time of concentration, length of main waterway and area were the main variables affecting ﬂood magnitude. The remaining 15 stations were divided into two homogenous regions using cluster analysis. Homogeneity of these two regions was confirmed using homogeneity and heterogeneity tests of L-moments. Based on the best-fit criteria of Zdist, GNO distribution with the statistic of 0.29 has the best fit for the entire region but for one and two homogeneous regions, GLO and GPA distributions with the statistics equal to 0.09 & 1.56 have the best fit respectively. After calculating parameter values for selected distributions, discharges with different return periods were estimated for all stations. Then, regression relations were obtained between peak discharge and factors affecting flood peak for each return periods at two homogeneous regions. Peak discharges at ungauged locations can be estimated for different recurrence interval using these relationships.