In all advanced weather forecasting centers, the use of probabilistic forecasts is very important and their application is increasing day by day. In this research, an 18-member ensemble system is formed, each member of which is an independent implementation of the WRF model with different physical configurations. Due to hardware limitations, creating an ensemble system with fewer members and maintaining efficiency is the main goal. In the BMA method, a weight is assigned to each ensemble member according to the model error history in a training period. In this study, by removing members with less weight, the size of the ensemble system has been reduced to 7 members. The comparison of probabilistic forecasts by both ensemble systems showed that the 7-member ensemble system has the same performance as the 18-member ensemble system.