In a high wind penetration scenario in electricity production, the availability of models for synthetically generating hourly wind speed data becomes increasingly important for both sizing and modeling integrated renewable systems. A methodology for generating hourly wind speed time series is presented in this paper, adopting a physical-statistical approach based on some known aggregate input data. The proposed approach, developed in a previous study by the same authors, has been here improved using a new diurnal wind speed profile function. The improved methodology is first described and validated through experimental wind speed data, comparing the results with those obtained using the basic model. Then, hourly wind speeds generated with the improved methodology are used as input data for the optimal design of grid-connected and off-grid wind power plants. For each configuration, the influence of the diurnal wind speed profile and wind regime on system sizing and its economic parameters has been evaluated. Results have shown that the diurnal variation of wind speed does not affect the size of wind turbine, but strongly influences the storage capacity of off-grid wind power plants. (C) 2013 Elsevier Ltd. All rights reserved.