Factors which influence wind power impact on a power system include variability, uncertainty, geographical spread and correlations of different sites. Presented is the wind power variability of three sites in Kenya and their potential impact on the system reserve. The sites' power was generated from wind speeds using wind farm models developed in Matlab/Simulink software. The wind power and load data were analysed for hourly variations and the load following reserve requirements in five different futuristic sites integration scenarios. However, forecast errors were not considered. Results showed that power variations depended on distinctive wind regimes and the number of turbines. Ngong (25 MW) power was evenly distributed between 0 and 100%; Kinangop (60 MW) was concentrated between 20 and 60% and Turkana (300 MW) between 20 and 80% of their respective peak values. The standard deviations reduced as farm capacity increased due to the turbine smoothing effect. The reserve requirements increased on average, 30 MW per percentage wind integration. The combined Ngong/Turkana (325 MW) reserve requirements were less than for Ngong/Kinangop (85 MW), indicating the significance of site correlations.. Also, when Ngong was up-scaled to the same output level as the three interconnected sites (385 MW) reserve requirement increases by 25.5%, thus indicating the importance of geographical spread. (C) 2015 Elsevier Ltd. All rights reserved.