Wind speeds at different wind farms with close geographical location have a strong correlation. The impact of wind speed correlation on probabilistic load flow（PLF） is studied by using a hybrid method of auto-regressive and moving average（ARMA） model and time-shifting technique. The factors with different influencing degrees are analyzed. The research focuses on three aspects of influence including probability density distribution of node voltage， wind farm maximum capacity and minimum reactive power compensation in order to meet certain requirements on voltage quality. The research results reveal that it can help to assess the impact of multiple wind farms on network voltage profile more reasonably and select better network planning scheme and operating modes if wind speed correlation is considered.