With large-scale integration of wind farms with wind speed correlation into the bulk power system, it is necessary to take account of both the wind speed correlation and the load correlation in the probabilistic load flow (PLF) calculation. The existing PLF methods usually utilize the Monte Carlo simulation to deal with these two kinds of correlation. But this algorithm needs a large amount of computation and it neglects the difference of wind speed and load in probability distribution. In this paper, a procedure is established for calculating PLF based on the classified consideration of wind speed correlation and load correlation. The MC simulation combined with Latin hypercube sampling is used to acquire the probability density function of total wind power with correlations. An analytical method is used to get the normal distribution function of total loads as load always follows normal distribution. Then the convolution method is used to obtain the final result. As the generation of load samples is avoided, the proposed method can improve the PLF calculation efficiency. The effectiveness and accuracy of the proposed method is verified via the comparative tests on the IEEE RTS-96 system and the IEEE 118 system.