The power outputs of photovoltaic generators (PVs) are strongly random

and appreciably correlated in close proximity. The research on probabilistic load flow of distribution network mainly focuses on how to deal with correlated random variables. Nataf transformation is used to sample correlated non-normal random variables and the method of combining Nataf transformation with the point estimate method (PEM) is proposed to develop a new correlated probabilistic load flow method. Compared with the orthogonal transformation

Nataf transformation is able to reflect the changes of correlation coefficients in different sampling spaces

and therefore it is more accurate. The method is applied in a modified IEEE 33-bus test system with PVs integration for calculation of average relative errors in bus voltages. Compared with orthogonal transformation

the advantage of Nataf transformation is verified. Furthermore

a sensitivity analysis is made on three factors

the correlation coefficients of solar radiation

the load power fluctuation and different estimating points. This work is supported by National Basic Research Program of China (973 Program) (No. 2013CB228203).