Due to the ever-increasing use of non-linear loads and their undesired effects on distribution systems operation,
harmonic analysis should be taken into consideration. On the other hand, the probabilistic nature of power
systems makes it necessary to consider the harmonic analysis in a probabilistic environment. In this paper, a data
clustering based algorithm is used for probabilistic assessment of harmonic load flow, for the frst time. Despite
the previous probabilistic harmonic load flow (PHLF), in which uncertainties are considered on grid connected
renewable generations, load demands, generators, transmission lines probable failure, etc., this paper considers
uncertainties on the location and non-linear load portion of nodal loads. Moreover, an organized PHLF algorithm
is formulated in this paper. In order to show the superior abilities of the proposed method, the method is applied
on the IEEE 37 node test system and the results are compared by the Monte Carlo simulation (MCS) method.