Stochastic load flow is a method for calculation of the effects of inaccuracies in input data on all output quantities through the load flow calculations. This gives a range of values (confidence limit) for each output quantity, which represent the operative condition of the system, to a high degree of probability or confidence. This paper presents a new method for stochastic load flow analysis using artificial neural networks. It is desirable to know the state of the power system in a range with certain confidence, with consideration of input data uncertainties and inaccuracies, on instant-to-instant basis in the fastest possible way. Present method using artificial neural networks to stochastic load flow problem is an effort in that direction and will be a very useful technique in effectively dealing with demand side uncertainties for power system planning and operation. The proposed artificial neural network model has been tested on a sample power system using two different training algorithms and simulation results are presented.