Power systems operation is widely monitored through load flow analyses. The three main methods used in these analyses are Newton-Raphson (NR), GaussSeidel and Fast-Decoupled method. These methods involve long calculations and numerous iteration which leads to an increase in computation time. However, fast analyses are required for an efficient power system protection scheme. Therefore, an alternative method which consumes much lesser time to compute and able to provide accurate results are necessary. In this paper, the accuracy of Artificial Neural Network (ANN) is studied by comparing numerical and analytical results for an IEEE 14-bus power system network model. The study is done by varying the load parameters at bus 6 and the output voltages (in per unit values) of all the buses are recorded for ANN training and testing purposes. The ANN is coded using the MATLAB software and the result obtained is compared with the analytical and simulation results. Findings from this study suggest that the ANN could possibly be an alternative method for load flow solutions.