To enhance the cost effectiveness of the distribution automation system (DAS), this thesis proposes the Artificial Neural Networks (ANNs) to derive the Line Switch Unit Commitment by minimizing the total cost of customer service outage and investment cost of line switches. A brief introduction of the smart grids and the DAS implemented by Taipower is described. The customer interruption cost is determined according to the customer type, loading, outage frequency and number of automated line switches in the feeder. The ANNs models were created for a radial feeder and an open tie feeder, and then implemented with the load growth in order to determine the year for the next line switch to be added. The neural network model for the line switch unit commitment is derived after performing the training using MATLAB/Neural Network Toolbox. A sensitivity analysis of the impacts of the loading and the outage frequency in the line switch unit commitment is studied in this thesis and a comparison between the radial feeder and the open tie feeder is also shown in the results. After the creation of the neural network model for the two types of feeder topology, we implement the model to determine the unit commitment of line switches for two Panamanian distribution feeders. The results of computer simulation show how many automatic line switches should be installed on the feeder for the first year and in which year the line switch should be added. It is found that the total cost function of customer outage and line switch investment can be minimized by considering the load growth of distribution feeders over the study period.