Abstract:
Presents the application of an artificial neural network in the distribution transformer load forecast using kWh consumption of the connected electrical consumers as inputs of the neural network, which was trained in advance by the sample of data. The designed outputs are kW load factor and power factor at peak load, which will be used to calculate the peak load, utilization factor, and average loading of each transformer. The results are then compared with the outcomes from a statistical method. This thesis applys the modified back-propagation (MBP) algorithm to train two layer feed-forward neural network using sampled transformer data from the direct measurement for data collection, and other related information from Provincial Electrical Authority (PEA) data base systems as the training data. The developed forecast method is practical for further applications purposes. The test results show that an average error of the forecasted values obtained from neural network is satisfactory compared to that obtained from statistical method.