Development of distribution feeder loss models by artificial neural networks

This paper proposes an artificial neural network (ANN) model for distribution system loss analysis. To derive the hourly loading of each load bus more accurately, the topology process has been performed for network connectivity and all customers served by each distribution transformer has been identified by automatic mapping/facility management (AM/FM). According to the typical daily load patterns and the energy consumption by customers, the hourly loading of distribution transformers can therefore be obtained. A three-phase load flow program is applied to solve feeder loss for different scenarios of input neurons to create the training data set for the neural network. The Levenburg-Marquardt (LM) algorithm is used for ANN to derive the neural based simplified loss model of distribution feeders. Two distribution feeders of Taipower system are selected for computer simulation to demonstrate the effectiveness of the proposed method for distribution system loss analysis. It is found that the daily loss pattern of distribution feeders can be solved in a very efficient manner with ANN loss model by considering the key factors of feeder loading, feeder length and transformer capacity.