engleski

Load Forecesting in Power Distribution System Using Neural Networks

This paper describes and analyses history data measured in a transformer substation in DSO (Distribution System Operator) Elektra Slavonski Brod, part of Croatian Electricity Board. Prediction process, which is based on neural network logic, uses collected data of two power lines, Ljufina and Grad. Detailed forecasting model is given, based on load-temperature dependence. Neural network behavior in learning and function approximation process is shown with various learning factor and momentum. Comparison graphs are also shown, indicating the relative prediction error between measured and estimated load.

Distribution power system; load forecast; neural networks; comparasion; measured values

nije evidentirano