engleski

Neural networks for predicting hourly natural gas consumption

The aim of the paper is to create a prediction model of natural gas consumption on a regional level by using neural network methodology, and to analyze the results in order to improve prediction accuracy in further research. The output variable consisted of the next-day consumption of natural gas in hourly intervals, while the input space included previous-day consumption in addition to exogenous variables such as meteorological data (temperature prognoses, wind velocity, wind direction), season detection fuzzy variable, month, day type and day of the week. After conducting a feature selection procedure, two neural network algorithms were trained and tested: the multilayer perceptron and the radial basis function network with different activation functions. The results were analyzed regarding critical periods of time where the error is over 10%. Some critical hours within a day, as well as problematic days within the test sample were identified.

natural gas consumption; neural networks; multilayer perceptron; radial basis function network; fuzzy variable

nije evidentirano