Forecasting of Weekly Electricity Consumption Using Neural Networks

Abstract: Neural networks have been shown to be effective in modelling time series, with applications in the forecasting of electricity consumption. In applying neural networks to weekly electricity consumption data, several issues, such as selection of network architecture, network structure and input structure need to be addressed. This paper addresses these issues in relation to the current application and also demonstrates that considerable value is to be gained from incorporating the lessons learned from linear time series modelling into the current nonlinear analysis. Results for national Irish weekly electricity data demonstrate the potential improvements which can be obtained using the neural network approach