سال انتشار: ۱۳۸۵

محل انتشار: اولین کنفرانس بین المللی مدیریت و برنامه ریزی انرژی

تعداد صفحات: ۶

نویسنده(ها):

A.Azadeh – Research Institute of Energy Management and Planning and Department of Industrial Engineering, Faculty of Engineering, University of Tehran, Iran
A.Kheirkhah – Department of Industrial Engineering, Faculty of Engineering, Bu- Ali Sina University, Hamadan,Iran
M. Saberi – Department of Industrial Engineering, Faculty of Engineering, Bu- Ali Sina University, Hamadan,Iran

چکیده:

By looking at the forecasting of Electricity consumption we will explain the application of neural networks to time series analysis. Electricity consumption represents two essential attributes; firstly it shows the strong monthly changes and secondly, clearly shows the increasing
trend. The multilayer perceptron with back propagation is used which is a supervised learning strategy and ideally suited to forecast problems.
Neural network is a strong rival of regression and time series in forecasting. In this paper shown that using neural networks with preprocessed input data would have less error than neural network with raw input data. Also it is shown that neural networks dominate time series approach from point of yielding less mean absolute percentage error( MAPE). The purpose of this model is to find the essential structure of data and eliminate the trend of it with preprocessing techniques to forecast future consumption with less error.