Abstract:
Interest in the topic of prediction has increased in recent years and appeared modern methods such as Artificial Neural Networks models, if these methods are able to learn and adapt self with any model, and does not require assumptions on the nature of the time series. On the other hand, the methods currently used to predict the classic method such as Box-Jenkins may be difficult to diagnose chain and modeling because they assume strict conditions.
So there was a need to compare the traditional methods used to predict the time chained with neural networks method to find the most efficient method to predict, and this is the purpose of this study.
Contributes to predict future demand for electricity in the electric power sector to solve problems through future planning to meet changes in the demand for electricity increases. Experience has shown there is no way of certain predict appropriate for all cases, but that in each case the way of a private predict is needed to find and use. However, taking more than one way may lead to raising the future accuracy of the estimates.
The present study aims to shed light on some of the statistical methods used to predict future demand for electricity for the Southern District, as well as a reference to more accurate methods to predict the future of energy. It has been used a number of methods to predict , such as econometric modeling technique, style and Box- Jenkins method of artificial neural network. And service to the goal of the study, which is based upon the premise that search: the neural network models more accurate than traditional models in long-term. As it is the most efficient and more accurate than other conventional models in dealing with non-linear time-series data.
We have been using the annual electrical energy consumption data for the Southern District to conduct a comparison of the program through the application of SPSS and Minitab for statistical analysis, and Matlab language has been used to build a program in neural networks, and through the practical application it was found that neural networks gives better results and more efficient than the classic way.
