Electrical energy prediction study case based on neural networks

TL;DR

This paper explores neural network-based methods for predicting electrical energy consumption, focusing on the Banat region, and demonstrates how preprocessing improves prediction accuracy, with potential applications beyond energy forecasting.

Contribution

It introduces a neural network approach with preprocessing techniques for energy prediction, applicable to general prediction tasks beyond energy consumption.

Findings

Neural networks can effectively predict energy consumption.

Preprocessing enhances prediction accuracy.

Method applicable to various prediction problems.

Abstract

This paper presents some considerations regarding the prediction of the electrical energy consumption. It is well known that the central element of a microeconomic analysis is represented by the economical agents actions, actions that follow their own interest such as: the consumer maximization of his satisfaction, the producer maximization of his profit. The study case is focused on the prediction of the sold energy in Banat region. The goal of this study case is to optimize the electrical energy quantity purchased from the producer by the energy distributor in Banat region. The prediction is based on neural networks. There are used feed-forward and Elman type neural networks. In order to enhance the prediction accuracy there have been used both linear and nonlinear preprocessing units. The aspects considered in this paper can be extrapolated in any general case of prediction based application, not only in the already stated case of electrical energy.