Predicción de la Demanda Eléctrica en Barra de 69 kV del Sistema Eléctrico de Subtransmisión de la Empresa Eléctrica Riobamba S.A.

Abstract

This document presents a model of an Artificial Neural Network (ANN) for the medium term electrical demand forecasting in a 69 kV bus of the Electrical Sub-transmission System of Riobamba’s Electrical Company. Different cases were developed for various conditions in order to find the best Artificial Neural Network, for that matter, the ANNs were trained with monthly electrical energy historical data, economical and/or demographical of Chimborazo’s province. The cases were analyzed by following different ANN topologies where 4 cases have individual outputs (next month consumption) while 5 cases have simultaneous outputs (consumption in the next 12 months). It was found that the simultaneous output models improve the data forecasting of the actual method used by Riobamba’s Electrical Company in certain ways. Cases that involve individual outputs didn’t have success in achieving the data modeling. The main objective is to find a valid model that allows to forecast the electrical demand in a 69 kV bus of the Electrical Sub-transmission System of Riobamba’s Electrical Company, these data correspond to the monthly Electrical energy consumption in MWh, so for that matter it was necessary to create a table with monthly data (12 months) from the year 2008 to the year 2015 (8 years) for a total of 96 monthly data that will be used as inputs in the training of the Neural Network and a total of 12 monthly data correspondent to the year 2016 to validate the network, Feed-forward multilayer Perceptron type Neural Networks were implemented while applying the back-propagation algorithm for the training. Data sets were divided into a training set, along with a testing set to compare real observations of the year 2016 and the ones using the ANN. The mentioned mid term goes from January 2017 to December 2017.