Multinodal Load Forecast Using Euclidean ARTMAP Neural Network

Forecasting electric demand is a fundamental part of the electric power systems, since, it provides useful information on several aspects of the network, acting directly in the planning of generation, transmission and distribution of energy and consequently in the economy of the resources. This work seeks to explore the application of artificial neural networks on the prediction of electric load considering several points of the electrical network (multinodal prediction). A neural model based on adaptive resonance theory (ART), called the Euclidean ARTMAP neural network, was used. This methodology can obtain significant results for the electrical load prediction in a fast, accurate and reliable way. In order to carry out the prediction, the Euclidean ARTMAP neural network was applied in each module (substation) as a Predictive Load System of the Substation (SPCS), which performs the prediction of the loads in an individualized way. Thus, to verify the efficiency of the proposed system, historical data of electrical loads of three substations of the New Zealand Electrical Company were used, aiming to obtain forecasts with a horizon of 24 hours ahead.