The main objective of this paper was to perform medium-term electrical energy consumption forecast in several types of commercial building loads. The historical data were used to produce forecasting models using nonlinear autoregressive exogenous (NARX) input in neural network time series (NNTS). Three training algorithms were being tested, that is Bayesian Regularization (BR), Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG). Next, performance analysis by determining the mean squared error (MSE) was carried out to measure the accuracy of the models. The results exhibited the advantages of neural network time series as an accurate forecast model method and revealed that Bayesian Regularization (BR) gives the best accuracy among the three algorithms tested.
Volume 12 | 04-Special Issue
Pages: 1555-1560
DOI: 10.5373/JARDCS/V12SP4/20201635