Estimates of hydroelectric energy generation in Turkey with Jaya algorithm-optimized artificial neural networks

Abstract

The main purpose of this study was to establish an artificial neural network (ANN) model trained by a Jaya algorithm, and use the model to predict Turkey’s future hydroelectric energy generation (HEG). Population, gross domestic product (GDP), installed capacity, energy consumption, gross electricity energy demand (GEED), and average yearly temperature (AYT) data were inputted as independent variables in the model. ANN-Jaya was compared with ANN models trained by the other two high performance optimization methods, namely back-propagation (BP) and artificial bee colony (ABC) algorithms, to test its accuracy. The ANN-Jaya model converged to smaller error values than were obtained with the ANN-BP and ANN-ABC models for both the training and test datasets. When the average relative error (RE) values calculated for the test set are taken into account, ANN-Jaya performs 19.3% better than ANN-ABC and 31.2% better than ANN-BP. Therefore, Turkey’s HEG projections were made out to the year 2030 using an ANN-Jaya model in a low and a high energy demand scenario. According to the developed projections, HEG values in Turkey in 2030 will be in the range of 104.81–124.66 TWh. The present results affirm that HEG can be modeled accurately with an ANN-Jaya technique and this method was shown to be advantageous for predicting future HEG.

Keywords

• jaya algorithm
• artificial bee colony algorithm
• hydropower generation
• neural networks