Multi-Neuron Functional Link Artificial Neural Network: A Novel Architecture and its Performance for Wind Energy Prediction

Abstract

In this paper, a novel architecture, multi-neuron functional link artificial neural network (MNFLANN), has been proposed and its performance in predicting wind energy is compared with the other conventional network models, i.e. ANN, multi-layer perceptrons (MLP) and functional link artificial neural networks (FLANN). The name, i.e. MNFLANN is given as per its structure which consists of multiple neurons unlike the conventional FLANN that consists of only one neuron in the output layer. The real-time wind energy data of October month of recent three years from Sotavento wind farm located in Spain has been taken into consideration to evaluate the performance of MNFLANN. Results show that the mean absolute percentage error (MAPE) during testing is so less, i.e. -1.32% for MNFLANN, compared to other conventional architectures, i.e. -9.47% for ANN, - 8.44% for MLP and 15.19% for FLANN. The proposed MNFLANN architecture effectively handles the nonlinearity in input data compared to other conventional architectures due to its improved structure.