Biogeography-Based Optimization (BBO) Trained Neural Networks for Wind Speed Forecasting

Abstract

Due to the advantages associated with artificial neural networks, it is used for forecasting problems. The accuracy of ANN depends on the structure, training data sets, and training algorithm. ANN training process is a complex task and can be considered as an optimization process with the aim to evaluate and train the ANN. There are many parameters in ANN which can be optimized such as the number of input, hidden and output nodes, weights, learning-rate, momentum rate, bias, minimum error, and activation/transfer function. In recent years, many researchers have proposed various optimization algorithms for optimizing the training of neural networks. In this paper, the biogeography-based optimization (BBO) algorithm for feed-forward neural network training is proposed. The BBOANN training algorithm effectiveness and accuracy is investigated on wind speed forecasting problem. The BBOANN training algorithm proposed is evaluated with the wind speed taken from Jaipur, Rajasthan, India; and the forecasting performance is compared with real values with reference to statistical error parameters. The comparison of forecasting performance is presented with existing methods for wind speed forecasting. Results clearly demonstrate the effectiveness of BBOANN algorithms with respect to forecasting accuracy, convergent speed, etc.