Performance Evaluation of Neural Network Techniques for Short-Range Wind Energy Estimation

Several countries all over the world are experiencing major consequences for the environment, and the most viable approach for mitigating these is the consumption of renewable energy, particularly wind power. However, the utilization of marine wind energy is fast expanding to satisfy rising electricity demand. Researchers have realized the need to sacrifice a near-perfect forecast of output power. Wind energy is impacted by variations in climate conditions, particularly wind speed. In this study, we estimated wind turbine output power using a multilayer feedforward neural network (MLFFNN). The MLFFNN was built with the model of the wind speed and used for simulation purposes as major input data and calculated power (Output power) as its output. The anticipated MLFFNN was trained by applying two types of machine learning algorithms. Firstly, Levenberg-Marquardt (LM) which produces a very low mean square error (MSE) and training errors. For checking the effectiveness, another set of data, that was not applied for training, was used to test and examine the performance of the trained MLFFNN-LM. This indicated that the trained MLFFNN-LM predicted power accurately and efficiently. Secondly, Error Back Propagation (EBP) was used for the training process with the same data which was used in MLFFNN-LM training. Evaluations of the training and testing procedures indicated that the MLFFNNEBP had slightly better results (lower MSE and training error) than the MLFFNN-LM.