An Artificial Intelligence Strategy for the Prediction of Wind Speed and Direction in Sarawak for Wind Energy Mapping

Accurate and reliable wind speed and direction prediction is one of the necessary concepts in implementing a wind energy system. In this paper, meteorological and geographical variables were modeled via artificial neural networks (ANNs), taking terrain elevation and roughness class into account. The feedforward neural network (FFNN) with back propagation trained with Levenberg–Marquardt algorithm was utilized, with wind speed and direction as the target function in each model. The results obtained using the formulated topographical models showed a regression value R in the range of 0.8256–0.9883. The optimum network based on the lower mean square error and fast computation time was 9-152-1. Thus, the developed topographical feedforward neural network (T-FFNN) is efficient to predict the wind speed and direction properly.