A COMPARISON OF NEURAL NETWORK ARCHITECTURES FOR ULTRA SHORT TERM WIND POWER FORECASTING: A CASE STUDY OF THAI HOA WIND FARM IN BINH THUAN

Abstract

This paper proposes a comparison of three types of artificial neural networks, including the multilayer perceptron neural network (MLP), the radial basis function (RBF) neural network, and the Elman neural network, for ultra-short-term forecasting of wind turbine power generation. Experimental results, based on the power generation data from the 5MW wind turbines at the Thai Hoa, Binh Thuan wind farm, indicate that the multilayer perceptron neural network achieves optimal computational speed and accuracy. This research helps to provide a suitable method for the ultra-short-term forecasting of wind turbine power generation in cases with limited computational capacity and input data