Multi-kernel based Random Vector Functional Link Neural Network for Short-term Prediction of Wind Speed

Abstract

This work provides a wind speed prediction technique which is the combination of kernel functions and the random vector functional link neural network (RVFLN). The nonlinear kernel functions used in RVFLN called as MKRVFLN replace the traditional trial and error method to decide the number of neurons in hidden layer and also their appropriate activation functions. The MATLAB results demonstrates a comparison between ELM, RVFLN and MKRVFLN model. From comparison, the MKRVFLN forecasting model shows greater prediction accuracy. For wind seed prediction, the samples are collected at 10 minute, 30 minute, 1 hour and 3hour intervals of time from the wind farm named Sotavento locate in Spain.