With linear combination operation on B-spline kernel functions including B1, B3, and B5 and radial basis function (RBF) kernel functions, combined kernel functions (referred to as the B-RBF function) are proposed. Further, least squares support vector machines (LSSVM) using particle swarm optimization (PSO) based B-RBF functions are developed, termed PSO-B-RBF. To predict fluctuating wind velocity, optimization is implemented on penalty parameters and kernel parameters using the PSO algorithm. For comparison, prediction results of PSO-RBF-LSSVM are also taken into consideration. The numerical analyses demonstrate that PSO-B3-RBF-LSSVM has better performance in predicting fluctuating wind velocity with respect to PSO-B1-RBF-LSSVM, PSO-B5-RBF-LSSVM, and PSO-RBF-LSSVM.