A Study of Wind Speed Prediction Based on Particle Swarm Algorithm to Optimize the Parameters of Sparse Least Squares Support Vector

Wind speed forecasting can accurately improve prediction efficiency of wind power in wind farm, decrease failure probability of wind turbine, and extend life cycle. An innovative algorithm is proposed to optimize both the parameters of least squares support vector machine (LSSVM) and the procedure of finding sparse support vector. Firstly, the defects of support vector are analyzed. Then inequality constraints are replaced by equality constrains. Quadratic programming problem is transformed into linear equations through Lagrange method to solve goal function. Solving process for least squares support vector is deduced and the reasons why LSSVM does not possess the sparse property are analyzed. Parameters of the LSSVM model and the procedure of finding sparse support vector are optimized by particle swarm optimization (PSO) algorithm. Then based on actual wind speed, prediction performances of three kinds of forecasting methods, including sparse LSSVM optimized by particle swarm algorithm (SPSO-LSSVM), auto regressive moving average (ARMA) and artificial neural networks (ANN), are compared. The results show that the forecasting performance of SPSO-LSSVM is the best. The effectiveness of proposed algorithm is verified by simulation.