A short-term wind speed forecasting model based on improved QPSO optimizing LSSVM

The accuracy of short-term wind forecasting is important to guarantee the accuracy of wind farm power forecasting. An improved QPSO(Quantum Particle Swarm Optimization) algorithm for LSSVM(Least Squares Support Vector Machine) parameters selection is proposed based on the analysis of the QPSO and LSSVM. And then, with the weight factor m_best (average optimal position of the particle swarm) being introduced, the global search capability of QPSO is improved to optimize important parameters during the modeling process, by which the generalization capability and learning performance of LSSVM model is improved. The simulation results show that the proposed method can significantly improve the predicting accuracy. However, the mean error of the predicted wind velocity is only 2.43%, which satisfies the requirements of predicting accuracy.