基于混合蚁群和粒子群优化 LSSVM的脉动风速预测

摘要

为提高最小二乘支持向量机（LSSVM）对脉动风速预测的精确性，提出了基于混合蚁群和粒子群优化LSSVM的预测方法。对 LSSVM参数进行搜索寻优，主要分为两阶段：第一阶段，利用蚁群算法在参数空间进行全局搜索，实现对 LSSVM参数的初步寻优；第二阶段，利用蚁群算法获得的寻优结果初始化粒子群粒子位置，实行进一步的粒子群搜索寻优，获得更为精确的 LSSVM。运用基于混合蚁群和粒子群优化的 LSSVM对脉动风速时程进行预测，并与分别基于蚁群和粒子群优化的 LSSSVM预测结果进行对比。数值分析表明，基于混合蚁群和粒子群优化的 LSSVM预测方法精度高、鲁棒性强，具有工程应用前景。%In order to enhance the accuracy of least square support vector machines (LSSVMs)for fluctuating wind velocity prediction,the LSSVMwith hybrid ant colony optimization (ACO)and particle swarm optimization(PSO)technique was proposed here.A two-stage meta-heuristic optimization framework was introduced to find the optimal parameters of LSSVM.In the first stage,the global search in the parameter space was accomplished using ACO to realize the preliminary optimization for parameters of LSSVM.In the second phase,the particle swarm's particle positions were initialized with the first phase results and then the further optimization was implemented with PSO to acquire more accurate LSSVM.Employing this hybrid intelligent optimal LSSVM,the fluctuating wind velocity's time histories were predicted and compared with those using LSSVMwith ACO and LSSVM with PSO,respectively.The numerical results showed that the proposed method can promote the prediction accuracy and the robust of LSSVM,and has good engineering application prospects.