作　　者： ; ;

机构地区： 广东工业大学自动化学院

出　　处： 《广东电力》 2013年第12期26-30,共5页

摘　　要： 针对风速序列非平稳变化的特性,首先通过经验模态分解(empirical mode decomposition,EMD)将原始风速序列分解为一系列较为平稳的子序列,再使用支持向量回归(support vector regression,SVR)模型分别对每一个子序列进行预测,为了克服SVR模型盲目选取学习参数的弊端,在和声搜索优化算法中加入了进化理论中优胜劣汰的思想,提出采用进化和声搜索(evolutional harmony search,EHS)算法对每一个SVR模型进行参数寻优。实例研究表明,EHS算法全局搜索能力强,收敛速度快,提出的EHS-EMD-SVR方法能有效提高短期风速预测的准确性。 Aiming at characteristics of unsteady changing wind speed sequence, this paper firstly studies method of decompo- sing original wind speed sequence into a series of comparatively steady subsequence by using empirical mode decomposition method. Then by making use of support vector regression model, it discusses prediction on each subsequence respectively. For overcoming disadvantage of selecting learning parameters by means of SVR model, it introduces survival of the fittest think- ing of evolutional theory in harmony search optimization algorithm and proposes to use evolutional harmony search algo- rithm to realize parameter optimizing for each SVR model. The example study indicates that EHS algorithm is provided with merits of strong global search ability and rapid convergence rate. In addition, the proposed EHS-EMD-SVR algorithm is ef- fective in improving veracity of predicting short-time wind speed.

关 键 词： 风力发电 风速预测 进化和声搜索 经验模态分解 支持向量回归

领　　域： [电气工程]