基于聚类经验模态分解-样本熵和优化极限学习机的风电功率多步区间预测

发布时间：2018-04-18 00:05

本文选题：多步区间预测 + 聚类经验模态分解?样本熵　；参考：《电网技术》2016年07期

【摘要】：针对风电功率序列的不确定性和随机性特征,提出一种基于聚类经验模态分解-样本熵和优化极限学习机的多步区间预测模型。首先,利用聚类经验模态分解-样本熵方法将原始风电功率序列分解为一系列复杂度差异明显的子序列。然后,分别对各子序列建立基于上下界直接估量的区间预测模型。为分析不同区间构造的差异,提出一种体现训练目标值偏离区间范围影响的新型区间预测评估指标作为目标函数,并采用基于混沌萤火虫结合多策略融合自适应差分进化的优化算法寻求其最优解,以提高模型预测性能。最后,以某一风电场实际功率数据为算例,验证了所提模型能获得可靠优良的多步区间预测结果,可为风电功率多步不确定性预测提供一种新的有效途径。
[Abstract]:In view of the uncertainty and randomness of wind power series, a multi-step interval prediction model based on clustering empirical mode decomposition-sample entropy and optimal extreme learning machine is proposed.Firstly, the original wind power series is decomposed into a series of sub-sequences with obvious difference in complexity by cluster empirical mode decomposition-sample entropy method.Then, the interval prediction models based on the upper and lower bounds are established for each sub-sequence.In order to analyze the differences of different interval structures, a new type of interval prediction evaluation index is proposed as the objective function, which reflects the effect of the training target value deviating from the range of the interval.The optimization algorithm based on chaos firefly and multi-strategy fusion adaptive differential evolution is used to find the optimal solution to improve the prediction performance of the model.Finally, taking the actual power data of a wind farm as an example, it is verified that the proposed model can obtain reliable and excellent multi-step interval prediction results, and can provide a new effective way for wind power multi-step uncertainty prediction.
【作者单位】：武汉大学电气工程学院;国网湖北省电力公司经济技术研究院;
【基金】：国家重点基础研究发展计划项目(973项目)(2012CB215101) 国家自然科学基金项目(51309258)~~
【分类号】：TM614

【相似文献】