计及历史数据熵关联信息挖掘的短期风电功率预测
发布时间:2018-03-11 10:43
本文选题:关联信息挖掘 切入点:熵相关系数 出处:《电力系统自动化》2017年03期 论文类型:期刊论文
【摘要】:对风电功率历史数据进行关联信息挖掘,将有助于提高短期风电功率预测的准确度和计算效率。为解决风电功率预测模型的输入、输出变量的相关性冗余问题,尝试采用了一种基于信息熵和互信息的熵相关系数指标,旨在量化评估不同历史日风电样本与待预测日参考样本间的复杂非线性映射关系,并与线性相关系数、秩相关系数、欧氏距离指标进行了对比研究。同时,设计了一种BP神经网络改进模型,通过亲密样本筛选、隐含层结构寻优、网络权重赋初值等环节,克服了传统预测模型的训练数据冗余度大、收敛速度慢问题,提高了预测模型的泛化能力和计算效率。对某风电场实测数据的算例分析表明,所提出的方法在改善短期风电功率预测性能方面具有应用可行性。
[Abstract]:The association information mining of wind power historical data will help to improve the accuracy and computational efficiency of short-term wind power prediction. In order to solve the problem of input and output redundancy of wind power prediction model, An index of entropy correlation coefficient based on information entropy and mutual information is used to quantitatively evaluate the complex nonlinear mapping relationship between wind power samples and reference samples to be forecasted on different historical days, and to be related to linear correlation coefficient, rank correlation coefficient, linear correlation coefficient, rank correlation coefficient, linear correlation coefficient, rank correlation coefficient, linear correlation coefficient and rank correlation coefficient. The Euclidean distance index is compared and studied. At the same time, an improved BP neural network model is designed. Through the close sample selection, the hidden layer structure is optimized, the network weight is assigned initial value, and so on. It overcomes the problems of large redundancy of training data and slow convergence speed of the traditional prediction model, and improves the generalization ability and computational efficiency of the prediction model. The proposed method is feasible in improving the prediction performance of short-term wind power.
【作者单位】: 清华大学电机工程与应用电子技术系;电力系统及发电设备控制和仿真国家重点实验室清华大学;国网吉林省电力有限公司;
【基金】:国家自然科学基金资助项目(51077078) 国家科技支撑计划资助项目(2015BAA01B01)~~
【分类号】:TM614
,
本文编号:1597816
本文链接:https://www.wllwen.com/kejilunwen/dianlidianqilunwen/1597816.html