中国电机工程学报Issue(23)：5985-5994,10.DOI:10.13334/j.0258-8013.pcsee.2015.23.004
区域风电场短期风电功率预测的最大相关-最小冗余数值天气预报特征选取策略
A Numerical Weather Prediction Feature Selection Approach Based on Minimal-redundancy-maximal-relevance Strategy for Short-term Regional Wind Power Prediction
摘要
Abstract
Regional wind power prediction is of great significance for large-scale wind power integration. Normally regional prediction error is less than that of an individual wind farm due to the spatial smoothing effect. In this paper, a feature selection method of minimal-redundancy-maximal-relevance (mRMR) based on mutual information theory was presented to select optimal feature subset from available numerical weather prediction (NWP) variables for regional wind power prediction. The selected optimal feature subset could maximize relevant information and minimize redundant information and noises from numerous original NWP datasets. An artificial neural network model was adopted to predict regional wind power to verify the validity of the selected optimal feature subset. As well, influence of subset cardinality on prediction accuracy was analyzed. Case study indicates that the optimal subset with a small number of features can not only improve regional prediction accuracy effectively than existing conventional methods,but also reduce computational cost and data-resource dependency significantly. In addition, the relationship between selected optimal feature subset and the spatial distribution of wind farms was investigated. Results show that the proposed approach can be used for improving the accuracy of the regional wind power prediction practically.关键词
风电功率/区域预测/互信息/特征选取/数值天气预报Key words
wind power/regional prediction/mutual information/feature selection/numerical weather prediction分类
信息技术与安全科学引用本文复制引用
赵永宁,叶林..区域风电场短期风电功率预测的最大相关-最小冗余数值天气预报特征选取策略[J].中国电机工程学报,2015,(23):5985-5994,10.基金项目
国家自然科学基金研究项目(51477174，51077126)。 Project Supported by National Natural Science Foundation of China (51477174,51077126) （51477174，51077126）
