基于四分位-MAD和朴素贝叶斯的风电数据处理方法

Wind power data processing method based on quartile-MAD and naive bayes

针对风电机组运行数据存在的异常和缺失问题,提出了一种基于四分位-MAD和朴素贝叶斯的风电数据处理方法.文中对实际风功率数据进行研究,分析风速-功率的分布特征,建立了四分位-MAD异常数据识别模型进行异常数据识别,并利用识别后的正常数据构造朴素贝叶斯分类器对剔除的异常数据和原始缺失数据进行补偿.仿真结果表明,所述异常识别方法能够有效识别不同分布类型的异常数据,采用所提补偿方法可以有效补偿数据且改善了数据完整性;经过上述方法处理后,风电功率的预测精度较数据处理前提高了17%,验证了所提方法的可行性和有效性,对风电场功率预测研究具有一定的现实应用价值.

A wind power data processing method based on quartile-MAD and naive Bayes is proposed to address the anomalies and missing problems in wind turbine operation data.Based on the actual wind power data,this paper analyzed the distribution characteristics of wind speed and power,established the quadrillion-MAD abnormal data identification model to identify abnormal data,The identified normal data is then used to construct a naive Bayes classifier,which compensates for excluded abnormal data and original missing data.The simulation results demostrate the effectiveness of the proposed method in identifying different distribution types of abnormal data,moreover,the proposed compensation method can effectively compensate the data and improves the data integrity.After applying the proposed method,the prediction accuracy of wind power increases by 17%compared to the accuracy before data processing.These results validate the feasibility and effectiveness of the proposed method and highlight its practical value in wind farm power prediction research.