基于四分位数和Sigmoid改进模型的风电数据清洗方法  

Wind Power Data Cleaning Method Based on Quartiles and Sigmoid Improved Model

作　　者：宫永立 王玉超 刘志文 陆旭峰 骆可 GONG Yongli;WANG Yuchao;LIU Zhiwen;LU Xufeng;LUO Ke(CCCC HaiFeng New Energy Technologies(Shanwei)Co.,Ltd.,Shanwei 516600,China;TBEA Sunoasis Co,.Ltd.,Urumqi 830000,China)

机构地区：[1]中交海峰新能源科技(汕尾)有限公司,广东汕尾516600 [2]特变电工新疆新能源股份有限公司,新疆乌鲁木齐830000

出　　处：《电力科学与工程》2025年第3期55-62,共8页Electric Power Science and Engineering

基　　金：新疆维吾尔自治区重大科技专项基金资助项目(2022A01007-6)。

摘　　要：风电机组实际运行工况复杂,导致所采集的风功率数据中存在大量异常点,不利于功率曲线准确拟合。为解决该问题,分析了风速-功率散点的分布特征、所有异常点产生原因及分布情况,并在此基础上应用了四分位法对稀疏异常点进行剔除。针对结果中仍存在的异常簇,充分运用Sigmoid函数与功率曲线的相似性,建立含有四参数的Sigmoid改进模型,将四分位法和Sigmoid模型充分结合,以完成风功率数据的进一步清洗。该方法首先应用四分位法将大部分异常数据剔除,使算法执行速度快、精度高;其次,基于Sigmoid改进模型剔除剩余异常点,尤其是存在限功率的点,使算法具有较好的普适性。在存在所有异常数据点情况的风功率数据上验证了该方法的可行性。In the actual operation process of wind turbines,the working conditions are complex,which can lead to a large number of abnormal points in the wind power data,resulting in inaccurate fitting of the power curve.To solve this problem,the distribution characteristics of wind speed-power scatter points,the causes and distribution of all abnormal points are analyzed,and based on this,the quartile method is applied to remove sparse abnormal points.To address the remaining abnormal clusters in the results,the similarity between the sigmoid function and the power curve is fully utilized to establish an improved sigmoid model with four parameters,which fully combines the quartile method and sigmoid model to accomplish further cleaning of wind power data.The advantages of this method are:firstly,the quartile method is applied to remove most of the abnormal data,which makes the algorithm has the characteristics of fast execution speed and high accuracy.Then,based on the sigmoid improved model,it removes the remaining abnormal points,especially those with limited power,making the algorithm more universal.The feasibility of this method is demonstrated through practical application on wind power data with all abnormal data points present.

关 键 词：风功率数据 四分位法 Sigmoid改进模型 目标函数 高斯牛顿法 LM优化算法 数据清洗 

分 类 号：TM614[电气工程—电力系统及自动化]