基于规则库与PRRL模型的风电功率数据清洗方法
WIND POWER DATA CLEANING METHOD BASED ON RULE BASE AND PRRL MODEL
杨海能 1唐杰 2邵武 3刘白杨 3陈日恒3
作者信息
- 1. 邵阳学院机械与能源工程学院,邵阳 422000;多电源地区电网运行与控制湖南省重点实验室(邵阳学院),邵阳 422000;邵阳职业技术学院汽车与智能制造学院,邵阳 422000
- 2. 邵阳学院机械与能源工程学院,邵阳 422000;多电源地区电网运行与控制湖南省重点实验室(邵阳学院),邵阳 422000
- 3. 多电源地区电网运行与控制湖南省重点实验室(邵阳学院),邵阳 422000
- 折叠
摘要
为提升风电场原始数据中异常数据的识别精度,提出一种结合规则库与PRRL模型的风电场数据清洗方法.首先依据风电场装机容量等参数建立规则库,提高数据集中正常数据占比.其次,以RANSAC稳健回归算法为核心,线性回归模型为基础,将风速数据作为输出,同时扩展输入变量的多项式特征来捕捉风速与功率之间的非线性关系,构建PRRL稳健回归模型.该模型经过规则库处理后的数据训练,可有效识别原始数据中的异常数据.通过对湖南某风电场的实例数据进行应用测试,结果显示该方法在处理异常数据占比较高数据时,能有效识别其中的异常数据,并降低风电功率预测模型的预测误差.
Abstract
In order to improve the identification accuracy of abnormal data in the original data of wind farms,a wind farm data cleaning method combining a rule base and the PRRL model is proposed.First,a rule base is established based on parameters such as the installed capacity of the wind farm to increase the proportion of normal data in the dataset.Second,the PRRL robust regression model is constructed using the RANSAC robust regression algorithm as the core and the linear regression model as the foundation,taking wind speed data as the output and expanding the polynomial features of input variables to capture the nonlinear relationship between wind speed and power.The model is trained on data processed by the rule base,effectively identifying abnormal data in the original dataset.Application testing using sample data from a wind farm in Hunan shows that this method can effectively identify abnormal data when processing the data with a high proportion of abnormal data,and reduce the prediction error of the wind power forecasting model.
关键词
风电场/数据清洗/异常检测/规则库/回归分析/稳健回归分析Key words
wind farm/data cleaning/anomaly detection/rule base/regression analysis/robust regression analysis引用本文复制引用
出版年
2024
NETL
NSTL
万方数据