以湖北巴东光伏电站实采数据为基础,分析了温度、辐照度、系统出力3类原始数据特征及变化特点,以融合准确率和召回率的综合指标为评价依据,提出采集数据的在线清洗与修复方法,最后以电站实采数据为例,进行数据清洗与修复的可行性验证。结果表明：根据趋势距离对变化趋势统一的温度数据进行清洗、通过设置动态阈值区间对辐照数据进行分时刻清洗,以及对独立分析困难但适宜联合分析的出力-辐照数据进行匹配度核算和清洗,均可较好地实现离群数据的筛选识别;此外,对残缺数据,根据相应清洗指标最小化原则,进行数据修复,可较好地再现缺失点的数据原貌。
Abstract
Based on the actual data collected in Hubei Badong photovoltaic power station, the characteristics and variation feature of three kinds of raw data, such as temperature, radiation and system output, are analyzed. Based on the comprehensive index of accuracy and recall rate, the on-line cleaning and repairing method of the corresponding data is proposed. Finally, the feasibility of data cleaning and repairing is validated by taking the actual data collected from power stations as an example. The results show that the temperature data with uniform variation trend can be cleaned according to the trend distance, the irradiation data can be cleaned in time by setting dynamic threshold interval. By Calculating and cleaning the matching degree of output-irradiation data, which is difficult to analyze independently but suitable for joint analysis, the outlier data can be screened and identified. In addition, the original data of missing points can be better reproduced by data repairing according to the principle of minimizing the corresponding cleaning index.
关键词
光伏系统 /
数据筛选 /
数据修复 /
数据联合匹配度
Key words
photovoltaic system /
data filtering /
data recovery /
data union matching degree
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 吴涛. 深度学习在光伏电站中的应用研究[D]. 西安: 西安热工研究院有限公司, 2019.
WU T.Research on the application of deep learning in photovoltaic power stations[D]. Xi’an: Xi’an Thermal Power Research Institute Co., Ltd., 2019.
[2] 王进, 余薇, 许景益, 等. 一种基于特征转换多标签学习的光伏发电大数据预测方法, 中国: CN110084412A[P].2019-08-02.
WANG J, YU W, XU J Y, et al. A photovoltaic power generationdata prediction method based on feature transformationand multi label learning, China: CN110084412A[P].2019-08-02.
[3] 戴巨川, 曹俊伟, 张帆, 等. 风电场SCADA数据预处理方法及评价策略[J]. 太阳能学报, 2017, 38(9): 2597-2604.
DAI J C, CAO J W, ZHANG F, et al.Data pre-processing method and its evaluation strategy of scada data from wind farm[J]. Acta energiae solaris sinica, 2017, 38(9): 2597-2604.
[4] 魏娅. 风光储能电站数据分析与评价模型研究[D]. 北京: 北方工业大学, 2016.
WEI Y.Research on data analysis and evaluating models of collective data from hybrid power stations of wind solar &energy-storage[D]. Beijing: NorthChina University of Technology, 2016.
[5] 包从剑. 数据清洗的若干关键技术研究[D]. 镇江: 江苏大学, 2007.
BAO C J.Some main technology’s research of data cleaning[D]. Zhenjiang: Jiangsu University, 2007.
[6] KLYUEV A L, DAVYDOV A D, GRAFOV B M.Measurementand Chebyshev spectral analysis of Brownian coordinate of electrochemical noise[J]. Measurement, 2020, 165: 108130.
[7] 刁赢龙, 盛万兴, 刘科研, 等. 大规模配电网负荷数据在线清洗与修复方法研究[J]. 电网技术, 2015, 39(11): 3134-3140.
DIAO Y L, SHENG W X, LIU K Y, et al.Research on online cleaning and repair methods of large-scale distribution network load data[J]. Power system technology, 2015, 39(11): 3134-3140.
[8] 张海宁, 张节潭, 杨立滨, 等. 一种用于光伏电站数据采集系统的数据处理方法[J]. 电器与能效管理技术, 2016(6): 8-13.
ZHANG H N, ZHANG J T, YANG L B, et al.Data processing method for PV power station data acquisition system[J]. Electrical & energy management technology, 2016(6): 8-13.
[9] 沈才良. 光伏电站故障监测系统的设计与实现[D]. 哈尔滨: 哈尔滨工业大学, 2015.
SHEN C L.Design and implementation of fault monitor system for PV power station[D]. Harbin: Harbin Institute of Technology, 2015.
[10] LIN J Q, LI L S, BI M H, et al.A study on performance improvement of IMDD-UFMC with modified K-means non-uniform quantization[J]. Optics communications, 2020, 476: 126324.
[11] 恽旻, 孙晓, 周滢, 等. 辐照度和温度对光伏组件光电转换性能测量的影响[J]. 现代测量与实验室管理, 2015, 23(4): 3-5,28.
YUN M, SUN X, ZHOU Y, et al.Influence of irradiance and temperature on measurement of photovoltaic module photoelectric conversion performance[J]. Advanced measurement and laboratory management, 2015, 23(4): 3-5,28.
[12] 邓建新, 单路宝, 贺德强, 等. 缺失数据的处理方法及其发展趋势[J]. 统计与决策, 2019, 35(23): 28-34.
DENG J X, SHAN L B, HE D Q, et al.Processing methods of missing data and its development trend[J]. Statistics & decision, 2019, 35(23): 28-34.
基金
国家重点研发计划（2018YFB1500800）; 国家电网有限公司科技项目（SGTJDK00DYJS2000148）; 中央科研基本业务费支持项目（PA2020GDGP0053）
PDF(2028 KB)
