基于ARIMA和指数平滑法的直流输电线路地面合成电场实测离群值的判别  

Discrimination of Measured Outliers of Ground Total Electric Field of DC Transmission Line Based on ARIMA and Exponential Smoothing Method

作　　者：谢辉春 倪园 张建功 王延召 张业茂 干喆渊 XIE Huichun;NI Yuan;ZHANG Jiangong;WANG Yanzhao;ZHANG Yemao;GAN Zheyuan(State Key Laboratory of Power Grid Environmental Protection,China Electric Power Research Institute,Wuhan 430074,China)

机构地区：[1]电网环境保护国家重点实验室(中国电力科学研究院有限公司),武汉430074

出　　处：《高电压技术》2023年第5期2161-2170,共10页High Voltage Engineering

基　　金：国家电网公司科技项目“±800 kV直流输电工程线路地面合成电场长期实测与预测对比分析研究”(GYW11201907738)。

摘　　要：直流输电线路地面合成电场是公众广泛关注的环境影响因子之一。合成电场极易受环境条件影响,实测结果波动较大,往往出现可疑异常值(离群值)。目前合成电场理论模型尚无法全面考虑环境影响,难以准确判别离群值。对此,从时间序列分析的角度,提出了基于自回归差分移动平均模型(auto regressive integrated moving average, ARIMA)和指数平滑法(exponential smoothing method, ETS)的离群值判别方法。经算例分析表明,该方法预测的均方根误差(root mean square error, RMSE)仅为1 kV/m左右,能全部判别算例中12处离群值,而经典数理统计方法仅能判别3处离群值。该方法可以有效处理相对平稳、具有固定趋势以及缓慢大幅波动等特征的数据序列,尤其是数据序列中有突然增大的非极端值离群值的情况。该方法比经典数理统计方法更适于判别合成电场实测结果的离群值。The ground total electric field of DC transmission line is one of the environmental impact factors of public concern.The total electric field is vulnerable to the environmental conditions,which leads to great fluctuation of measured results,and suspicious abnormal data(outliers)often appear.At present,the environmental impact is not fully taken into consideration in the theoretical model of total electric field,thus it is difficult to accurately discriminate those outliers.From the perspective of time series analysis,this paper proposes an outlier discrimination method based on auto regressive integrated moving average model(ARIMA)and exponential smoothing method(ETS).The numerical case analysis shows that the proposed method,by which the root mean square error(RMSE)of prediction is about 1 kV/m,can be adopted to discriminate all 12 outliers,while the classical mathematical statistics method can only be adopted to discriminate 3 outliers.This method can be adopted to effectively deal with the typical cases such as relatively stable data,fixed trend data and slowly but significantly fluctuant data series,especially the case of suddenly increased non-extreme value in data series.This method is more reasonable to discriminate outlier of measured results of total electric field than classical mathematical statistics.

关 键 词：直流输电线路 合成电场 时间序列 离群值 ARIMA 指数平滑法 

分 类 号：TM75[电气工程—电力系统及自动化]