| 引用本文: | 谭风雷,陈昊,何嘉弘.基于K-means聚类方法和相似时段的特高压并联电抗器油温预测[J].电力自动化设备,2021,41(6): |
| TAN Fenglei,CHEN Hao,HE Jiahong.Oil temperature forecasting of UHV shunt reactor based on K-means clustering method and similar period[J].Electric Power Automation Equipment,2021,41(6): |
|
| 摘要: |
| 为及时评估特高压并联电抗器内部的绝缘状态，提出了一种特高压并联电抗器油温预测方法。首先基于特高压并联电抗器油温曲线的“单峰单谷”变化趋势，利用K-means聚类方法将油温曲线划分为低谷时段和高峰时段2个部分。然后分析了特高压并联电抗器油温的主要影响因素，并根据K-means聚类和时间“近大远小”原则，基于气象、油温和时间3个因素对油温历史样本进行3次分类后，选择得到了相似时段。最后详细给出了基于相似时段的油温预测方法的计算流程。将所提方法应用到江苏地区某特高压并联电抗器油温的预测工作中，结果表明所提方法的平均误差为1.03 %，最大误差为4.84 %，具有较高预测精度，从而验证了方法的有效性。 |
| 关键词: 特高压并联电抗器 油温预测 单峰单谷 K-means聚类 相似时段 |
| DOI：10.16081/j.epae.202103001 |
| 分类号:TM472 |
| 基金项目:国家自然科学基金资助项目(51807028)；江苏省自然科学基金资助项目(BK20170672)；国网江苏省电力有限公司科技项目(J2018014) |
|
| Oil temperature forecasting of UHV shunt reactor based on K-means clustering method and similar period |
|
TAN Fenglei1, CHEN Hao1, HE Jiahong2
|
|
1.Maintenance Branch Company of State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 211102, China;2.School of Electrical Engineering, Southeast University, Nanjing 210096, China
|
| Abstract: |
| A method of oil temperature forecasting of UHV(Ultra High Voltage) shunt reactor is proposed to evaluate the insulation state of UHV shunt reactor in time. Firstly, based on the “single peak single valley” changing trend of UHV shunt reactor’s oil temperature curve, the oil temperature curve is divided into two parts of valley period and peak period by using K-means clustering method. Then, the main influencing factors of the oil temperature of UHV shunt reactor are analyzed. According to K-means clustering and the time principle of “near big and far small”,the similar period is obtained by three classifications based on weather, oil temperature and time factors. Finally, the calculation process of the oil temperature forecasting method based on similar period is given in detail. The proposed method is applied to the oil temperature forecasting work of a UHV shunt reactor in Jiangsu area, the results show that the average error of the proposed method is 1.03 % and the max error is 4.84 %,which shows the high forecasting accuracy and effec-tiveness of the proposed method. |
| Key words: UHV shunt reactor oil temperature forecasting single peak single valley K-means clustering similar period |
