电网雷电监测数据统计与挖掘分析  

Statistical and Mining Analysis of Lightning Detection Data in Power Grid

作　　者：谷山强[1,2] 王剑[3] 冯万兴[1,2] 王佩[1,2] 王海涛[1,2] 郭钧天[1,2] 

机构地区：[1]南瑞集团公司(国网电力科学研究院),南京211106 [2]国网电力科学研究院武汉南瑞有限责任公司,武汉430074 [3]国家电网公司,北京100031

出　　处：《高电压技术》2016年第11期3383-3391,共9页High Voltage Engineering

基　　金：国家电网公司科技项目(WHJ151-0049)~~

摘　　要：电网雷电监测网已积累了大量雷电数据,为研究和获取雷电特征,采用统计分析、数据挖掘并结合近年来大数据分析方法,对2005—2014年10年雷电监测数据进行了分析。分析结果表明：全国10年合计落雷1.28亿次,正极性比率呈现总体上升趋势,平均为7.50%;各区域雷电活动差异较大,总数南方最高,地闪密度华东最高,全国平均地闪密度为1.33次/（km2·a）;雷电流幅值10年分布规律比较接近,累积概率分布拟合曲线a、b值为31.33 k A和2.76,与国际上IEEE和CIGRE推荐取值非常接近。对武汉和杭州两城雷电与降水量相关性分析表明：2者高峰期大体一致,月分布除个别年份外呈现中强相关性;两城雷电正极性比率与降水量相关性一致,除个别年份外呈现弱相关性;平均电流与降水量武汉为弱相关性、杭州除个别年份外也为弱相关性。基于雷电监测数据,融入大气电场、卫星云图、气象雷达数据,采用数据挖掘技术预测雷暴团运动趋势实现雷电预警。具体实例表明,预警提前时间为10～30 min,预警平均空间误差为5～15 km,预警有效率达75%～85%。The Lightning Detection Network of China Power Grid has been accumulating a large amount of lightning data. In order to study and obtain the lightning characteristics, we applied the statistics and mining analysis methods as well as the big data analysis methods in recent years to analyze 10 years of lightning data from 2005 to 2014. Analysis results show that the lightning amount in 10 years is about 128 million, and the positive ratio shows a general upward trend with an average value of 7.50%. Lightning activities in all areas are different where the amount in the South China and the cloud-ground flash density （GFD） in the East China are the highest ,respectively, and the national average GFD is 1.33 flashes/（km^2·a）. The lightning current amplitude has a similar distribution regular in each year, and a, b of the fitting curve of cumulative probability is 31.33 kA, 2.76, which is very close to the recommended values of IEEE and CIGRE. Moreo- ver, we discussed the relationship between lightning and precipitation of Wuhan and Hangzhou and found that the lighting activities in the years from 2005 to 2008 have the generally same peak as the precipitation, and their monthly distribution shows a moderately strong correlation except for a few years. There are no significant correlations between the positive rate of lightning and the precipitation except for a few years. The correlations between the average lightning current am- plitude and the precipitation of the tow cities have good consistence, and are both very weak except individual year in Hangzhou. On the basis of the lightning detection data, integrating the atmospheric electric field, satellite cloud image, and meteorologic radar data, we applied data mining technology to realize the lightning warning through forecasting the movement trend of lightning clusters. The specific examples show that the lead time of lightning warning is about 10 to 30 minutes, the average space error of warning target area is about 5 to 15 km, and the warning effective ra

关 键 词：雷电监测 监测数据 数据挖掘 地闪密度 雷电流 累积概率 降水量 雷电预警 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TM863[自动化与计算机技术—计算机科学与技术]