基于空间密度聚类与LS-SVM的雷云预测方法

摘要

Lightning threatens the stability operation of power system.Based on the data of lightning location system,thunderstorm cloud tracking and predicting were realized in this paper.Firstly,the data of cloud-to-ground flash was counted by using equal-area grid,and then they were clustered into thunderstorm cloud cluster by Density-based spatial clustering of application with noise algorithm (DBSCAN).And the cluster centroid could be calculated.The time and location of thunderstorm cloud were integrated for the same dimension as input.Finally,the nonlinear moving tracking of thunderstorm cloud centroid was predicted by least squares support vector machine method (LV-SVM),thus the thunderstorm cloud was followed and predicted.Taking data of a certain power grid lighting location system as a simulation example,the results show that the proposed algorithm is with high accuracy in practical.Comparing with the linear extrapolated method and BP neutral network method,it has the practical application value.%雷电威胁着电力系统稳定运行,以雷电定位监测系统数据为基础,实现雷暴云团跟踪与预测.首先,利用网格化方法统计落雷数据;然后,利用空间密度聚类将落雷点聚类为雷暴云团,确定雷云团的重心位置;将雷云出现的时间与地理位置作为输入量整合为同一量纲;最后,利用最小二乘支持向量机算法,动态预测雷云重心的非线性运动轨迹,从而实现雷云的跟踪与预测.以某供电局监测系统数据为例进行仿真分析,结果表明:该算法预测结果准确,与现有的线性外推及BP神经网络算法比较,更具实际应用价值.