为了提高雷电定位精度与算法模型性能,在定位计算中引入了改进的聚类分析算法,评估了不同算法在国家雷电监测网中的性能。仿真结果表明,具有噪声的基于自适应密度的空间聚类(ADBSCAN)能根据实际定位结果间聚合程度自适应确定邻域参数,实际应用效果最好。最后提出了一套基于改进ADBSCAN聚类分析与网格搜索优化的雷电定位算法(ADGLLA)。在相同回击数量的条件下,ADG-LLA相较于国家雷电监测网得到了更多的定位结果,闪电回击数据利用率从42.78%提高至56.37%,且定位效果较好,精确度更高,平均定位误差也比国家雷电监测网降低了30.78%。文中设计的雷电定位算法能有效识别噪声数据,克服了传统迭代算法计算量大且易于陷入局部最优等缺点,可稳定并精确求解出雷击点。

In order to improve the lightning location accuracy and the performance of the algorithm model, the improved clustering analysis algorithms are introduced in the location calculation, and the performances of different algorithms in the national lightning monitoring network are evaluated. The simulation results show that the adaptive-density-based spatial clustering of applications with noise (ADBSCAN) can adaptively determine the neighborhood parameters according to the degree of aggregation between the actual location results, and the actual application effect is the best. Finally, the ADBSCAN and grid-search lighting location algorithm (ADG-LLA) is proposed. Under the same number of return attacks, ADG-LLA can obtain more location results than the national lightning monitoring network. The utilization rate of lightning return data has increased from 42.78% to 56.37%, and the location effect is better, the accuracy is higher, and the average location error is also 30.78% lower than the national lightning monitoring network. The lightning location algorithm designed in this paper can effectively identify the noise data, overcome the shortcomings of the traditional iterative algorithm such as large calculation amount and easy to fall into the local optimum, and calculate the lightning strike point stably and accurately.

P427. 32

国家自然科学基金资助项目(61806139);国家重点基础研究发展计划资助项目(2020YFC1506002);青海省科技厅创新平台建设专项资助项目(2019-ZJT03)