Abstract Time difference of arrival (TDOA) technique is one of many bases to determine lightning strike location employed in a lightning locating system (LLS). In this technique, at least four measurement sensors are required to correctly locate a lightning strike. Usage of fewer number of sensors will result in non-unique solutions to the generated hyperbolas, and hence wrong lightning strike point. This research aims to correctly determine the strike point even if only three measuring sensors are utilized. An artificial neural network (ANN) based algorithm was developed for a 400 km2 coverage area in Southern Malaysia using time of arrival data collected at the three measuring stations over a certain period. The Levenberg–Marquardt algorithm is demonstrated to correctly identify the lightning strike coordinates with an average error of 350 m. The algorithm has helped the three-station TDOA-based LLS to successfully locate the lightning strike point with a remarkable accuracy comparable to that of commercial systems.

中文翻译：

---

人工神经网络在闪电定位系统中的应用

摘要 到达时间差（TDOA）技术是雷击定位系统（LLS）中确定雷击位置的众多依据之一。在这种技术中，至少需要四个测量传感器才能正确定位雷击。使用较少数量的传感器将导致生成的双曲线的非唯一解，从而导致错误的雷击点。这项研究旨在即使仅使用三个测量传感器也能正确确定打击点。一种基于人工神经网络 (ANN) 的算法是针对马来西亚南部 400 平方公里的覆盖区域开发的，使用在一定时期内在三个测量站收集的到达时间数据。Levenberg-Marquardt 算法被证明可以正确识别雷击坐标，平均误差为 350 m。