Wildfires, also known as bushfires, happened more and more frequently in the last decades. Especially in countries like Australia, the dry and warm climate there make bushfire become one of the most frequent local hazards. Among different kinds of causes of bushfires, overhead powerline vegetation fault is one of the most common causes that relate to human activities. Reducing the bushfire risk from this perspective has attracted many scholars to study efficient strategies and systems. However, most of them started their research from the angle of powerline faults, while limited literature has explored the characteristics of the vegetations and their ignition features. The objective of this study is to explore and discover the numerical patterns from the contact to the ignition process between different upper story vegetations and the powerlines. Those patterns can not only help provide real-time warnings of bushfire caused by powerline vegetation faults but also avoid false alarm. To achieve this, we collected the voltage and current records of 188 ignition field tests that simulated the powerline vegetation faults. To explore the numerical patterns behind and develop a real-time alarming system, this study proposed a machine learning-based model, namely Hybrid Step XGBoost. According to the tests, the model could identify the safe contacts or the danger contacts between the powerlines and the upper story vegetation with an accuracy of 98.17%. Its performance also surpassed some advanced deep learning networks in our experiments.

在过去的几十年中，野火（又称丛林大火）发生的频率越来越高。特别是在像澳大利亚这样的国家中，那里干燥干燥的气候使丛林大火成为当地最常见的危害之一。在造成丛林大火的各种原因中，架空电力线植被断裂是与人类活动有关的最常见原因之一。从这个角度降低林区大火的风险吸引了许多学者来研究有效的策略和系统。然而，大多数人是从电力线故障的角度开始研究的，而有限的文献已经探索了植被的特征及其着火特征。这项研究的目的是探索和发现不同高层植被和电力线之间从接触到点火过程的数值模式。这些模式不仅有助于提供由电力线植被故障引起的林区大火的实时警告，而且还可以避免误报。为了达到这个目的，我们收集了188次点火现场测试的电压和电流记录，这些仿真模拟了电力线植被的故障。为了探索背后的数值模式并开发实时警报系统，本研究提出了一种基于机器学习的模型，即Hybrid Step XGBoost。根据测试，该模型可以识别电力线与高层植被之间的安全接触或危险接触，准确度为98.17％。在我们的实验中，它的性能也超过了一些高级深度学习网络。我们收集了188次点火现场测试的电压和电流记录，以模拟电力线植被的故障。为了探索背后的数值模式并开发实时警报系统，本研究提出了一种基于机器学习的模型，即Hybrid Step XGBoost。根据测试，该模型可以识别电力线与高层植被之间的安全接触或危险接触，准确度为98.17％。在我们的实验中，它的性能也超过了一些高级深度学习网络。我们收集了188次点火现场测试的电压和电流记录，以模拟电力线植被的故障。为了探索背后的数值模式并开发实时警报系统，本研究提出了一种基于机器学习的模型，即Hybrid Step XGBoost。根据测试，该模型可以识别电力线与高层植被之间的安全接触或危险接触，准确度为98.17％。在我们的实验中，它的性能也超过了一些高级深度学习网络。该模型可以识别电力线与高层植被之间的安全接触或危险接触，准确度为98.17％。在我们的实验中，它的性能也超过了一些高级深度学习网络。该模型可以识别电力线与高层植被之间的安全接触或危险接触，准确度为98.17％。在我们的实验中，它的性能也超过了一些高级深度学习网络。