Study on Localized Lightning Strike Prediction Using Machine Learning: Validation Using the Vector Autoregression Model

In Japan, a significant number of lightning strikes occurs annually, with the number sometimes exceeding one million in a single month. Therefore, there have been numerous cases of human casualties, including fatalities and incidents where individuals suffer from loss of consciousness due to lightning strikes. According to the White Paper on Police from 1994 to 2003, approximately 100 people were affected by lightning strikes during this period, with over 30 of them resulting in fatalities. That is to say, approximately 30 % of lightning strike incidents result in fatalities. Lightning strikes induced damage occurs across various settings, including wind and solar power generation facilities, golf courses, fields, and maritime environments. As an example, when lightning strikes a wind turbine, one of the most significant accidents affecting the operation is “blade damage.” Among the cases of blade damage, there have been cases where parts of a rotating blade, after being struck by lightning, have scattered around the surrounding area, falling near roads and residential areas. Therefore, it is crucial to accurately predict the risk of lightning strikes and take preventive measures to mitigate damage in advance. This paper provides a brief overview of the development of the localized meteorological data collection system to measure weather conditions around Chubu University. Additionally, the paper reports on the construction of a system that predicts lightning strike risk using data obtained from the localized meteorological data collection system and machine learning techniques.