LIGHTNING LOCAL DATA: ENHANCED LIGHTNING PERFORMANCE ASSESSMENT OF OVERHEAD DISTRIBUTION LINES AND ALTERNATIVE DETECTION METHOD BASED ON MACHINE LEARNING

Lightning is a complex natural phenomenon with significant risk implications. It can cause injuries to humans and animals and substantial damage to power lines, telecommunication systems, and structures, especially if they are tall or isolated such as wind turbines. Despite a growing research interest, the full understanding of lightning properties remains uncomplete due to its inherent complexity. This thesis explores the potential of using Lightning Location System (LLS) data to locally characterize lightning properties and enhance the evaluation of the lightning performance on medium-voltage overhead distribution lines. Two supporting tasks are included: i) processing 10-year data recorded by a LLS operating over the Italian territory to derive distributions for the main lightning parameters and ii) developing an efficient computational method to simulate lightning-induced overvoltages. Additionally, this thesis proposes an innovative approach using Machine Learning (ML) to estimate lightning peak currents and locations starting from the observation of lightning-induced overvoltages on transmission and distribution lines, offering a cost-effective and promising alternative to conventional lightning measurements methods. In this thesis it is proved how locally sourced lightning data can provide a more actual and accurate assessment of lightning risk, addressing limitations of conventional data sources. By integrating atmospheric insights and ML tools with Electro Magnetic Compatibility (EMC) requirements, the research contributes to improved risk mitigation strategies and advancements in lightning-related technologies.