Emilia Sobieska
supervisor: Jacek Starzyński, Konrad Sobolewski
The procedure for designing lightning protection systems defined by the PN-EN 62305 standard is characterized by some imperfections and limitations. When determining the lightning risk, the values of some coefficients describing the lightning threat, are selected on the basis of the subjective assessment of the designer. As a result, there are cases of overestimating or underestimating the parameters of the lightning protection system. The process of estimating this risk may be improved by the use of artificial neural networks to correctly classify the values of the aforementioned coefficients on the basis of input data, such as photos of the object and its vicinity. Neural networks construct the models needed by the user using the learning process based on examples. Such action is aimed at automatically creating the necessary data structure in the memory. The network, based on a self-created data structure, performs all the functions related to the use of the created model after completing the learning process.
In the scope of the research work, it is planned to perform an analysis to verify the possibility of implementing at a certain level of automation of relevant calculations with the use of a neural network and to conduct the process of inference about the optimal type of external lightning protection. The tests will be carried out for such facilities as, for example, a rectangular building, a free-standing photovoltaic power plant and a high voltage power line.