GRNN applied to study the lightning shielding performance for EHV&UHV transmission line

Lightning shielding failure is the main factor to threat the safety and reliability of EHV transmission lines. In this paper, the small-scale model experiments of lightning shielding performance of EHV transmission lines were made, taking the various influencing factors into account, including protection angle alpha, the gradient of the hillside theta, the height of the gully H, the conductor potential U, non-vertical leader angle beta and the position of lightning stroke point (x,y), showing the lightning stroke probability P. has the hidden non-linear function relation with the factors above. Due to the fact that the generalized regression neural network (GRNN) has several advantages over the back-propagation network (BP) such as very short training time and guaranteed performance even with sparse data, it is applied to simulate the complex non-linear function relationship between the lightning stroke probability P. and these various influencing factors. A lot of simulation was carried out in Matlab and the feasibility and accuracy of the proposed GRNN are verified.