基于梯度下降原理的雷电流幅值累积概率分布曲线拟合方法OA

Lightning activity exhibits significant spatiotemporal variability,which necessitates the derivation of cumulative probability distribution(CPD)formula of lightning current magnitude(LCM)through statistical fitting as needed.The CPD formula of LCM recommended by IEEE is taken as the prototype,and a new fitting method based on the principle of gradient descent is designed and implemented via a Python program.This method establishes appropriate initial values and improves the iteration step size from a constant parameter to a constant vector tailored to the dimensionality of the parameter vector so as to accelerate the convergence process.The calculation of the cumulative probability statistical sample of nationwide cloud-to-ground LCM from 2005 to 2015 suggests that the statistical data for each year converges to the preset accuracy tolerance after 19 to 29 rounds of iterations,and that,compared with the existing nonlinear fitting methods in previous research,the proposed approach offers greater computational simplicity while yielding nearly identical final fitting results,which demonstrates its effectiveness and provides practical guiding significance for lightning protection engineering practice.