Data-driven Accurate Verification Method for Short-circuit Current Over-limited Based on Critical Over-limited Samples Expansion
In the verification of short-circuit current over-limited,aiming at the weaknesses of insufficient accuracy and easy to misjudge in existing methods in critical over-limited scenarios,this paper proposes a data-driven accurate verification method for short-circuit current over-limited based on critical over-limited samples expansion.Firstly,the Generative Adversarial Networks(GAN)is adopted in the short-circuit current field for the first time to generate many samples with the same effects as the Monte Carlo simulation.Then,the critical over-limited samples are filtered.Secondly,a new method of sample composition is designed,which has the characteristic of a high proportion of critical over-limited and a low proportion of non-critical over-limited.Based on this,the data-driven sample set is formed by blending GAN critical over-limited and Monte Carlo simulation samples.Then,LightGBM,the representative regression algorithm in data-driven methods,is used to verify short-circuit current over-limitations.Finally,the simulation results show that the proposed method can effectively improve the verification accuracy of short-circuit current over-limitations.Compared to other physical computing and data-driven methods,it has higher efficiency,computational accuracy,and faster calculation speed.
short-circuit current over-limited verificationcritical over-limited samples expansionGANdata-drivenLightGBM
万方数据
维普
