Fault location method based on deep learning in new power system

Abstract

Due to the large amount of uncertain energy, the new power system makes the power grid status data more nonlinear and non-stationary, which poses a great challenge to the fault location of protection devices. In this paper, empirical mode decomposition algorithm and deep network model are combined to realize accurate fault location analysis and judgment for complex power system. The introduction of permutation entropy function can realize the processing of set empirical mode decomposition algorithm and optimize the sensitive components that can best represent the characteristics of fault signals. The state data of power system has obvious time characteristics. Using long-term and short-term memory network for fault location analysis can effectively extract the effective information in the fault characteristic data and achieve accurate and efficient fault location analysis. The results prove that the PE-CEEMDAN-LSTM method can obviously realize the accurate fault analysis of complex power system, which the maximum error distance is only 15 m.