• Assessment of electrical quantity fault degree by current amplitude estimation. • Assessment of Switching quantity fault degree by corrected Bayesian network model. • An improved fusion algorithm is proposed to fuse multi-source fault degrees. When short circuit fault occur in the direct current (DC) distribution network, in view of the problems such as inaccurate representation of fault degrees, low location accuracy and location errors in existing fault location methods, this paper proposes a fault section location method based on multi-source fault information fusion during the fault recovery period in complex topology DC distribution network. Firstly, the high-frequency fault current sparse vector is extracted based on the compressed sensing reconstruction algorithm to preliminarily identify the fault range. The nonlinear attenuation characteristics of high-frequency fault current amplitudes are analyzed, and the normal distribution models are constructed to quantitatively characterize the attenuation trend to evaluate the electrical quantity fault degree. Secondly, the local switching quantity operation information of DC distribution network is determined based on the precise identification of high-frequency voltage signal generated at the moment of circuit breaker trip, and the erroneous state information of elements in Bayesian network models are effectively corrected to enhance the accuracy of switching quantity fault degree. Then, the traditional Dempster-Shafer evidence theory (D-S evidence theory) is improved from the perspective of the weight of evidence body, and based on the improved D-S evidence theory, the fault degrees are fused to achieve accurate fault section location. Finally, the simulation model of DC distribution network based on Power Systems Computer Aided Design (PSCAD) platform verifies the validity and superiority of the fault section location method proposed in this paper.
Wang et al. (Wed,) studied this question.