Uncontrolled coal fires are natural disasters that may cause mineral loss and environmental damage． The traditional loop source transient electromagnetic method can effectively detect the low-resistivity region of coal fires，but its detection efficiency is not so good for high-resistivity regions． In view of this limitation，a technique based on electrical source transient electromagnetics is proposed in this paper to detect high-resistivity regions in the spontaneous combustion process of coal． Considering the complex geometry of the coal fire area，an unstructured tetrahedral grid is used in this study to realize the spatial discretization of the model，and solve the electromagnetic field based on a vector finite element algorithm． Numerical analysis is used to investigate methods for detecting coal fires and the characteristics of effective anomalies are further examined to provide guidance for practical detection．