Arcgrounding fault is the most common type of single-phase grounding fault in medium-voltage distribution networks. The research on the Arc grounding fault in medium-voltage distribution networks is of great significance for fault detection and fire prevention. First, This paper proposes an improved” control theory model” based on the classical arc model. Then, to overcome the shortcoming that the improved “control theory model” can not accurately model when the fault grounding medium is different, an adaptive parametric arc modeling method is proposed. According to the arc data of different grounding media, the normalized standard model of arc classification is established, which can identify the grounding media according to the actual fault data. Finally, using MATLAB-PSCAD co-simulation, the normalized model with determinable classification is taken as the target model. PSO parameter optimization algorithm is used to determine the optimal parameters of the arc model in PSCAD. The comparison with the measured data proves the validity and correctness of the model.